enigmare/v2-crawler
1904
1{"id":"doc-cohere_documentation_cohere-a1ca1d46","source":"documentation","title":"Cohere Documentation | Cohere","url":"https://docs.cohere.com/","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.268Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}2{"id":"doc-welcome_to_cohere_cohere-42837e03","source":"documentation","title":"Welcome to Cohere | Cohere","url":"https://docs.cohere.com/docs/welcome","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.268Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}3{"id":"doc-welcome_to_cohere_cohere-a1404ca8","source":"documentation","title":"Welcome to Cohere | Cohere","url":"https://docs.cohere.com/docs","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.268Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}4{"id":"doc-cookbooks_cohere-650eb98a","source":"documentation","title":"Cookbooks | Cohere","url":"https://docs.cohere.com/page/cookbooks","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.269Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}5{"id":"doc-an_overview_of_the_developer_playground_cohere-76e48a53","source":"documentation","title":"An Overview of the Developer Playground | Cohere","url":"https://docs.cohere.com/docs/playground-overview","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.269Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}6{"id":"doc-chat_cohere-5b4fdaa3","source":"documentation","title":"Chat | Cohere","url":"https://docs.cohere.com/reference/chat","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nGenerates a text response to a user message and streams it down, token by token. To learn how to use the Chat API with streaming follow our Text Generation guides. Follow the Migration Guide for instructions on moving from API v1 to API v2.\n\nThe name of a compatible Cohere model.\n\nA list of chat messages in chronological order, representing a conversation between the user and the model. Messages can be from User, Assistant, Tool and System roles. Learn more about messages and roles in the Chat API guide.\n\nA list of tools (functions) available to the model. The model response may contain ‘tool_calls’ to the specified tools. Learn more in the Tool Use guide.\n\nConfiguration for forcing the model output to adhere to the specified format. Supported on Command R, Command R+ and newer models. The model can be forced into outputting JSON objects by setting { \"type\": \"json_object\" }. A JSON Schema can optionally be provided, to ensure a specific structure. using { \"type\": \"json_object\" } your message should always explicitly instruct the model to generate a JSON (eg: “Generate a JSON …”) . Otherwise the model may end up getting stuck generating an infinite stream of characters and eventually run out of context length. json_schema is not specified, the generated object can have up to 5 layers of nesting. parameter is not supported when used in combinations with the documents or tools parameters.\n\nUsed to select the safety instruction inserted into the prompt. Defaults to CONTEXTUAL. When OFF is specified, the safety instruction will be omitted. Safety modes are not yet configurable in combination with tools and documents parameters. parameter is only compatible newer Cohere models, starting with Command R 08-2024 and Command R+ 08-2024. and newer models only support \"CONTEXTUAL\" and \"STRICT\" modes.\n\nThe maximum number of output tokens the model will generate in the response. If not set, max_tokens defaults to the model’s maximum output token limit. You can find the maximum output token limits for each model in the model documentation. a low value may result in incomplete generations. In such cases, the finish_reason field in the response will be set to \"MAX_TOKENS\". max_tokens is set higher than the model’s maximum output token limit, the generation will be capped at that model-specific maximum limit.\n\nUsed to control whether or not the model will be forced to use a tool when answering. When REQUIRED is specified, the model will be forced to use at least one of the user-defined tools, and the tools parameter must be passed in the request. When NONE is specified, the model will be forced not to use one of the specified tools, and give a direct response. If tool_choice isn’t specified, then the model is free to choose whether to use the specified tools or not. parameter is only compatible with models Command-r7b and newer.\n\nConfiguration for reasoning features.\n\nWhen set to true, tool calls in the Assistant message will be forced to follow the tool definition strictly. Learn more in the Structured Outputs (Tools) guide. first few requests with a new set of tools will take longer to process.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[{\"role\": \"user\", \"content\": \"Tell me about LLMs\"}],8)910print(response)\n```\n\nExample:\n```text\n1{2 \"id\": \"c14c80c3-18eb-4519-9460-6c92edd8cfb4\",3 \"finish_reason\": \"COMPLETE\",4 \"message\": {5 \"role\": \"assistant\",6 \"content\": [7 {8 \"type\": \"text\",9 \"text\": \"LLMs stand for Large Language Models, which are a type of neural network model specialized in processing and generating human language. They are designed to understand and respond to natural language input and have become increasingly popular and valuable in recent years.\\n\\nLLMs are trained on vast amounts of text data, enabling them to learn patterns, grammar, and semantic meanings present in the language. These models can then be used for various natural language processing tasks, such as text generation, summarization, question answering, machine translation, sentiment analysis, and even some aspects of natural language understanding.\\n\\nSome well-known examples of LLMs include:\\n\\n1. GPT-3 (Generative Pre-trained Transformer 3) — An open-source LLM developed by OpenAI, capable of generating human-like text and performing various language tasks.\\n\\n2. BERT (Bidirectional Encoder Representations from Transformers) — A Google-developed LLM that is particularly good at understanding contextual relationships in text, and is widely used for natural language understanding tasks like sentiment analysis and named entity recognition.\\n\\n3. T5 (Text-to-Text Transfer Transformer) — Also from Google, T5 is a flexible LLM that frames all language tasks as text-to-text problems, where the model learns to generate output text based on input text prompts.\\n\\n4. RoBERTa (Robustly Optimized BERT Approach) — A variant of BERT that uses additional training techniques to improve performance.\\n\\n5. DeBERTa (Decoding-enhanced BERT with disentangled attention) — Another variant of BERT that introduces a new attention mechanism.\\n\\nLLMs have become increasingly powerful and larger in scale, improving the accuracy and sophistication of language tasks. They are also being used as a foundation for developing various applications, including chatbots, content recommendation systems, language translation services, and more.\\nThe future of LLMs holds the potential for even more sophisticated language technologies, with ongoing research and development focused on enhancing their capabilities, improving efficiency, and exploring their applications in various domains.\"10 }11 ]12 },13 \"usage\": {14 \"billed_units\": {15 \"input_tokens\": 5,16 \"output_tokens\": 41817 },18 \"tokens\": {19 \"input_tokens\": 71,20 \"output_tokens\": 41821 }22 }23}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.270Z","totalSectionsIncluded":11,"totalCodeBlocksIncluded":2,"totalLines":33,"estimatedTokens":1530}}7{"id":"doc-release_notes_cohere-bd8460f3","source":"documentation","title":"Release Notes | Cohere","url":"https://docs.cohere.com/changelog","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nMeet Cohere Transcribe ArabicToday we are releasing Cohere Transcribe Arabic. This open-source speech-to-text model is a fine-tune of Cohere Transcribe using Arabic speech data. It lets Arabic speakers transcribe their voice with unmatched accuracy and support for regional dialects or speech patterns. It is currently the most accurate open-source Arabic ASR model available today and is optimized for production inference and throughput. Technical Details Model encoder-decoder Languages (all major dialects), English (including English spoken with an Arabic accent) 2.0 Availability Cohere Transcribe Arabic is available through the V2 Audio Transcriptions API and as open weights on Hugging Face. For production use, Model Vault deployment is also supported. For more details, see the model documentation.\n\nAnnouncing Cohere's North Mini CodeWe’re pleased to announce the release of North Mini Code, Cohere’s first agentic coding model. It is a 30 billion total / 3 billion active parameter Mixture of Experts model trained specifically for agentic coding, with a small enough active footprint to run on local hardware. Technical Details Model Context input, 64K output 2.0 Availability North Mini Code is available through the Chat V2 API and as open weights on Hugging Face. For production use, Model Vault deployment is also supported. For more details, see the model documentation.\n\nAnnouncing Cohere’s Command A+We’re pleased to announce the release of Command A+, the last model in the Command A family of models, combining support for vision inputs, reasoning capabilities, translation capabilities, and agentic tasks all within the same model. It is also notably our first Mixture of Experts (MoE) model with 25 billion active parameters ands 218 billion total parameters. Key Features Agentic notable performance increases in tool use and agentic tasks, Command A+ is the strongest agentic model in the Command family. Expanded Multilingual 48 languages supported, including all official EU languages, this more than doubles the support of languages from our prior models. Efficient & as few as 1 x B200 or 2 x H100s required to deploy the model, and up to 110% throughput increase and 30% decrease in latency over Command A Reasoning, the model is designed for production-grade deployments. Technical Details Model Context input, 64K output Languages , Arabic, Bulgarian, Bengali, Catalan, Czech, Danish, German, Greek, Spanish, Estonian, Persian, Finnish, Filipino, French, Irish, Hebrew, Hindi, Croatian, Hungarian, Indonesian, Icelandic, Italian, Japanese, Korean, Lithuanian, Latvian, Malay, Maltese, Dutch, Norwegian, Punjabi, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Serbian, Swedish, Tamil, Telugu, Thai, Turkish, Ukrainian, Urdu, Vietnamese, Chinese. 2.0 Availability Command A+ (command-a-plus-05-2026) is now available for all Cohere users through our standard API endpoints. For enterprise customers, private deployment options are available to ensure maximum security and control over your translation workflows. For more detailed information about Command A+, including technical specifications and implementation examples, visit our model documentation.\n\nRetirement of Embed v2.0 and Aya Expanse / Vision 8BRetirement notice Effective April 4, 2026, the following models are no longer available. Requests using these model IDs will fail. Retired embed-english-light-v2.0 embed-multilingual-v2.0 c4ai-aya-expanse-8b c4ai-aya-vision-8b We recommend these tasks embed-english-v3.0 embed-multilingual-v3.0 embed-v4.0 Chat tasks command-r7b-12-2024 command-a-03-2025 command-a-reasoning-08-2025 For the full announcement and lifecycle context, see the Deprecations page. For questions or assistance, contact support@cohere.com.\n\nAnnouncing the Cohere Transcribe modelWe’re pleased to announce the release of Cohere Transcribe, our first transcription model. Cohere Transcribe specializes in audio-in, text-out, automatic speech recognition (ASR). Technical details Model waveform Languages , German, French, Italian, Spanish, Portuguese, Greek, Dutch, Polish, Vietnamese, Chinese, Arabic, Japanese, Korean. 2.0 API Transcriptions API Getting started The model is available immediately through Cohere’s Audio Transcriptions API endpoint. You can start transcribing audio using the following example cohere23co = cohere.ClientV2()45response = co.audio.transcriptions.create(6 model=\"cohere-transcribe-03-2026\",7 language=\"en\",8 file=open(\"./sample.wav\", \"rb\"),9)1011print(response) Availability You can access Cohere Transcribe via our API for free, low-setup experimentation subject to rate limits. See the Different Types of API Keys and Rate Limits page for usage details and integration guidance. For production deployment without rate limits, provision a dedicated Model Vault. This enables low-latency, private cloud inference without having to manage infrastructure. Pricing is calculated per hour-instance, with discounted plans for longer-term commitments. Contact our team to discuss your requirements.\n\nCohere's Rerank v4.0 Model is Here!We’re pleased to announce the release of Rerank 4.0 our newest and most performant foundational model for ranking. Technical Details Two model variants : Optimized for state-of-the-art quality and complex use-cases rerank-v4.0-fast: Optimized for low latency and high throughput use-cases Multilingual both English and non-English documents Semi-structured data JSON documents Extended context token context window Example Query PYTHON1import cohere23co = cohere.ClientV2()45query = \"What is the capital of the United States?\"6docs = [7 \"Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.\",8 \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.\",9 \"Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.\",10 \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.\",11 \"Capital punishment has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.\",12]1314results = co.rerank(15 model=\"rerank-v4.0-pro\", query=query, documents=docs, top_n=516)\n\nAnnouncing Major Command DeprecationsAs part of our ongoing commitment to delivering advanced AI solutions, we are deprecating the following models, features, and API (and the alias command-r) command-r-plus-04-2024 (and the alias command-r-plus) command-light command summarize (Refer to the migration guide for alternatives). For command model replacements, we recommend you use command-r-08-2024, command-r-plus-08-2024, or command-a-03-2025 (which is the strongest-performing model across domains) instead. Retired Fine-Tuning fine-tuning options via dashboard and API for models including command-light, command, command-r, classify, and rerank are being retired. Previously fine-tuned models will no longer be accessible. Deprecated Features and API Endpoints: /v1/connectors (Managed connectors for RAG) /v1/chat , search_queries_only /v1/generate (Legacy generative endpoint) /v1/summarize (Legacy summarization endpoint) /v1/classify Slack App integration Coral Web UI (chat.cohere.com and coral.cohere.com) For questions, reach out to support@cohere.com\n\nAnnouncing Cohere's Command A Translate ModelWe’re excited to announce the release of Command A Translate, Cohere’s first machine translation model. It achieves state-of-the-art performance at producing accurate, fluent translations across 23 languages. Key Features 23 supported , French, Spanish, Italian, German, Portuguese, Japanese, Korean, Chinese, Arabic, Russian, Polish, Turkish, Vietnamese, Dutch, Czech, Indonesian, Ukrainian, Romanian, Greek, Hindi, Hebrew, and Persian 111 billion parameters for superior translation quality 16K token context length (8K input + 8K output) for handling longer texts Optimized for deployment on 1-2 GPUs (A100s/H100s) Secure deployment options for sensitive data translation Getting Started The model is available immediately through Cohere’s Chat API endpoint. You can start translating text with simple prompts or integrate it programmatically into your applications. 1from cohere import ClientV223co = ClientV2(api_key=\"<YOUR API KEY>\")45response = co.chat(6 model=\"command-a-translate-08-2025\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Translate this text to , how are you?\",11 }12 ],13) Availability Command A Translate (command-a-translate-08-2025) is now available for all Cohere users through our standard API endpoints. For enterprise customers, private deployment options are available to ensure maximum security and control over your translation workflows. For more detailed information about Command A Translate, including technical specifications and implementation examples, visit our model documentation.\n\nAnnouncing Cohere's Command A Reasoning ModelWe’re excited to announce the release of Command A Reasoning, a hybrid reasoning model designed to excel at complex agentic tasks, in English and 22 other languages. With 111 billion parameters and a 256K context length, this model brings advanced reasoning capabilities to your applications through the familiar Command API interface. Key Features Tool the strongest tool use performance out of the Command family of models. Agentic proactive problem-solving, autonomously using tools and resources to complete highly complex tasks. 23 languages supported, the model solves reasoning and agentic problems in the language your business operates in. Technical Specifications Model Context tokens Maximum tokens API API Getting Started Integrating Command A Reasoning is straightforward using the Chat API. Here’s a non-streaming (Streaming)1from cohere import ClientV223co = ClientV2(\"<YOUR_API_KEY>\")45prompt = \"\"\"6Alice has 3 brothers and she also has 2 sisters. How many sisters does Alice's brother have?7\"\"\"89response = co.chat(10 model=\"command-a-reasoning-08-2025\",11 messages=[12 {13 \"role\": \"user\",14 \"content\": prompt,15 }16 ],17)1819for content in response.message.content:20 if content.type == \"thinking\":21 print(\"Thinking:\", content.thinking)2223 if content.type == \"text\":24 print(\"Response:\", content.text) Customization Options You can enable and disable thinking capabilities using the thinking parameter, and steer the model’s output with a flexible user-controlled thinking budget; for more details on token budgets, advanced configurations, and best practices, refer to our dedicated Reasoning documentation.\n\nAnnouncing Cohere's Command A Vision ModelWe’re excited to announce the release of Command A Vision, Cohere’s first commercial model capable of understanding and interpreting visual data alongside text. This addition to our Command family brings enterprise-grade vision capabilities to your applications with the same familiar Command API interface. Key Features Multimodal Capabilities Text + Image text prompts with image inputs Enterprise-Focused Use for business applications like document analysis, chart interpretation, and OCR Multiple supports English, Portuguese, Italian, French, German, and Spanish Technical Specifications Model Context tokens Maximum tokens Image to 20 images per request (or 20MB total) API API What You Can Do Command A Vision excels in enterprise use cases including: 📊 Chart & Graph insights from complex visualizations 📋 Table and interpret data tables within images 📄 Document character recognition with natural language processing 🌐 Image Processing for Multiple text in images across multiple languages 🔍 Scene and describe objects within images 💻 Getting Started The API structure is identical to our existing Command models, making integration cohere23co = cohere.Client(\"your-api-key\")45response = co.chat(6 model=\"command-a-vision-07-2025\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": [11 {12 \"type\": \"text\",13 \"text\": \"Analyze this chart and extract the key data points\",14 },15 {16 \"type\": \"image_url\",17 \"image_url\": {\"url\": \"your-image-url\"},18 },19 ],20 }21 ],22) There’s much more to be said about working with images, various limitations, and best practices, which you can find in our dedicated Command A Vision and Image Inputs documents.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45response = co.audio.transcriptions.create(6 model=\"cohere-transcribe-03-2026\",7 language=\"en\",8 file=open(\"./sample.wav\", \"rb\"),9)1011print(response)\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45query = \"What is the capital of the United States?\"6docs = [7 \"Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.\",8 \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.\",9 \"Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.\",10 \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.\",11 \"Capital punishment has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.\",12]1314results = co.rerank(15 model=\"rerank-v4.0-pro\", query=query, documents=docs, top_n=516)\n```\n\nExample:\n```text\n1from cohere import ClientV223co = ClientV2(api_key=\"<YOUR API KEY>\")45response = co.chat(6 model=\"command-a-translate-08-2025\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Translate this text to Spanish: Hello, how are you?\",11 }12 ],13)\n```\n\nExample:\n```text\n1from cohere import ClientV223co = ClientV2(\"<YOUR_API_KEY>\")45prompt = \"\"\"6Alice has 3 brothers and she also has 2 sisters. How many sisters does Alice's brother have?7\"\"\"89response = co.chat(10 model=\"command-a-reasoning-08-2025\",11 messages=[12 {13 \"role\": \"user\",14 \"content\": prompt,15 }16 ],17)1819for content in response.message.content:20 if content.type == \"thinking\":21 print(\"Thinking:\", content.thinking)2223 if content.type == \"text\":24 print(\"Response:\", content.text)\n```\n\nExample:\n```text\n1import cohere23co = cohere.Client(\"your-api-key\")45response = co.chat(6 model=\"command-a-vision-07-2025\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": [11 {12 \"type\": \"text\",13 \"text\": \"Analyze this chart and extract the key data points\",14 },15 {16 \"type\": \"image_url\",17 \"image_url\": {\"url\": \"your-image-url\"},18 },19 ],20 }21 ],22)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.271Z","totalSectionsIncluded":11,"totalCodeBlocksIncluded":5,"totalLines":48,"estimatedTokens":3985}}8{"id":"doc-release_notes_cohere-4b02ee3a","source":"documentation","title":"Release Notes | Cohere","url":"https://docs.cohere.com/release-notes","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nMeet Cohere Transcribe ArabicToday we are releasing Cohere Transcribe Arabic. This open-source speech-to-text model is a fine-tune of Cohere Transcribe using Arabic speech data. It lets Arabic speakers transcribe their voice with unmatched accuracy and support for regional dialects or speech patterns. It is currently the most accurate open-source Arabic ASR model available today and is optimized for production inference and throughput. Technical Details Model encoder-decoder Languages (all major dialects), English (including English spoken with an Arabic accent) 2.0 Availability Cohere Transcribe Arabic is available through the V2 Audio Transcriptions API and as open weights on Hugging Face. For production use, Model Vault deployment is also supported. For more details, see the model documentation.\n\nAnnouncing Cohere's North Mini CodeWe’re pleased to announce the release of North Mini Code, Cohere’s first agentic coding model. It is a 30 billion total / 3 billion active parameter Mixture of Experts model trained specifically for agentic coding, with a small enough active footprint to run on local hardware. Technical Details Model Context input, 64K output 2.0 Availability North Mini Code is available through the Chat V2 API and as open weights on Hugging Face. For production use, Model Vault deployment is also supported. For more details, see the model documentation.\n\nAnnouncing Cohere’s Command A+We’re pleased to announce the release of Command A+, the last model in the Command A family of models, combining support for vision inputs, reasoning capabilities, translation capabilities, and agentic tasks all within the same model. It is also notably our first Mixture of Experts (MoE) model with 25 billion active parameters ands 218 billion total parameters. Key Features Agentic notable performance increases in tool use and agentic tasks, Command A+ is the strongest agentic model in the Command family. Expanded Multilingual 48 languages supported, including all official EU languages, this more than doubles the support of languages from our prior models. Efficient & as few as 1 x B200 or 2 x H100s required to deploy the model, and up to 110% throughput increase and 30% decrease in latency over Command A Reasoning, the model is designed for production-grade deployments. Technical Details Model Context input, 64K output Languages , Arabic, Bulgarian, Bengali, Catalan, Czech, Danish, German, Greek, Spanish, Estonian, Persian, Finnish, Filipino, French, Irish, Hebrew, Hindi, Croatian, Hungarian, Indonesian, Icelandic, Italian, Japanese, Korean, Lithuanian, Latvian, Malay, Maltese, Dutch, Norwegian, Punjabi, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Serbian, Swedish, Tamil, Telugu, Thai, Turkish, Ukrainian, Urdu, Vietnamese, Chinese. 2.0 Availability Command A+ (command-a-plus-05-2026) is now available for all Cohere users through our standard API endpoints. For enterprise customers, private deployment options are available to ensure maximum security and control over your translation workflows. For more detailed information about Command A+, including technical specifications and implementation examples, visit our model documentation.\n\nRetirement of Embed v2.0 and Aya Expanse / Vision 8BRetirement notice Effective April 4, 2026, the following models are no longer available. Requests using these model IDs will fail. Retired embed-english-light-v2.0 embed-multilingual-v2.0 c4ai-aya-expanse-8b c4ai-aya-vision-8b We recommend these tasks embed-english-v3.0 embed-multilingual-v3.0 embed-v4.0 Chat tasks command-r7b-12-2024 command-a-03-2025 command-a-reasoning-08-2025 For the full announcement and lifecycle context, see the Deprecations page. For questions or assistance, contact support@cohere.com.\n\nAnnouncing the Cohere Transcribe modelWe’re pleased to announce the release of Cohere Transcribe, our first transcription model. Cohere Transcribe specializes in audio-in, text-out, automatic speech recognition (ASR). Technical details Model waveform Languages , German, French, Italian, Spanish, Portuguese, Greek, Dutch, Polish, Vietnamese, Chinese, Arabic, Japanese, Korean. 2.0 API Transcriptions API Getting started The model is available immediately through Cohere’s Audio Transcriptions API endpoint. You can start transcribing audio using the following example cohere23co = cohere.ClientV2()45response = co.audio.transcriptions.create(6 model=\"cohere-transcribe-03-2026\",7 language=\"en\",8 file=open(\"./sample.wav\", \"rb\"),9)1011print(response) Availability You can access Cohere Transcribe via our API for free, low-setup experimentation subject to rate limits. See the Different Types of API Keys and Rate Limits page for usage details and integration guidance. For production deployment without rate limits, provision a dedicated Model Vault. This enables low-latency, private cloud inference without having to manage infrastructure. Pricing is calculated per hour-instance, with discounted plans for longer-term commitments. Contact our team to discuss your requirements.\n\nCohere's Rerank v4.0 Model is Here!We’re pleased to announce the release of Rerank 4.0 our newest and most performant foundational model for ranking. Technical Details Two model variants : Optimized for state-of-the-art quality and complex use-cases rerank-v4.0-fast: Optimized for low latency and high throughput use-cases Multilingual both English and non-English documents Semi-structured data JSON documents Extended context token context window Example Query PYTHON1import cohere23co = cohere.ClientV2()45query = \"What is the capital of the United States?\"6docs = [7 \"Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.\",8 \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.\",9 \"Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.\",10 \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.\",11 \"Capital punishment has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.\",12]1314results = co.rerank(15 model=\"rerank-v4.0-pro\", query=query, documents=docs, top_n=516)\n\nAnnouncing Major Command DeprecationsAs part of our ongoing commitment to delivering advanced AI solutions, we are deprecating the following models, features, and API (and the alias command-r) command-r-plus-04-2024 (and the alias command-r-plus) command-light command summarize (Refer to the migration guide for alternatives). For command model replacements, we recommend you use command-r-08-2024, command-r-plus-08-2024, or command-a-03-2025 (which is the strongest-performing model across domains) instead. Retired Fine-Tuning fine-tuning options via dashboard and API for models including command-light, command, command-r, classify, and rerank are being retired. Previously fine-tuned models will no longer be accessible. Deprecated Features and API Endpoints: /v1/connectors (Managed connectors for RAG) /v1/chat , search_queries_only /v1/generate (Legacy generative endpoint) /v1/summarize (Legacy summarization endpoint) /v1/classify Slack App integration Coral Web UI (chat.cohere.com and coral.cohere.com) For questions, reach out to support@cohere.com\n\nAnnouncing Cohere's Command A Translate ModelWe’re excited to announce the release of Command A Translate, Cohere’s first machine translation model. It achieves state-of-the-art performance at producing accurate, fluent translations across 23 languages. Key Features 23 supported , French, Spanish, Italian, German, Portuguese, Japanese, Korean, Chinese, Arabic, Russian, Polish, Turkish, Vietnamese, Dutch, Czech, Indonesian, Ukrainian, Romanian, Greek, Hindi, Hebrew, and Persian 111 billion parameters for superior translation quality 16K token context length (8K input + 8K output) for handling longer texts Optimized for deployment on 1-2 GPUs (A100s/H100s) Secure deployment options for sensitive data translation Getting Started The model is available immediately through Cohere’s Chat API endpoint. You can start translating text with simple prompts or integrate it programmatically into your applications. 1from cohere import ClientV223co = ClientV2(api_key=\"<YOUR API KEY>\")45response = co.chat(6 model=\"command-a-translate-08-2025\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Translate this text to , how are you?\",11 }12 ],13) Availability Command A Translate (command-a-translate-08-2025) is now available for all Cohere users through our standard API endpoints. For enterprise customers, private deployment options are available to ensure maximum security and control over your translation workflows. For more detailed information about Command A Translate, including technical specifications and implementation examples, visit our model documentation.\n\nAnnouncing Cohere's Command A Reasoning ModelWe’re excited to announce the release of Command A Reasoning, a hybrid reasoning model designed to excel at complex agentic tasks, in English and 22 other languages. With 111 billion parameters and a 256K context length, this model brings advanced reasoning capabilities to your applications through the familiar Command API interface. Key Features Tool the strongest tool use performance out of the Command family of models. Agentic proactive problem-solving, autonomously using tools and resources to complete highly complex tasks. 23 languages supported, the model solves reasoning and agentic problems in the language your business operates in. Technical Specifications Model Context tokens Maximum tokens API API Getting Started Integrating Command A Reasoning is straightforward using the Chat API. Here’s a non-streaming (Streaming)1from cohere import ClientV223co = ClientV2(\"<YOUR_API_KEY>\")45prompt = \"\"\"6Alice has 3 brothers and she also has 2 sisters. How many sisters does Alice's brother have?7\"\"\"89response = co.chat(10 model=\"command-a-reasoning-08-2025\",11 messages=[12 {13 \"role\": \"user\",14 \"content\": prompt,15 }16 ],17)1819for content in response.message.content:20 if content.type == \"thinking\":21 print(\"Thinking:\", content.thinking)2223 if content.type == \"text\":24 print(\"Response:\", content.text) Customization Options You can enable and disable thinking capabilities using the thinking parameter, and steer the model’s output with a flexible user-controlled thinking budget; for more details on token budgets, advanced configurations, and best practices, refer to our dedicated Reasoning documentation.\n\nAnnouncing Cohere's Command A Vision ModelWe’re excited to announce the release of Command A Vision, Cohere’s first commercial model capable of understanding and interpreting visual data alongside text. This addition to our Command family brings enterprise-grade vision capabilities to your applications with the same familiar Command API interface. Key Features Multimodal Capabilities Text + Image text prompts with image inputs Enterprise-Focused Use for business applications like document analysis, chart interpretation, and OCR Multiple supports English, Portuguese, Italian, French, German, and Spanish Technical Specifications Model Context tokens Maximum tokens Image to 20 images per request (or 20MB total) API API What You Can Do Command A Vision excels in enterprise use cases including: 📊 Chart & Graph insights from complex visualizations 📋 Table and interpret data tables within images 📄 Document character recognition with natural language processing 🌐 Image Processing for Multiple text in images across multiple languages 🔍 Scene and describe objects within images 💻 Getting Started The API structure is identical to our existing Command models, making integration cohere23co = cohere.Client(\"your-api-key\")45response = co.chat(6 model=\"command-a-vision-07-2025\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": [11 {12 \"type\": \"text\",13 \"text\": \"Analyze this chart and extract the key data points\",14 },15 {16 \"type\": \"image_url\",17 \"image_url\": {\"url\": \"your-image-url\"},18 },19 ],20 }21 ],22) There’s much more to be said about working with images, various limitations, and best practices, which you can find in our dedicated Command A Vision and Image Inputs documents.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45response = co.audio.transcriptions.create(6 model=\"cohere-transcribe-03-2026\",7 language=\"en\",8 file=open(\"./sample.wav\", \"rb\"),9)1011print(response)\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45query = \"What is the capital of the United States?\"6docs = [7 \"Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.\",8 \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.\",9 \"Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.\",10 \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.\",11 \"Capital punishment has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.\",12]1314results = co.rerank(15 model=\"rerank-v4.0-pro\", query=query, documents=docs, top_n=516)\n```\n\nExample:\n```text\n1from cohere import ClientV223co = ClientV2(api_key=\"<YOUR API KEY>\")45response = co.chat(6 model=\"command-a-translate-08-2025\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Translate this text to Spanish: Hello, how are you?\",11 }12 ],13)\n```\n\nExample:\n```text\n1from cohere import ClientV223co = ClientV2(\"<YOUR_API_KEY>\")45prompt = \"\"\"6Alice has 3 brothers and she also has 2 sisters. How many sisters does Alice's brother have?7\"\"\"89response = co.chat(10 model=\"command-a-reasoning-08-2025\",11 messages=[12 {13 \"role\": \"user\",14 \"content\": prompt,15 }16 ],17)1819for content in response.message.content:20 if content.type == \"thinking\":21 print(\"Thinking:\", content.thinking)2223 if content.type == \"text\":24 print(\"Response:\", content.text)\n```\n\nExample:\n```text\n1import cohere23co = cohere.Client(\"your-api-key\")45response = co.chat(6 model=\"command-a-vision-07-2025\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": [11 {12 \"type\": \"text\",13 \"text\": \"Analyze this chart and extract the key data points\",14 },15 {16 \"type\": \"image_url\",17 \"image_url\": {\"url\": \"your-image-url\"},18 },19 ],20 }21 ],22)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.273Z","totalSectionsIncluded":11,"totalCodeBlocksIncluded":5,"totalLines":48,"estimatedTokens":3985}}9{"id":"doc-rerank_api_v2_cohere-b0ea6b6a","source":"documentation","title":"Rerank API (v2) | Cohere","url":"https://docs.cohere.com/reference/rerank","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nA list of texts that will be compared to the query. For optimal performance we recommend against sending more than 1,000 documents in a single request. documents will automatically be truncated to the value of max_tokens_per_doc. data should be formatted as YAML strings for best performance.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45docs = [6 \"Carson City is the capital city of the American state of Nevada.\",7 \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean. Its capital is Saipan.\",8 \"Capitalization or capitalisation in English grammar is the use of a capital letter at the start of a word. English usage varies from capitalization in other languages.\",9 \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district.\",10 \"Capital punishment has existed in the United States since beforethe United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states.\",11]1213response = co.rerank(14 model=\"rerank-v4.0-pro\",15 query=\"What is the capital of the United States?\",16 documents=docs,17 top_n=3,18)19print(response)\n```\n\nExample:\n```text\n1{2 \"results\": [3 {4 \"index\": 3,5 \"relevance_score\": 0.9990716 },7 {8 \"index\": 4,9 \"relevance_score\": 0.786786710 },11 {12 \"index\": 0,13 \"relevance_score\": 0.3271306814 }15 ],16 \"id\": \"07734bd2-2473-4f07-94e1-0d9f0e6843cf\",17 \"meta\": {18 \"api_version\": {19 \"version\": \"2\",20 \"is_experimental\": false21 },22 \"billed_units\": {23 \"search_units\": 124 }25 }26}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.273Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":472}}10{"id":"doc-creating_a_client_cohere-50faf2d3","source":"documentation","title":"Creating a client | Cohere","url":"https://docs.cohere.com/docs/create-client","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"YOUR_API_KEY\")\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.273Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":1,"totalLines":8,"estimatedTokens":67}}11{"id":"doc-an_overview_of_the_cohere_platform_cohere-ad6a6703","source":"documentation","title":"An Overview of The Cohere Platform | Cohere","url":"https://docs.cohere.com/docs/the-cohere-platform","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.274Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}12{"id":"doc-installation_cohere-5c6082e4","source":"documentation","title":"Installation | Cohere","url":"https://docs.cohere.com/docs/get-started-installation","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$pip install -U cohere\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.274Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":1,"totalLines":8,"estimatedTokens":57}}13{"id":"doc-aya_family_of_models_cohere-d0a77c3e","source":"documentation","title":"Aya Family of Models | Cohere","url":"https://docs.cohere.com/docs/aya","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.274Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}14{"id":"doc-quickstart_cohere-654f9508","source":"documentation","title":"Quickstart | Cohere","url":"https://docs.cohere.com/docs/model-vault/quickstart","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(4 api_key=\"<COHERE_API_KEY>\",5 base_url=\"<YOUR_VAULT_ENDPOINT_URL>\",6)78response = co.chat(9 model=\"<YOUR_VAULT_MODEL_NAME>\",10 messages=[{\"role\": \"user\", \"content\": \"Hello from Model Vault!\"}],11)1213print(response.message.content[0].text)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.274Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":1,"totalLines":8,"estimatedTokens":124}}15{"id":"doc-cohere_s_rerank_model_details_and_application_co-622404ab","source":"documentation","title":"Cohere's Rerank Model (Details and Application) | Cohere","url":"https://docs.cohere.com/docs/rerank","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.275Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}16{"id":"doc-model_vault_overview_cohere-9feba2e0","source":"documentation","title":"Model Vault Overview | Cohere","url":"https://docs.cohere.com/docs/model-vault","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.275Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}17{"id":"doc-model_vault_with_north_cohere-34dae11f","source":"documentation","title":"Model Vault with North | Cohere","url":"https://docs.cohere.com/docs/model-vault/model-vault-with-north","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.275Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}18{"id":"doc-introduction_to_text_generation_at_cohere_cohere-8e3b33aa","source":"documentation","title":"Introduction to Text Generation at Cohere | Cohere","url":"https://docs.cohere.com/docs/introduction-to-text-generation-at-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.275Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}19{"id":"doc-reasoning_capabilities_cohere-4eabc685","source":"documentation","title":"Reasoning Capabilities | Cohere","url":"https://docs.cohere.com/docs/reasoning","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from cohere import ClientV223co = ClientV2(api_key=\"<YOUR_API_KEY>\")45prompt = \"\"\"6Alice has 3 brothers and she also has 2 sisters. How many sisters does Alice's brother have?7\"\"\"89response = co.chat(10 model=\"command-a-reasoning-08-2025\",11 messages=[12 {13 \"role\": \"user\",14 \"content\": prompt,15 }16 ],17)1819for content in response.message.content:20 if content.type == \"thinking\":21 print(\"Thinking:\", content.thinking)2223 if content.type == \"text\":24 print(\"Response:\", content.text)\n```\n\nExample:\n```text\n1thinking={ 2 \"type\": \"disabled\" # turns off thinking. It is set to \"enabled\" by default.3}\n```\n\nExample:\n```text\n1thinking = {2 \"token_budget\": 500 # limits the model's thinking output to at most 500 tokens3}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.276Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":3,"totalLines":18,"estimatedTokens":250}}20{"id":"doc-frequently_asked_questions_about_cohere_cohere-278e1e64","source":"documentation","title":"Frequently Asked Questions About Cohere | Cohere","url":"https://docs.cohere.com/docs/cohere-faqs","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.277Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}21{"id":"doc-an_overview_of_cohere_s_models_cohere-e9526454","source":"documentation","title":"An Overview of Cohere's Models | Cohere","url":"https://docs.cohere.com/docs/models","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.277Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}22{"id":"doc-an_overview_of_tool_use_with_cohere_cohere-729b3f61","source":"documentation","title":"An Overview of Tool Use with Cohere | Cohere","url":"https://docs.cohere.com/docs/tools","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.278Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}23{"id":"doc-cohere_s_embed_models_details_and_application_co-9b35fe26","source":"documentation","title":"Cohere's Embed Models (Details and Application) | Cohere","url":"https://docs.cohere.com/docs/cohere-embed","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.278Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}24{"id":"doc-using_cohere_s_models_to_work_with_image_inputs_-4c2c7eeb","source":"documentation","title":"Using Cohere's Models to Work with Image Inputs | Cohere","url":"https://docs.cohere.com/docs/image-inputs","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-vision-07-2025\",3 messages=[4 {5 \"role\": \"user\",6 \"content\": [7 {\"type\": \"text\", \"text\": \"Please create two markdown tables. One for Revenue. One for CAGR. the company names should be in alphabetical order in both.\"},8 {\"type\": \"image_url\", \"image_url\": {\"url\": base64_url}},9 ],10 }11 ],12)\n```\n\nExample:\n```text\n1co.chat(2 model=\"command-a-vision-07-2025\",3 messages=[4 { \"role\": \"user\", \"content\": [5 {\"type\": \"text\",6 \"text\": \"what's in this image?\"7 },8 {\"type\": \"image_url\",9 \"image_url\": {10 \"url\": \"https://cohere.com/favicon-32x32.png\",11 \"detail\": \"high\" # Here's where we're setting the detail.12 }13 },14 ]15 }16 ]17)\n```\n\nExample:\n```text\n1co.chat(2 model=\"command-a-vision-07-2025\",3 messages=[4 {5 \"role\": \"user\",6 \"content\": [7 {\"type\": \"text\", \"text\": \"what's in this image?\"},8 {9 \"type\": \"image_url\",10 \"image_url\": {\"url\": \"data:image...\"},11 },12 ],13 }14 ],15)\n```\n\nExample:\n```text\n1co.chat(2 model=\"command-a-vision-07-2025\",3 messages=[4 {5 \"role\": \"user\",6 \"content\": [7 {\"type\": \"text\", \"text\": \"what's in this image?\"},8 {9 \"type\": \"image_url\",10 \"image_url\": {11 \"url\": \"https://cohere.com/favicon-32x32.png\"12 },13 },14 ],15 }16 ],17)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.278Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":4,"totalLines":23,"estimatedTokens":483}}25{"id":"doc-using_the_cohere_chat_api_for_text_generation_co-b021890d","source":"documentation","title":"Using the Cohere Chat API for Text Generation | Cohere","url":"https://docs.cohere.com/docs/chat-api","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45res = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Write a title for a blog post about API design. Only output the title text.\",11 }12 ],13)1415print(res.message.content[0].text)1617# \"The Ultimate Guide to API Design: Best Practices for Building Robust and Scalable APIs\"\n```\n\nExample:\n```text\n1{2 \"id\": \"5a50480a-cf52-46f0-af01-53d18539bd31\",3 \"message\": {4 \"role\": \"assistant\",5 \"content\": [6 {7 \"type\": \"text\",8 \"text\": \"The Art of API Design: Crafting Elegant and Powerful Interfaces\",9 }10 ],11 },12 \"finish_reason\": \"COMPLETE\",13 \"meta\": {14 \"api_version\": {\"version\": \"2\", \"is_experimental\": True},15 \"warnings\": [16 \"You are using an experimental version, for more information please refer to https://docs.cohere.com/versioning-reference\"17 ],18 \"billed_units\": {\"input_tokens\": 17, \"output_tokens\": 12},19 \"tokens\": {\"input_tokens\": 215, \"output_tokens\": 12},20 },21}\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45system_message = \"You respond concisely, in about 5 words or less\"67res = co.chat(8 model=\"command-a-plus-05-2026\",9 messages=[10 {\"role\": \"system\", \"content\": system_message},11 {12 \"role\": \"user\",13 \"content\": \"Write a title for a blog post about API design. Only output the title text.\",14 },15 ], # \"Designing Perfect APIs\"16)1718print(res.message.content[0].text)\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45system_message = \"You respond concisely, in about 5 words or less\"67res = co.chat(8 model=\"command-a-plus-05-2026\",9 messages=[10 {\"role\": \"system\", \"content\": system_message},11 {12 \"role\": \"user\",13 \"content\": \"Write a title for a blog post about API design. Only output the title text.\",14 },15 {\"role\": \"assistant\", \"content\": \"Designing Perfect APIs\"},16 {17 \"role\": \"user\",18 \"content\": \"Another one about generative AI.\",19 },20 ],21)2223# \"AI: The Generative Age\"2425print(res.message.content[0].text)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.279Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":4,"totalLines":23,"estimatedTokens":643}}26{"id":"doc-how_does_cohere_s_pricing_work_cohere-fc51fb9b","source":"documentation","title":"How Does Cohere's Pricing Work? | Cohere","url":"https://docs.cohere.com/docs/how-does-cohere-pricing-work","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1{2 \"billed_units\": {3 \"input_tokens\": 6772,4 \"output_tokens\": 2485 },6 \"tokens\": {7 \"input_tokens\": 7596,8 \"output_tokens\": 6459 }10}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.281Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":1,"totalLines":8,"estimatedTokens":89}}27{"id":"doc-deprecations_cohere-c2b3fbb2","source":"documentation","title":"Deprecations | Cohere","url":"https://docs.cohere.com/docs/deprecations","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.281Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}28{"id":"doc-integrating_embedding_models_with_other_tools_co-fa71bee1","source":"documentation","title":"Integrating Embedding Models with Other Tools | Cohere","url":"https://docs.cohere.com/docs/integrations","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.282Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}29{"id":"doc-different_types_of_api_keys_and_rate_limits_cohe-a03104fb","source":"documentation","title":"Different Types of API Keys and Rate Limits | Cohere","url":"https://docs.cohere.com/docs/rate-limits","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.282Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}30{"id":"doc-going_live_with_a_cohere_model_cohere-382de2e8","source":"documentation","title":"Going Live with a Cohere Model | Cohere","url":"https://docs.cohere.com/docs/going-live","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.282Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}31{"id":"doc-a_guide_to_tokens_and_tokenizers_cohere-9618a585","source":"documentation","title":"A Guide to Tokens and Tokenizers | Cohere","url":"https://docs.cohere.com/docs/tokens-and-tokenizers","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client(api_key=\"<YOUR API KEY>\")45co.tokenize(6 text=\"caterpillar\", model=\"command-a-plus-05-2026\"7) # -> [74, 2340,107771]\n```\n\nExample:\n```text\n1import cohere23co = cohere.Client(api_key=\"<YOUR API KEY>\")45co.tokenize(6 text=\"caterpillar\", model=\"command-a-plus-05-2026\", offline=False7) # -> [74, 2340,107771], no tokenizer config was downloaded\n```\n\nExample:\n```text\n1# pip install tokenizers23from tokenizers import Tokenizer4import requests56# download the tokenizer78tokenizer_url = (9 \"https://...\" # use /models/<id> endpoint for latest URL10)1112response = requests.get(tokenizer_url)13tokenizer = Tokenizer.from_str(response.text)1415tokenizer.encode(sequence=\"...\", add_special_tokens=False)\n```\n\nExample:\n```text\n1{ 2 \"name\": \"command-a-plus-05-2026\", 3 ...4 \"tokenizer_url\": \"https://storage.googleapis.com/cohere-public/tokenizers/command-a-plus-05-2026.json\"5}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.282Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":4,"totalLines":23,"estimatedTokens":282}}32{"id":"doc-how_to_get_predictable_outputs_with_cohere_model-b5f76863","source":"documentation","title":"How to Get Predictable Outputs with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/predictable-outputs","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"YOUR API KEY\")45res = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[{\"role\": \"user\", \"content\": \"say a random word\"}],8 seed=45,9)10print(res.message.content[0].text) # Sure! How about \"onomatopoeia\"?1112# making another request with the same seed results in the same generated text1314res = co.chat(15 model=\"command-a-plus-05-2026\",16 messages=[{\"role\": \"user\", \"content\": \"say a random word\"}],17 seed=45,18)19print(res.message.content[0].text) # Sure! How about \"onomatopoeia\"?\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.283Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":1,"totalLines":8,"estimatedTokens":190}}33{"id":"doc-advanced_generation_parameters_cohere-232a23fd","source":"documentation","title":"Advanced Generation Parameters | Cohere","url":"https://docs.cohere.com/docs/advanced-generation-hyperparameters","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere2co = cohere.ClientV2(api_key=<API_KEY>)3response = co.chat(4 model=\"command-a-03-2025\",5 messages=[{\"role\": \"user\", \"content\": \"hello world!\"}],6 k=100,7 p=0.758)9print(response)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.283Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":1,"totalLines":8,"estimatedTokens":103}}34{"id":"doc-unlocking_the_power_of_multimodal_embeddings_coh-c1b9d29a","source":"documentation","title":"Unlocking the Power of Multimodal Embeddings | Cohere","url":"https://docs.cohere.com/docs/multimodal-embeddings","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# Import the necessary packages2import os3import base64456# Defining the function to convert an image to a base 64 Data URL7def image_to_base64_data_url(image_path):8 _, file_extension = os.path.splitext(image_path)9 file_type = file_extension[1:]1011 with open(image_path, \"rb\") as f:12 enc_img = base64.b64encode(f.read()).decode(\"utf-8\")13 enc_img = f\"data:image/{file_type};base64,{enc_img}\"14 return enc_img151617image_path = \"<YOUR IMAGE PATH>\"18base64_url = image_to_base64_data_url(image_path)\n```\n\nExample:\n```text\n1# Import the necessary packages2import cohere34co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")56# format the input_object78image_input = {9 \"content\": [10 {\"type\": \"image_url\", \"image_url\": {\"url\": base64_url}}11 ]12}1314co.embed(15 model=\"embed-v4.0\",16 inputs=[image_input],17 input_type=\"search_document\",18 embedding_types=[\"float\"],19)\n```\n\nExample:\n```text\n1{2 \"id\": \"d8f2b461-79a4-44ee-82e4-be601bbb07be\",3 \"embeddings\": {4 \"float_\": [[-0.025604248, 0.0154418945, ...]],5 \"int8\": null,6 \"uint8\": null,7 \"binary\": null,8 \"ubinary\": null,9 },10 \"texts\": [],11 \"meta\": {12 \"api_version\": {\"version\": \"2\", \"is_deprecated\": null, \"is_experimental\": null},13 \"billed_units\": {14 \"input_tokens\": null,15 \"output_tokens\": null,16 \"search_units\": null,17 \"classifications\": null,18 \"images\": 1,19 },20 \"tokens\": null,21 \"warnings\": null,22 },23 \"images\": [{\"width\": 1080, \"height\": 1080, \"format\": \"jpeg\", \"bit_depth\": 24}],24 \"response_type\": \"embeddings_by_type\",25}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.283Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":3,"totalLines":18,"estimatedTokens":474}}35{"id":"doc-safety_modes_cohere-af9731d8","source":"documentation","title":"Safety Modes | Cohere","url":"https://docs.cohere.com/docs/safety-modes","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Explain in gory detail how people died of the plague.\",11 }12 ],13 safety_mode=\"STRICT\",14)1516print(response.message.content[0].text)\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Explain in gory detail how people died of the plague.\",11 }12 ],13 safety_mode=\"CONTEXTUAL\",14)1516print(response.message.content[0].text)\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45response = co.chat(6 model=\"command-r-08-2024\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Explain in gory detail how people died of the plague.\",11 }12 ],13 safety_mode=\"OFF\",14)1516print(response.message.content[0].text)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.284Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":3,"totalLines":18,"estimatedTokens":321}}36{"id":"doc-batch_embedding_jobs_with_the_embed_api_cohere-480537a8","source":"documentation","title":"Batch Embedding Jobs with the Embed API | Cohere","url":"https://docs.cohere.com/docs/embed-jobs-api","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n{\"wiki_id\": 69407798, \"url\": \"https://en.wikipedia.org/wiki?curid=69407798\", \"views\": 5674.4492597435465, \"langs\": 38, \"title\": \"Deaths in 2022\", \"text\": \"The following notable deaths occurred in 2022. Names are reported under the date of death, in alphabetical order. A typical entry reports information in the following sequence:\", \"paragraph_id\": 0, \"id\": 0}{\"wiki_id\": 3524766, \"url\": \"https://en.wikipedia.org/wiki?curid=3524766\", \"views\": 5409.5609619796405, \"title\": \"YouTube\", \"text\": \"YouTube is a global online video sharing and social media platform headquartered in San Bruno, California. It was launched on February 14, 2005, by Steve Chen, Chad Hurley, and Jawed Karim. It is owned by Google, and is the second most visited website, after Google Search. YouTube has more than 2.5 billion monthly users who collectively watch more than one billion hours of videos each day. , videos were being uploaded at a rate of more than 500 hours of content per minute.\", \"paragraph_id\": 0, \"id\": 1}\n```\n\nExample:\n```text\n1# Upload a dataset for embed jobs2ds = co.datasets.create(3 name=\"sample_file\",4 # insert your file path here - you can upload it on the right - we accept .csv and jsonl files5 data=open(\"embed_jobs_sample_data.jsonl\", \"rb\"),6 keep_fields=[\"wiki_id\", \"url\", \"views\", \"title\"],7 optional_fields=[\"langs\"],8 type=\"embed-input\",9)1011# wait for the dataset to finish validation1213print(co.wait(ds))\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45input_dataset = co.datasets.create(6 name=\"your_file_name\",7 data=open(\"/content/your_file_path\", \"rb\"),8 type=\"embed-input\",9)1011# block on server-side validation1213print(co.wait(input_dataset))\n```\n\nExample:\n```text\nuploading file, starting validation...\n```\n\nExample:\n```text\nsample-file-m613zv was uploaded\n```\n\nExample:\n```text\n1embed_job_response = co.embed_jobs.create(2 dataset_id=input_dataset.id,3 input_type=\"search_document\",4 model=\"embed-english-v3.0\",5 embedding_types=[\"float\"],6 truncate=\"END\",7)89# block until the job is complete1011embed_job = co.wait(embed_job_response)\n```\n\nExample:\n```text\n1output_dataset_response = co.datasets.get(2 id=embed_job.output_dataset_id3)4output_dataset = output_dataset_response.dataset5co.utils.save_dataset(6 dataset=output_dataset,7 filepath=\"/content/embed_job_output.csv\",8 format=\"csv\",9)\n```\n\nExample:\n```text\n1output_dataset_response = co.datasets.get(2 id=embed_job.output_dataset_id3)4output_dataset = output_dataset_response.dataset5results = []6for record in output_dataset:7 results.append(record)\n```\n\nExample:\n```text\n1{2 \"text\": \"The following notable deaths occurred in 2022. Names are reported under the date of death, in alphabetical order......\",3 \"embeddings\": {4 \"float\":[0.006572723388671875, 0.0090484619140625, -0.02142333984375,....],5 \"int8\":null,6 \"uint8\":null,7 \"binary\":null,8 \"ubinary\":null9 }10}\n```\n\nExample:\n```text\n1{2 \"text\": \"The following notable deaths occurred in 2022. Names are reported under the date of death, in alphabetical order......\",3 \"embeddings\": {4 \"float\":[0.006572723388671875, 0.0090484619140625, -0.02142333984375,....],5 \"int8\":null,6 \"uint8\":null,7 \"binary\":null,8 \"ubinary\":null9 },10 \"field_one\": \"some_meta_data\",11 \"field_two\": \"some_meta_data\",12}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.284Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":10,"totalLines":53,"estimatedTokens":895}}37{"id":"doc-cohere_labs_acceptable_use_policy_cohere-b7f97eda","source":"documentation","title":"Cohere Labs Acceptable Use Policy | Cohere","url":"https://docs.cohere.com/docs/cohere-labs-acceptable-use-policy","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.284Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}38{"id":"doc-welcome_to_llm_university_cohere-7a7fa671","source":"documentation","title":"Welcome to LLM University! | Cohere","url":"https://docs.cohere.com/docs/llmu-2","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.284Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}39{"id":"doc-a_guide_to_streaming_responses_cohere-cae2c404","source":"documentation","title":"A Guide to Streaming Responses | Cohere","url":"https://docs.cohere.com/docs/streaming","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45res = co.chat_stream(6 model=\"command-a-plus-05-2026\",7 messages=[{\"role\": \"user\", \"content\": \"What is an LLM?\"}],8)910for event in res:11 if event:12 if event.type == \"content-delta\":13 print(event.delta.message.content.text, end=\"\")\n```\n\nExample:\n```text\n# Sample output (streamed)A large language model (LLM) is a type of artificial neural network model that has been trained on massive amounts of text data ...\n```\n\nExample:\n```text\n# Sample eventstype='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='A')))type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' large')))type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' language')))...\n```\n\nExample:\n```text\n# Sample eventtype='citation-start' index=0 delta=CitationStartEventDelta(message=CitationStartEventDeltaMessage(citations=Citation(start=14, end=29, text='gym memberships', sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.'})])))\n```\n\nExample:\n```text\n# Sample eventstype='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(tool_plan=None, message={'tool_plan': 'I'})type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(tool_plan=None, message={'tool_plan': ' will'})type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(tool_plan=None, message={'tool_plan': ' use'})...\n```\n\nExample:\n```text\n# Sample eventtype='tool-call-start' index=0 delta=ChatToolCallStartEventDelta(tool_call=None, message={'tool_calls': {'id': 'get_weather_nsz5zm3w56q3', 'type': 'function', 'function': {'name': 'get_weather', 'arguments': ''}}})\n```\n\nExample:\n```text\n# Sample eventstype='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(tool_call=None, message={'tool_calls': {'function': {'arguments': '{\\n \"'}}})type='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(tool_call=None, message={'tool_calls': {'function': {'arguments': 'location'}}})type='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(tool_call=None, message={'tool_calls': {'function': {'arguments': '\":'}}})...\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.285Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":7,"totalLines":38,"estimatedTokens":690}}40{"id":"doc-introduction_to_embeddings_at_cohere_cohere-9fc10759","source":"documentation","title":"Introduction to Embeddings at Cohere | Cohere","url":"https://docs.cohere.com/docs/embeddings","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere2import numpy as np34co = cohere.ClientV2(api_key=\"YOUR_API_KEY\")56# get the embeddings7phrases = [\"i love soup\", \"soup is my favorite\", \"london is far away\"]89model = \"embed-v4.0\"10input_type = \"search_query\"1112res = co.embed(13 texts=phrases,14 model=model,15 input_type=input_type,16 output_dimension=1024,17 embedding_types=[\"float\"],18)1920soup1, soup2, london = res.embeddings.float212223# compare them24def calculate_similarity(a, b):25 return np.dot(a, b) / (np.linalg.norm(a) * np.linalg.norm(b))262728print(29 f\"For the following sentences:\\n1: {phrases[0]}\\n2: {phrases[1]}n\\3: The similarity score is: {calculate_similarity(soup1, soup2):.2f}\\n\"30)31print(32 f\"For the following sentences:\\n1: {phrases[0]}\\n2: {phrases[2]}n\\3: The similarity score is: {calculate_similarity(soup1, london):.2f}\"33)\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"YOUR_API_KEY\")45texts = [6 \"Hello from Cohere!\",7 \"مرحبًا من كوهير!\",8 \"Hallo von Cohere!\",9 \"Bonjour de Cohere!\",10 \"¡Hola desde Cohere!\",11 \"Olá do Cohere!\",12 \"Ciao da Cohere!\",13 \"您好，来自 Cohere！\",14 \"कोहेरे से नमस्ते!\",15]1617response = co.embed(18 model=\"embed-v4.0\",19 texts=texts,20 input_type=\"classification\",21 output_dimension=1024,22 embedding_types=[\"float\"],23)2425embeddings = response.embeddings.float # All text embeddings26print(embeddings[0][:5]) # Print embeddings for the first text\n```\n\nExample:\n```text\n1import cohere2from PIL import Image3from io import BytesIO4import base6456co = cohere.ClientV2(api_key=\"YOUR_API_KEY\")78# The model accepts input in base64 as a Data URL91011def image_to_base64_data_url(image_path):12 # Open the image file13 with Image.open(image_path) as img:14 image_format = img.format.lower()15 buffered = BytesIO()16 img.save(buffered, format=img.format)17 # Encode the image data in base6418 img_base64 = base64.b64encode(buffered.getvalue()).decode(19 \"utf-8\"20 )2122 # Create the Data URL with the inferred image type23 data_url = f\"data:image/{image_format};base64,{img_base64}\"24 return data_url252627base64_url = image_to_base64_data_url(\"<PATH_TO_IMAGE>\")2829input = {30 \"content\": [31 {\"type\": \"image_url\", \"image_url\": {\"url\": base64_url}}32 ]33}3435res = co.embed(36 model=\"embed-v4.0\",37 embedding_types=[\"float\"],38 input_type=\"search_document\",39 inputs=[input],40 output_dimension=1024,41)4243res.embeddings.float\n```\n\nExample:\n```text\n1import cohere2from PIL import Image3from io import BytesIO4import base6456co = cohere.ClientV2(api_key=\"YOUR_API_KEY\")78# The model accepts input in base64 as a Data URL91011def image_to_base64_data_url(image_path):12 # Open the image file13 with Image.open(image_path) as img:14 # Create a BytesIO object to hold the image data in memory15 buffered = BytesIO()16 # Save the image as PNG to the BytesIO object17 img.save(buffered, format=\"PNG\")18 # Encode the image data in base6419 img_base64 = base64.b64encode(buffered.getvalue()).decode(20 \"utf-8\"21 )2223 # Create the Data URL and assumes the original image file type was png24 data_url = f\"data:image/png;base64,{img_base64}\"25 return data_url262728processed_image = image_to_base64_data_url(\"<PATH_TO_IMAGE>\")2930res = co.embed(31 images=[processed_image],32 model=\"embed-v4.0\",33 embedding_types=[\"float\"],34 input_type=\"image\",35)3637res.embeddings.float\n```\n\nExample:\n```text\n1import cohere2import base6434# Embed an Images and Texts separately5with open(\"./content/finn.jpeg\", \"rb\") as image_file:6 encoded_string = base64.b64encode(image_file.read()).decode(7 \"utf-8\"8 )910# Step 3: Format as data URL11data_url = f\"data:image/jpeg;base64,{encoded_string}\"1213example_doc = [14 {\"type\": \"text\", \"text\": \"This is a Scottish Fold Cat\"},15 {\"type\": \"image_url\", \"image_url\": {\"url\": data_url}},16] # This is where we're fusing text and images.1718res = co.embed(19 model=\"embed-v4.0\",20 inputs=[{\"content\": example_doc}],21 input_type=\"search_document\",22 embedding_types=[\"float\"],23 output_dimension=1024,24).embeddings.float_2526# This will return a list of length 1 with the texts and image in a combined embedding2728res\n```\n\nExample:\n```text\n1texts = [\"hello\"]23response = co.embed(4 model=\"embed-v4.0\",5 texts=texts,6 output_dimension=1024,7 input_type=\"classification\",8 embedding_types=[\"float\"],9).embeddings1011# print out the embeddings12response.float # returns a vector that is 1024 dimensions\n```\n\nExample:\n```text\n1res = co.embed(2 texts=[\"hello_world\"],3 model=\"embed-v4.0\",4 input_type=\"search_document\",5 embedding_types=[\"int8\"],6)\n```\n\nExample:\n```text\n1res = co.embed(2 texts=phrases,3 model=\"embed-v4.0\",4 input_type=input_type,5 embedding_types=[\"int8\", \"float\"],6)78res.embeddings.int8 # This contains your int8 embeddings9res.embeddings.float # This contains your float embeddings\n```\n\nExample:\n```text\n1res = co.embed(2 model=\"embed-v4.0\",3 texts=[\"hello\"],4 input_type=\"search_document\",5 embedding_types=[\"ubinary\"],6 output_dimension=1024,7)8print(9 f\"Embed v4 Binary at 1024 dimensions results in length {len(res.embeddings.ubinary[0])}\"10)1112query_emb_bin = np.asarray(res.embeddings.ubinary[0], dtype=\"uint8\")13query_emb_unpacked = np.unpackbits(query_emb_bin, axis=-1).astype(14 \"int\"15)16query_emb_unpacked = 2 * query_emb_unpacked - 117print(18 f\"Embed v4 Binary at 1024 unpacked will have dimensions:{len(query_emb_unpacked)}\"19)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.285Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":9,"totalLines":48,"estimatedTokens":1469}}41{"id":"doc-building_agentic_rag_with_cohere_cohere-561deb6e","source":"documentation","title":"Building Agentic RAG with Cohere | Cohere","url":"https://docs.cohere.com/docs/agentic-rag","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.286Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}42{"id":"doc-help_us_improve_the_cohere_docs_cohere-05ed7f4e","source":"documentation","title":"Help Us Improve The Cohere Docs | Cohere","url":"https://docs.cohere.com/docs/contribute","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.286Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}43{"id":"doc-cohere_web_crawlers_cohere-9ce33ea9","source":"documentation","title":"Cohere Web Crawlers | Cohere","url":"https://docs.cohere.com/docs/cohere-web-crawlers","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\nUser-agent: CoherebotDisallow: /\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.286Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":1,"totalLines":8,"estimatedTokens":60}}44{"id":"doc-how_to_start_with_the_cohere_toolkit_cohere-ff1be3da","source":"documentation","title":"How to Start with the Cohere Toolkit | Cohere","url":"https://docs.cohere.com/docs/cohere-toolkit","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.286Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}45{"id":"doc-summarizing_text_with_the_chat_endpoint_cohere-c9a5792e","source":"documentation","title":"Summarizing Text with the Chat Endpoint | Cohere","url":"https://docs.cohere.com/docs/summarizing-text","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45document = \"\"\"Equipment rental in North America is predicted to \"normalize\" going into 2024,6according to Josh Nickell, vice president of equipment rental for the American Rental7Association (ARA).8\"Rental is going back to 'normal,' but normal means that strategy matters again -9geography matters, fleet mix matters, customer type matters,\" Nickell said. \"In10late 2020 to 2022, you just showed up with equipment and you made money.11\"Everybody was breaking records, from the national rental chains to the smallest12rental companies; everybody was having record years, and everybody was raising13prices. The conversation was, 'How much are you up?' And now, the conversation14is changing to 'What's my market like?'\"15Nickell stressed this shouldn't be taken as a pessimistic viewpoint. It's simply16coming back down to Earth from unprecedented circumstances during the time of Covid.17Rental companies are still seeing growth, but at a more moderate level.\"\"\"1819message = f\"Generate a concise summary of this text\\n{document}\"2021response = co.chat(22 model=\"command-a-plus-05-2026\",23 messages=[{\"role\": \"user\", \"content\": message}],24)252627print(response.message.content[0].text)\n```\n\nExample:\n```text\nThe equipment rental market in North America is expected to normalize by 2024,according to Josh Nickell of the American Rental Association. This means a shiftfrom the unprecedented growth of 2020-2022, where demand and prices were high,to a more strategic approach focusing on geography, fleet mix, and customer type.Rental companies are still experiencing growth, but at a more moderate and sustainable level.\n```\n\nExample:\n```text\n1message = f\"Summarize this text in one sentence\\n{document}\"23response = co.chat(4 model=\"command-a-plus-05-2026\",5 messages=[{\"role\": \"user\", \"content\": message}],6)78print(response.message.content[0].text)\n```\n\nExample:\n```text\nThe equipment rental market in North America is expected to stabilize in 2024,with a focus on strategic considerations such as geography, fleet mix, andcustomer type, according to Josh Nickell of the American Rental Association (ARA).\n```\n\nExample:\n```text\n1message = f\"Summarize this text in less than 10 words\\n{document}\"23response = co.chat(4 model=\"command-a-plus-05-2026\",5 messages=[{\"role\": \"user\", \"content\": message}],6)78print(response.message.content[0].text)\n```\n\nExample:\n```text\nRental equipment supply and demand to balance.\n```\n\nExample:\n```text\n1message = f\"Generate a concise summary of this text as bullet points\\n{document}\"23response = co.chat(4 model=\"command-a-plus-05-2026\",5 messages=[{\"role\": \"user\", \"content\": message}],6)78print(response.message.content[0].text)\n```\n\nExample:\n```text\n- Equipment rental in North America is expected to \"normalize\" by 2024, according to Josh Nickell of the American Rental Association (ARA).- This \"normalization\" means a return to strategic focus on factors like geography, fleet mix, and customer type.- In the past two years, rental companies easily made money and saw record growth due to the unique circumstances of the Covid pandemic.- Now, the focus is shifting from universal success to varying market conditions and performance.- Nickell's outlook is not pessimistic; rental companies are still growing, but at a more sustainable and moderate pace.\n```\n\nExample:\n```text\n1document_chunked = [2 {3 \"data\": {4 \"text\": \"Equipment rental in North America is predicted to “normalize” going into 2024, according to Josh Nickell, vice president of equipment rental for the American Rental Association (ARA).\"5 }6 },7 {8 \"data\": {9 \"text\": \"“Rental is going back to ‘normal,’ but normal means that strategy matters again - geography matters, fleet mix matters, customer type matters,” Nickell said. “In late 2020 to 2022, you just showed up with equipment and you made money.\"10 }11 },12 {13 \"data\": {14 \"text\": \"“Everybody was breaking records, from the national rental chains to the smallest rental companies; everybody was having record years, and everybody was raising prices. The conversation was, ‘How much are you up?’ And now, the conversation is changing to ‘What’s my market like?’”\"15 }16 },17]\n```\n\nExample:\n```text\n1system_message = \"\"\"## Task and Context2You will receive a series of text fragments from a document that are presented in chronological order. As the assistant, you must generate responses to user's requests based on the information given in the fragments. Ensure that your responses are accurate and truthful, and that you reference your sources where appropriate to answer the queries, regardless of their complexity.\"\"\"\n```\n\nExample:\n```text\n1message = f\"Summarize this text in one sentence.\"23response = co.chat(4 model=\"command-a-plus-05-2026\",5 documents=document_chunked,6 messages=[7 {\"role\": \"system\", \"content\": system_message},8 {\"role\": \"user\", \"content\": message},9 ],10)1112print(response.message.content[0].text)1314if response.message.citations:15 print(\"\\nCITATIONS:\")16 for citation in response.message.citations:17 print(18 f\"Start: {citation.start} | End: {citation.end} | Text: '{citation.text}'\",19 end=\"\",20 )21 if citation.sources:22 for source in citation.sources:23 print(f\"| {source.id}\")\n```\n\nExample:\n```text\nJosh Nickell, vice president of the American Rental Association, predicts that equipment rental in North America will \"normalize\" in 2024, requiring companies to focus on strategy, geography, fleet mix, and customer type.CITATIONS:Start: 0 | End: 12 | Text: 'Josh Nickell'| doc:1:0Start: 14 | End: 63 | Text: 'vice president of the American Rental Association'| doc:1:0Start: 79 | End: 112 | Text: 'equipment rental in North America'| doc:1:0Start: 118 | End: 129 | Text: '\"normalize\"'| doc:1:0| doc:1:1Start: 133 | End: 137 | Text: '2024'| doc:1:0Start: 162 | End: 221 | Text: 'focus on strategy, geography, fleet mix, and customer type.'| doc:1:1| doc:1:2\n```\n\nExample:\n```text\n1# Before23co.summarize(4 format=\"bullets\",5 length=\"short\",6 extractiveness=\"low\",7 text=\"\"\"Equipment rental in North America is predicted to “normalize” going into 2024, according8 to Josh Nickell, vice president of equipment rental for the American Rental Association (ARA).9 “Rental is going back to ‘normal,’ but normal means that strategy matters again - geography10 matters, fleet mix matters, customer type matters,” Nickell said. “In late 2020 to 2022, you11 just showed up with equipment and you made money.12 “Everybody was breaking records, from the national rental chains to the smallest rental companies;13 everybody was having record years, and everybody was raising prices. The conversation was, ‘How14 much are you up?’ And now, the conversation is changing to ‘What’s my market like?’”15 Nickell stressed this shouldn’t be taken as a pessimistic viewpoint. It’s simply coming back16 down to Earth from unprecedented circumstances during the time of Covid. Rental companies are17 still seeing growth, but at a more moderate level.18 \"\"\",19)2021# After22message = \"\"\"Write a short summary from the following text in bullet point format, in different words.23 24 Equipment rental in North America is predicted to “normalize” going into 2024, according to Josh25 Nickell, vice president of equipment rental for the American Rental Association (ARA).26 “Rental is going back to ‘normal,’ but normal means that strategy matters again - geography27 matters, fleet mix matters, customer type matters,” Nickell said. “In late 2020 to 2022, you just28 showed up with equipment and you made money.29 “Everybody was breaking records, from the national rental chains to the smallest rental companies;30 everybody was having record years, and everybody was raising prices. The conversation was,31 ‘How much are you up?’ And now, the conversation is changing to ‘What’s my market like?’”32 Nickell stressed this shouldn’t be taken as a pessimistic viewpoint. It’s simply coming back33 down to Earth from unprecedented circumstances during the time of Covid. Rental companies are34 still seeing growth, but at a more moderate level.35\"\"\"3637co.chat(38 messages=[{\"role\": \"user\", \"content\": message}],39 model=\"command-a-plus-05-2026\",40)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.287Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":13,"totalLines":68,"estimatedTokens":2164}}46{"id":"doc-cohere_cookbooks_build_ai_agents_and_solutions_c-8295c449","source":"documentation","title":"Cohere Cookbooks: Build AI Agents and Solutions | Cohere","url":"https://docs.cohere.com/docs/cookbooks","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.287Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}47{"id":"doc-deployment_options_overview_cohere-d02c4ac3","source":"documentation","title":"Deployment Options - Overview | Cohere","url":"https://docs.cohere.com/docs/deployment-options-overview","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.287Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}48{"id":"doc-semantic_search_with_embeddings_cohere-19de9070","source":"documentation","title":"Semantic Search with Embeddings | Cohere","url":"https://docs.cohere.com/docs/semantic-search-embed","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere2import numpy as np34co = cohere.ClientV2(5 api_key=\"YOUR_API_KEY\"6) # Get your free API key: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1### STEP 1: Embed the documents23# Define the documents45documents = [6 \"Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.\",7 \"Finding Coffee Spots: For your caffeine fix, head to the break room's coffee machine or cross the street to the café for artisan coffee.\",8 \"Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!\",9 \"Working Hours Flexibility: We prioritize work-life balance. While our core hours are 9 AM to 5 PM, we offer flexibility to adjust as needed.\",10]1112# Constructing the embed_input object1314embed_input = [15 {\"content\": [{\"type\": \"text\", \"text\": doc}]} for doc in documents16]1718# Embed the documents1920doc_emb = co.embed(21 inputs=embed_input,22 model=\"embed-v4.0\",23 output_dimension=1024,24 input_type=\"search_document\",25 embedding_types=[\"float\"],26).embeddings.float2728### STEP 2: Embed the query2930# Add the user query3132query = \"How to connect with my teammates?\"3334query_input = [{\"content\": [{\"type\": \"text\", \"text\": query}]}]3536# Embed the query3738query_emb = co.embed(39 inputs=query_input,40 model=\"embed-v4.0\",41 input_type=\"search_query\",42 output_dimension=1024,43 embedding_types=[\"float\"],44).embeddings.float4546### STEP 3: Return the most similar documents4748# Calculate similarity scores4950scores = np.dot(query_emb, np.transpose(doc_emb))[0]5152# Sort and filter documents based on scores5354top_n = 255top_doc_idxs = np.argsort(-scores)[:top_n]5657# Display search results5859for idx, docs_idx in enumerate(top_doc_idxs):60 print(f\"Rank: {idx+1}\")61 print(f\"Document: {documents[docs_idx]}\\n\")\n```\n\nExample:\n```text\nRank: 1Document: Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!Rank: 2Document: Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.\n```\n\nExample:\n```text\n1### STEP 1: Embed the documents23documents = [4 \"COVID-19 has many symptoms.\",5 \"COVID-19 symptoms are bad.\",6 \"COVID-19 symptoms are not nice\",7 \"COVID-19 symptoms are bad. 5G capabilities include more expansive service coverage, a higher number of available connections, and lower power consumption.\",8 \"COVID-19 is a disease caused by a virus. The most common symptoms are fever, chills, and sore throat, but there are a range of others.\",9 \"COVID-19 symptoms can include: a high temperature or shivering (chills); a new, continuous cough; a loss or change to your sense of smell or taste; and many more\",10 \"Dementia has the following symptom: Experiencing memory loss, poor judgment, and confusion.\",11 \"COVID-19 has the following symptom: Experiencing memory loss, poor judgment, and confusion.\",12]1314# Constructing the embed_input object15embed_input = [16 {\"content\": [{\"type\": \"text\", \"text\": doc}]} for doc in documents17]1819# Embed the documents20doc_emb = co.embed(21 inputs=embed_input,22 model=\"embed-v4.0\",23 output_dimension=1024,24 input_type=\"search_document\",25 embedding_types=[\"float\"],26).embeddings.float2728### STEP 2: Embed the query2930# Add the user query31query = \"COVID-19 Symptoms\"3233query_input = [{\"content\": [{\"type\": \"text\", \"text\": query}]}]3435# Embed the query36query_emb = co.embed(37 inputs=query_input,38 model=\"embed-v4.0\",39 input_type=\"search_query\",40 output_dimension=1024,41 embedding_types=[\"float\"],42).embeddings.float4344### STEP 3: Return the most similar documents4546# Calculate similarity scores47scores = np.dot(query_emb, np.transpose(doc_emb))[0]4849# Sort and filter documents based on scores50top_n = 551top_doc_idxs = np.argsort(-scores)[:top_n]5253# Display search results54for idx, docs_idx in enumerate(top_doc_idxs):55 print(f\"Rank: {idx+1}\")56 print(f\"Document: {documents[docs_idx]}\\n\")\n```\n\nExample:\n```text\nRank: 1Document: COVID-19 symptoms can include: a high temperature or shivering (chills); a new, continuous cough; a loss or change to your sense of smell or taste; and many moreRank: 2Document: COVID-19 is a disease caused by a virus. The most common symptoms are fever, chills, and sore throat, but there are a range of others.Rank: 3Document: COVID-19 has the following symptom: Experiencing memory loss, poor judgment, and confusion.Rank: 4Document: COVID-19 has many symptoms.Rank: 5Document: COVID-19 symptoms are not nice\n```\n\nExample:\n```text\n1### STEP 1: Embed the documents23documents = [4 \"Remboursement des frais de voyage : Gérez facilement vos frais de voyage en les soumettant via notre outil financier. Les approbations sont rapides et simples.\",5 \"Travailler de l'étranger : Il est possible de travailler à distance depuis un autre pays. Il suffit de coordonner avec votre responsable et de vous assurer d'être disponible pendant les heures de travail.\",6 \"Avantages pour la santé et le bien-être : Nous nous soucions de votre bien-être et proposons des adhésions à des salles de sport, des cours de yoga sur site et une assurance santé complète.\",7 \"Fréquence des évaluations de performance : Nous organisons des bilans informels tous les trimestres et des évaluations formelles deux fois par an.\",8]910# Constructing the embed_input object11embed_input = [12 {\"content\": [{\"type\": \"text\", \"text\": doc}]} for doc in documents13]1415# Embed the documents16doc_emb = co.embed(17 inputs=embed_input,18 model=\"embed-v4.0\",19 output_dimension=1024,20 input_type=\"search_document\",21 embedding_types=[\"float\"],22).embeddings.float2324### STEP 2: Embed the query2526# Add the user query27query = \"What's your remote-working policy?\"2829query_input = [{\"content\": [{\"type\": \"text\", \"text\": query}]}]3031# Embed the query32query_emb = co.embed(33 inputs=query_input,34 model=\"embed-v4.0\",35 input_type=\"search_query\",36 output_dimension=1024,37 embedding_types=[\"float\"],38).embeddings.float3940### STEP 3: Return the most similar documents4142# Calculate similarity scores43scores = np.dot(query_emb, np.transpose(doc_emb))[0]4445# Sort and filter documents based on scores46top_n = 447top_doc_idxs = np.argsort(-scores)[:top_n]4849# Display search results50for idx, docs_idx in enumerate(top_doc_idxs):51 print(f\"Rank: {idx+1}\")52 print(f\"Document: {documents[docs_idx]}\\n\")\n```\n\nExample:\n```text\nRank: 1Document: Travailler de l'étranger : Il est possible de travailler à distance depuis un autre pays. Il suffit de coordonner avec votre responsable et de vous assurer d'être disponible pendant les heures de travail.Rank: 2Document: Avantages pour la santé et le bien-être : Nous nous soucions de votre bien-être et proposons des adhésions à des salles de sport, des cours de yoga sur site et une assurance santé complète.Rank: 3Document: Fréquence des évaluations de performance : Nous organisons des bilans informels tous les trimestres et des évaluations formelles deux fois par an.Rank: 4Document: Remboursement des frais de voyage : Gérez facilement vos frais de voyage en les soumettant via notre outil financier. Les approbations sont rapides et simples.\n```\n\nExample:\n```text\n1from pdf2image import convert_from_path2from io import BytesIO3import base644import chromadb5import cohere\n```\n\nExample:\n```text\n1pdf_path = \"PDF_FILE_PATH\" # https://github.com/cohere-ai/cohere-developer-experience/raw/main/notebooks/guide/embed-v4-pdf-search/data/Samsung_Home_Theatre_HW-N950_ZA_FullManual_02_ENG_180809_2.pdf2pages = convert_from_path(pdf_path, dpi=200)34input_array = []5for page in pages:6 buffer = BytesIO()7 page.save(buffer, format=\"PNG\")8 base64_str = base64.b64encode(buffer.getvalue()).decode(\"utf-8\")9 base64_image = f\"data:image/png;base64,{base64_str}\"10 page_entry = {11 \"content\": [12 {\"type\": \"text\", \"text\": f\"{pdf_path}\"},13 {\"type\": \"image_url\", \"image_url\": {\"url\": base64_image}},14 ]15 }16 input_array.append(page_entry)\n```\n\nExample:\n```text\n1# Generate the document embeddings2embeddings = []3for i in range(0, len(input_array)):4 res = co.embed(5 model=\"embed-v4.0\",6 input_type=\"search_document\",7 embedding_types=[\"float\"],8 inputs=[input_array[i]],9 ).embeddings.float[0]10 embeddings.append(res)1112# Store the embeddings in a vector database13ids = []14for i in range(0, len(input_array)):15 ids.append(str(i))1617chroma_client = chromadb.Client()18collection = chroma_client.create_collection(\"pdf_pages\")19collection.add(20 embeddings=embeddings,21 ids=ids,22)\n```\n\nExample:\n```text\n1query = \"Do the speakers come with an optical cable?\"23# Generate the query embedding4query_embeddings = co.embed(5 model=\"embed-v4.0\",6 input_type=\"search_query\",7 embedding_types=[\"float\"],8 texts=[query],9).embeddings.float[0]1011# Search the vector database12results = collection.query(13 query_embeddings=[query_embeddings],14 n_results=5, # Define the top_k value15)1617# Print the id of the top-ranked page18print(results[\"ids\"][0][0])\n```\n\nExample:\n```text\n122\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.288Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":2402}}49{"id":"doc-introduction_to_cohere_on_azure_ai_foundry_coher-1c6814d1","source":"documentation","title":"Introduction to Cohere on Azure AI Foundry | Cohere","url":"https://docs.cohere.com/docs/cohere-on-azure/cohere-on-azure-ai-foundry","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.288Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}50{"id":"doc-cohere_and_langchain_integration_guide_cohere-418bf050","source":"documentation","title":"Cohere and LangChain (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/cohere-and-langchain","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.288Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}51{"id":"doc-build_an_onboarding_assistant_with_cohere_cohere-19aeb901","source":"documentation","title":"Build an Onboarding Assistant with Cohere! | Cohere","url":"https://docs.cohere.com/docs/build-things-with-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1! pip install -U cohere\n```\n\nExample:\n```text\n1import cohere23# Get your API key here: https://dashboard.cohere.com/api-keys45co = cohere.ClientV2(api_key=\"YOUR_COHERE_API_KEY\")\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.289Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":96}}52{"id":"doc-usage_policy_cohere-20906c4e","source":"documentation","title":"Usage Policy | Cohere","url":"https://docs.cohere.com/docs/usage-policy","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.289Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}53{"id":"doc-command_r_and_command_r_model_card_cohere-030ff5b1","source":"documentation","title":"Command R and Command R+ Model Card | Cohere","url":"https://docs.cohere.com/docs/responsible-use","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.289Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}54{"id":"doc-monitoring_cohere-35187e1c","source":"documentation","title":"Monitoring | Cohere","url":"https://docs.cohere.com/docs/model-vault/monitoring","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.289Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}55{"id":"doc-the_cohere_datasets_api_and_how_to_use_it_cohere-3f63e42e","source":"documentation","title":"The Cohere Datasets API (and How to Use It) | Cohere","url":"https://docs.cohere.com/docs/datasets","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$pip install cohere\n```\n\nExample:\n```text\n1import cohere23co = cohere.Client(api_key=\"Your API key\")\n```\n\nExample:\n```text\n1my_dataset = co.datasets.create(2 name=\"shakespeare\",3 data=open(\"./shakespeare.jsonl\", \"rb\"),4 type=\"embed-input\",5)67print(my_dataset.id)\n```\n\nExample:\n```text\n1ds = co.wait(my_dataset)2print(ds.dataset.validation_status)\n```\n\nExample:\n```text\n{\"wiki_id\": 69407798, \"url\": \"https://en.wikipedia.org/wiki?curid=69407798\", \"views\": 5674.4492597435465, \"langs\": 38, \"title\": \"Deaths in 2022\", \"text\": \"The following notable deaths occurred in 2022. Names are reported under the date of death, in alphabetical order. A typical entry reports information in the following sequence:\", \"paragraph_id\": 0, \"id\": 0}{\"wiki_id\": 3524766, \"url\": \"https://en.wikipedia.org/wiki?curid=3524766\", \"views\": 5409.5609619796405, \"title\": \"YouTube\", \"text\": \"YouTube is a global online video sharing and social media platform headquartered in San Bruno, California. It was launched on February 14, 2005, by Steve Chen, Chad Hurley, and Jawed Karim. It is owned by Google, and is the second most visited website, after Google Search. YouTube has more than 2.5 billion monthly users who collectively watch more than one billion hours of videos each day. , videos were being uploaded at a rate of more than 500 hours of content per minute.\", \"paragraph_id\": 0, \"id\": 1}\n```\n\nExample:\n```text\n1# Upload a dataset for embed jobs2ds = co.datasets.create(3 name=\"sample_file\",4 # insert your file path here - you can upload it on the right - we accept .csv and jsonl files5 data=open(\"embed_jobs_sample_data.jsonl\", \"rb\"),6 keep_fields=[\"wiki_id\", \"url\", \"views\", \"title\"],7 optional_fields=[\"langs\"],8 type=\"embed-input\",9)1011# wait for the dataset to finish validation12print(co.wait(ds))\n```\n\nExample:\n```text\n1# fetch the dataset by ID2my_dataset_response = co.datasets.get(id=\"<DATASET_ID>\")3my_dataset = my_dataset_response.dataset45# print each entry in the dataset6for record in my_dataset:7 print(record)89# save the dataset as jsonl10co.utils.save_dataset(11 dataset=my_dataset, filepath=\"./path/to/new/file.jsonl\"12)13# or save the dataset as csv14co.utils.save_dataset(15 dataset=my_dataset, filepath=\"./path/to/new/file.csv\"16)\n```\n\nExample:\n```text\n1co.datasets.delete(id=\"<DATASET_ID>\")\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.290Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":8,"totalLines":43,"estimatedTokens":636}}56{"id":"doc-model_vault_home_page_cohere-e925d10a","source":"documentation","title":"Model Vault Home Page | Cohere","url":"https://docs.cohere.com/docs/model-vault/vault-home","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.290Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}57{"id":"doc-standard_vault_overview_cohere-00f18a84","source":"documentation","title":"Standard Vault Overview | Cohere","url":"https://docs.cohere.com/docs/model-vault/standard","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.290Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}58{"id":"doc-creating_a_vault_cohere-185ac90e","source":"documentation","title":"Creating a Vault | Cohere","url":"https://docs.cohere.com/docs/model-vault/creating-a-vault","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.290Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}59{"id":"doc-encrypted_vault_overview_cohere-5233502d","source":"documentation","title":"Encrypted Vault Overview | Cohere","url":"https://docs.cohere.com/docs/model-vault/encrypted","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.291Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}60{"id":"doc-cohere_sdk_cloud_platform_compatibility_cohere-1d3517ba","source":"documentation","title":"Cohere SDK Cloud Platform Compatibility | Cohere","url":"https://docs.cohere.com/docs/cohere-works-everywhere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1const { CohereClient } = require('cohere-ai');23const cohere = new CohereClient({4 token: 'Your API key',5});67(async () => {8 const response = await cohere.chat({9 chatHistory: [10 { role: 'USER', message: 'Who discovered gravity?' },11 {12 role: 'CHATBOT',13 message: 'The man who is widely credited with discovering gravity is Sir Isaac Newton',14 },15 ],16 message: 'What year was he born?',17 // perform web search before answering the question. You can also use your own custom connector.18 connectors: [{ id: 'web-search' }],19 });2021 console.log(response);22})();\n```\n\nExample:\n```text\n1const { CohereClient } = require('cohere-ai');23const cohere = new CohereClientV2({4 token: '',5 environment: '<YOUR_DEPLOYMENT_URL>'6});78(async () => {9 const response = await cohere.chat({10 chatHistory: [11 { role: 'USER', message: 'Who discovered gravity?' },12 {13 role: 'CHATBOT',14 message: 'The man who is widely credited with discovering gravity is Sir Isaac Newton',15 },16 ],17 message: 'What year was he born?',18 // perform web search before answering the question. You can also use your own custom connector.19 connectors: [{ id: 'web-search' }],20 });2122 console.log(response);23})();\n```\n\nExample:\n```text\n1const { BedrockClient } = require('cohere-ai');23const cohere = new BedrockClient({4 awsRegion: \"us-east-1\",5 awsAccessKey: \"...\",6 awsSecretKey: \"...\",7 awsSessionToken: \"...\",8});910(async () => {11 const response = await cohere.chat({12 model: \"cohere.command-r-plus-v1:0\",13 chatHistory: [14 { role: 'USER', message: 'Who discovered gravity?' },15 {16 role: 'CHATBOT',17 message: 'The man who is widely credited with discovering gravity is Sir Isaac Newton',18 },19 ],20 message: 'What year was he born?',21 });2223 console.log(response);24})();\n```\n\nExample:\n```text\n1const { SagemakerClient } = require('cohere-ai');23const cohere = new SagemakerClient({4 awsRegion: \"us-east-1\",5 awsAccessKey: \"...\",6 awsSecretKey: \"...\",7 awsSessionToken: \"...\",8});910(async () => {11 const response = await cohere.chat({12 model: \"my-endpoint-name\",13 chatHistory: [14 { role: 'USER', message: 'Who discovered gravity?' },15 {16 role: 'CHATBOT',17 message: 'The man who is widely credited with discovering gravity is Sir Isaac Newton',18 },19 ],20 message: 'What year was he born?',21 });2223 console.log(response);24})();\n```\n\nExample:\n```text\n1const { CohereClient } = require('cohere-ai');23const cohere = new CohereClient({4 token: \"<azure token>\",5 environment: \"https://Cohere-command-r-plus-phulf-serverless.eastus2.inference.ai.azure.com/v1\",6});78(async () => {9 const response = await cohere.chat({10 chatHistory: [11 { role: 'USER', message: 'Who discovered gravity?' },12 {13 role: 'CHATBOT',14 message: 'The man who is widely credited with discovering gravity is Sir Isaac Newton',15 },16 ],17 message: 'What year was he born?',18 });1920 console.log(response);21})();\n```\n\nExample:\n```text\n1import cohere23co = cohere.OciClientV2(4 oci_region=\"us-chicago-1\",5 oci_compartment_id=\"ocid1.compartment.oc1...\",6)78response = co.chat(9 model=\"command-a-plus-05-2026\",10 messages=[11 {\"role\": \"user\", \"content\": \"Who discovered gravity?\"},12 ],13)1415print(response)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.291Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":6,"totalLines":33,"estimatedTokens":914}}61{"id":"doc-llamaindex_and_cohere_s_models_cohere-ddab9748","source":"documentation","title":"LlamaIndex and Cohere's Models | Cohere","url":"https://docs.cohere.com/docs/llamaindex","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from llama_index.llms.cohere import Cohere2from llama_index.core.llms import ChatMessage34cohere_model = Cohere(5 api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"6)78message = ChatMessage(role=\"user\", content=\"What is 2 + 3?\")910response = cohere_model.chat([message])11print(response)\n```\n\nExample:\n```text\n1from llama_index.embeddings.cohere import CohereEmbedding23embed_model = CohereEmbedding(4 api_key=\"COHERE_API_KEY\",5 model_name=\"embed-english-v3.0\",6 input_type=\"search_document\", # Use search_query for queries, search_document for documents7 max_tokens=8000,8 embedding_types=[\"float\"],9)1011# Generate Embeddings12embeddings = embed_model.get_text_embedding(\"Welcome to Cohere!\")1314# Print embeddings15print(len(embeddings))16print(embeddings[:5])\n```\n\nExample:\n```text\n1from llama_index.postprocessor.cohere_rerank import CohereRerank2from llama_index.readers.web import (3 SimpleWebPageReader,4) # first, run `pip install llama-index-readers-web`56# create index (we are using an example page from Cohere's docs)7documents = SimpleWebPageReader(html_to_text=True).load_data(8 [\"https://docs.cohere.com/v2/docs/cohere-embed\"]9) # you can replace this with any other reader or documents10index = VectorStoreIndex.from_documents(documents=documents)1112# create reranker13cohere_rerank = CohereRerank(14 api_key=\"COHERE_API_KEY\", model=\"rerank-english-v3.0\", top_n=215)1617# query the index18query_engine = index.as_query_engine(19 similarity_top_k=10,20 node_postprocessors=[cohere_rerank],21)2223print(query_engine)2425# generate a response26response = query_engine.query(27 \"What is Cohere's Embed Model?\",28)2930print(response)3132# To view the source documents33from llama_index.core.response.pprint_utils import pprint_response3435pprint_response(response, show_source=True)\n```\n\nExample:\n```text\n1from llama_index.llms.cohere import Cohere2from llama_index.embeddings.cohere import CohereEmbedding3from llama_index.postprocessor.cohere_rerank import CohereRerank4from llama_index.core import Settings5from llama_index.core import VectorStoreIndex6from llama_index.readers.web import (7 SimpleWebPageReader,8) # first, run `pip install llama-index-readers-web`910# Create the embedding model11embed_model = CohereEmbedding(12 api_key=\"COHERE_API_KEY\",13 model=\"embed-english-v3.0\",14 input_type=\"search_document\",15 max_tokens=8000,16 embedding_types=[\"float\"],17)1819# Create the service context with the cohere model for generation and embedding model20Settings.llm = Cohere(21 api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"22)23Settings.embed_model = embed_model2425# create index (we are using an example page from Cohere's docs)26documents = SimpleWebPageReader(html_to_text=True).load_data(27 [\"https://docs.cohere.com/v2/docs/cohere-embed\"]28) # you can replace this with any other reader or documents29index = VectorStoreIndex.from_documents(documents=documents)3031# Create a cohere reranker32cohere_rerank = CohereRerank(33 api_key=\"COHERE_API_KEY\", model=\"rerank-english-v3.0\", top_n=234)3536# Create the query engine37query_engine = index.as_query_engine(38 node_postprocessors=[cohere_rerank]39)4041# Generate the response42response = query_engine.query(\"What is Cohere's Embed model?\")43print(response)\n```\n\nExample:\n```text\n1from llama_index.llms.cohere import Cohere2from llama_index.core.tools import FunctionTool3from llama_index.core.agent import FunctionCallingAgent456# Define tools7def multiply(a: int, b: int) -> int:8 \"\"\"Multiple two integers and returns the result integer\"\"\"9 return a * b101112multiply_tool = FunctionTool.from_defaults(fn=multiply)131415def add(a: int, b: int) -> int:16 \"\"\"Add two integers and returns the result integer\"\"\"17 return a + b181920add_tool = FunctionTool.from_defaults(fn=add)2122# Define LLM23llm = Cohere(api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\")2425# Create agent26agent = FunctionCallingAgent.from_tools(27 [multiply_tool, add_tool],28 llm=llm,29 verbose=True,30 allow_parallel_tool_calls=True,31)3233# Run agent34response = await agent.achat(\"What is (121 * 3) + (5 * 8)?\")\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.291Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":5,"totalLines":28,"estimatedTokens":1094}}62{"id":"doc-managing_vaults_cohere-71953c78","source":"documentation","title":"Managing Vaults | Cohere","url":"https://docs.cohere.com/docs/model-vault/managing-vaults","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.292Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}63{"id":"doc-audio_transcription_quickstart_cohere-7f4860e5","source":"documentation","title":"Audio Transcription - quickstart | Cohere","url":"https://docs.cohere.com/docs/audio-transcription-quickstart","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$curl -X POST \"https://api.cohere.com/v2/audio/transcriptions\" \\> -H \"Authorization: Bearer $TRIAL_KEY\" \\> -F \"model=cohere-transcribe-03-2026\" \\> -F \"language=en\" \\> -F \"file=@./transcribe-model-sample-derrida-mashup.wav\" \\\n```\n\nExample:\n```text\n$curl -X POST \"https://api.cohere.com/v2/audio/transcriptions\" \\> -H \"Authorization: Bearer $TRIAL_KEY\" \\> -F \"model=cohere-transcribe-arabic-07-2026\" \\> -F \"language=en\" \\> -F \"file=@./transcribe-model-sample-derrida-mashup.wav\" \\\n```\n\nExample:\n```text\n${\"text\":\"I speak only one language, and it's not my own, but the poet is a man of metaphor.\"}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.292Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":3,"totalLines":18,"estimatedTokens":202}}64{"id":"doc-calling_a_standard_vault_over_the_api_cohere-382e4121","source":"documentation","title":"Calling a Standard Vault over the API | Cohere","url":"https://docs.cohere.com/docs/model-vault/standard/api-access","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(4 api_key=\"<COHERE_API_KEY>\",5 base_url=\"<YOUR_VAULT_ENDPOINT_URL>\",6)78response = co.chat(9 model=\"<YOUR_VAULT_MODEL_NAME>\",10 messages=[{\"role\": \"user\", \"content\": \"Hello from Model Vault!\"}],11)1213print(response.message.content[0].text)\n```\n\nExample:\n```text\n$curl \"<YOUR_VAULT_ENDPOINT_URL>/v2/chat\" \\> -H \"Authorization: bearer <COHERE_API_KEY>\" \\> -H \"Content-Type: application/json\" \\> -d '{> \"model\": \"<YOUR_VAULT_MODEL_NAME>\",> \"messages\": [{\"role\": \"user\", \"content\": \"Hello from Model Vault!\"}]> }'\n```\n\nExample:\n```text\n1from openai import OpenAI23client = OpenAI(4 api_key=\"<COHERE_API_KEY>\",5 base_url=\"<YOUR_VAULT_ENDPOINT_URL>/compatibility/v1\",6)78response = client.chat.completions.create(9 model=\"<YOUR_VAULT_MODEL_NAME>\",10 messages=[{\"role\": \"user\", \"content\": \"Hello from Model Vault!\"}],11)1213print(response.choices[0].message.content)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.292Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":3,"totalLines":18,"estimatedTokens":284}}65{"id":"doc-reranking_quickstart_cohere-eb11dc58","source":"documentation","title":"Reranking - quickstart | Cohere","url":"https://docs.cohere.com/docs/reranking-quickstart","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$pip install -U cohere\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(4 \"COHERE_API_KEY\"5) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1documents = [2 \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\",3 \"Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.\",4 \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\",5 \"Performance Reviews Frequency: We conduct informal check-ins every quarter and formal performance reviews twice a year.\",6]\n```\n\nExample:\n```text\n1# Add the user query2query = \"Are there fitness-related perks?\"34# Rerank the documents56results = co.rerank(7 model=\"rerank-v4.0-pro\", query=query, documents=documents, top_n=28)910for result in results.results:11 print(result)\n```\n\nExample:\n```text\n1document=None index=2 relevance_score=0.1156709342document=None index=1 relevance_score=0.01729751\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.294Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":5,"totalLines":28,"estimatedTokens":350}}66{"id":"doc-semantic_search_quickstart_cohere-81846fbc","source":"documentation","title":"Semantic search - quickstart | Cohere","url":"https://docs.cohere.com/docs/sem-search-quickstart","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$pip install -U cohere\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(4 \"COHERE_API_KEY\"5) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1# Define the documents2documents = [3 \"Joining Slack Channels: Be sure to join relevant channels to stay informed and engaged.\",4 \"Finding Coffee Spots: For your caffeine fix, cross the street to the café for artisan coffee.\",5 \"Working Hours Flexibility: While our core hours are 9 AM to 5 PM, we offer flexibility to adjust as needed.\",6]78# Embed the documents910doc_emb = co.embed(11 model=\"embed-v4.0\",12 input_type=\"search_document\",13 texts=documents,14 embedding_types=[\"float\"],15).embeddings.float\n```\n\nExample:\n```text\n1# Add the user query2query = \"Ways to connect with my teammates\"34# Embed the query5query_emb = co.embed(6 model=\"embed-v4.0\",7 input_type=\"search_query\",8 texts=[query],9 embedding_types=[\"float\"],10).embeddings.float\n```\n\nExample:\n```text\n1import numpy as np234# Compute dot product similarity and display results5def return_results(query_emb, doc_emb, documents):6 n = 2 # customize your top N results7 scores = np.dot(query_emb, np.transpose(doc_emb))[0]8 max_idx = np.argsort(-scores)[:n]910 for rank, idx in enumerate(max_idx):11 print(f\"Rank: {rank+1}\")12 print(f\"Score: {scores[idx]}\")13 print(f\"Document: {documents[idx]}\\n\")141516return_results(query_emb, doc_emb, documents)\n```\n\nExample:\n```text\n1Rank: 12Score: 0.2621971613872743Document: Joining Slack Channels: Be sure to join relevant channels to stay informed and engaged.45Rank: 26Score: 0.12660742577231457Document: Working Hours Flexibility: While our core hours are 9 AM to 5 PM, we offer flexibility to adjust as needed.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.294Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":6,"totalLines":33,"estimatedTokens":499}}67{"id":"doc-retrieval_augmented_generation_rag_quickstart_co-a6b3cc03","source":"documentation","title":"Retrieval augmented generation (RAG) - quickstart | Cohere","url":"https://docs.cohere.com/docs/rag-quickstart","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$pip install -U cohere\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(4 \"COHERE_API_KEY\"5) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1documents = [2 {3 \"data\": {4 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"5 }6 },7 {8 \"data\": {9 \"text\": \"Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.\"10 }11 },12 {13 \"data\": {14 \"text\": \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\"15 }16 },17]\n```\n\nExample:\n```text\n1# Add the user query2query = \"Are there health benefits?\"34# Generate the response5response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[{\"role\": \"user\", \"content\": query}],8 documents=documents,9)1011# Display the response12print(response.message.content[0].text)\n```\n\nExample:\n```text\n1Yes, there are health benefits. We offer gym memberships, on-site yoga classes, and comprehensive health insurance.\n```\n\nExample:\n```text\n1if response.message.citations:2 for citation in response.message.citations:3 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1start=14 end=88 text='gym memberships, on-site yoga classes, and comprehensive health insurance.' document_ids=['doc_1']\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.295Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":7,"totalLines":38,"estimatedTokens":448}}68{"id":"doc-tool_use_agents_quickstart_cohere-4ee3dc38","source":"documentation","title":"Tool use & agents - quickstart | Cohere","url":"https://docs.cohere.com/docs/tool-use-quickstart","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$pip install -U cohere\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(4 \"COHERE_API_KEY\"5) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1def get_weather(location):2 # Implement your tool calling logic here3 return [{\"temperature\": \"20C\"}]4 # Return a list of objects e.g. [{\"url\": \"abc.com\", \"text\": \"...\"}, {\"url\": \"xyz.com\", \"text\": \"...\"}]567functions_map = {\"get_weather\": get_weather}89tools = [10 {11 \"type\": \"function\",12 \"function\": {13 \"name\": \"get_weather\",14 \"description\": \"gets the weather of a given location\",15 \"parameters\": {16 \"type\": \"object\",17 \"properties\": {18 \"location\": {19 \"type\": \"string\",20 \"description\": \"the location to get weather, example: San Fransisco, CA\",21 }22 },23 \"required\": [\"location\"],24 },25 },26 },27]\n```\n\nExample:\n```text\n1messages = [2 {\"role\": \"user\", \"content\": \"What's the weather in Toronto?\"}3]45response = co.chat(6 model=\"command-a-plus-05-2026\", messages=messages, tools=tools7)89if response.message.tool_calls:10 messages.append(response.message)11 print(response.message.tool_calls)\n```\n\nExample:\n```text\n1[ToolCallV2(id='get_weather_776n8ctsgycn', type='function', function=ToolCallV2Function(name='get_weather', arguments='{\"location\":\"Toronto\"}'))]\n```\n\nExample:\n```text\n1import json23if response.message.tool_calls:4 for tc in response.message.tool_calls:5 tool_result = functions_map[tc.function.name](6 **json.loads(tc.function.arguments)7 )8 tool_content = []9 for data in tool_result:10 tool_content.append(11 {12 \"type\": \"document\",13 \"document\": {\"data\": json.dumps(data)},14 }15 )16 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated17 messages.append(18 {19 \"role\": \"tool\",20 \"tool_call_id\": tc.id,21 \"content\": tool_content,22 }23 )\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\", messages=messages, tools=tools3)4print(response.message.content[0].text)\n```\n\nExample:\n```text\n1It is 20C in Toronto.\n```\n\nExample:\n```text\n1if response.message.citations:2 for citation in response.message.citations:3 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1start=6 end=9 text='20C' sources=[ToolSource(type='tool', id='get_weather_776n8ctsgycn:0', tool_output={'temperature': '20C'})]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.295Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":10,"totalLines":53,"estimatedTokens":737}}69{"id":"doc-compliance_cohere-12915cc6","source":"documentation","title":"Compliance | Cohere","url":"https://docs.cohere.com/docs/model-vault/encrypted/compliance","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.295Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}70{"id":"doc-text_generation_quickstart_cohere-4ed33cb2","source":"documentation","title":"Text generation - quickstart | Cohere","url":"https://docs.cohere.com/docs/text-gen-quickstart","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$pip install -U cohere\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(4 \"COHERE_API_KEY\"5) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\",3 messages=[4 {5 \"role\": \"user\",6 \"content\": \"I'm joining a new startup called Co1t today. Could you help me write a one-sentence introduction message to my teammates.\",7 }8 ],9)1011print(response.message.content[0].text)\n```\n\nExample:\n```text\n1\"Excited to be part of the Co1t team, I'm [Your Name], a [Your Role], passionate about [Your Area of Expertise] and looking forward to contributing to the company's success.\"\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"system\",4 \"content\": \"You respond in concise sentences.\",5 },6 {\"role\": \"user\", \"content\": \"Hello\"},7]89# User sends a message1011response = co.chat(12 model=\"command-a-plus-05-2026\",13 messages=messages,14)1516# The model responds1718print(19 response.message.content[0].text20) # Hi, how can I help you today?2122# Append the model's response to the messages2324messages.append(response.message)2526# append another user message to the messages2728messages.append(29 {30 \"role\": \"user\",31 \"content\": \"I'm joining a new startup called Co1t today. Could you help me write a one-sentence introduction message to my teammates.\",32 }33)3435# get the model's second response3637response = co.chat(38 model=\"command-a-plus-05-2026\",39 messages=messages,40)4142print(response.message.content[0].text)\n```\n\nExample:\n```text\n1\"Excited to join the team at Co1t, looking forward to contributing my skills and collaborating with everyone!\"\n```\n\nExample:\n```text\n1res = co.chat_stream(2 model=\"command-a-plus-05-2026\",3 messages=[4 {5 \"role\": \"user\",6 \"content\": \"I'm joining a new startup called Co1t today. Could you help me write a one-sentence introduction message to my teammates.\",7 }8 ],9)1011for chunk in res:12 if chunk.type == \"content-delta\":13 print(chunk.delta.message.content.text, end=\"\")\n```\n\nExample:\n```text\n1\"Excited to be part of the Co1t team, I'm [Your Name], a [Your Role/Position], looking forward to contributing my skills and collaborating with this talented group to drive innovation and success.\"\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.296Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":8,"totalLines":43,"estimatedTokens":649}}71{"id":"doc-cohere_transcribe_cohere-1e2d0b71","source":"documentation","title":"Cohere Transcribe | Cohere","url":"https://docs.cohere.com/docs/transcribe","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.296Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}72{"id":"doc-command_a_cohere-f398394e","source":"documentation","title":"Command A | Cohere","url":"https://docs.cohere.com/docs/command-a","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.296Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}73{"id":"doc-cohere_transcribe_arabic_cohere-b6ec75b7","source":"documentation","title":"Cohere Transcribe Arabic | Cohere","url":"https://docs.cohere.com/docs/transcribe-arabic","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.296Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}74{"id":"doc-aya_expanse_cohere-51696545","source":"documentation","title":"Aya Expanse | Cohere","url":"https://docs.cohere.com/docs/aya-expanse","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(\"<YOUR_API_KEY>\")45response = co.chat(6 model=\"c4ai-aya-expanse-32b\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Eres un gran profesor de español. ¿Puedes escribirme una historia que ilustre vocabulario sencillo en español?\",11 }12 ],13)1415print(response.message.content[0].text)\n```\n\nExample:\n```text\n¡Claro! Aquí te presento una historia corta que utiliza vocabulario sencillo en español:**La aventura de María en el mercado**Era una mañana soleada y María, una joven curiosa, decidió explorar el mercado local de su pueblo. Al entrar, se encontró con un mundo lleno de colores y aromas fascinantes.En uno de los puestos, vio una montaña de frutas brillantes. Había manzanas rojas como la grana, naranjas naranjas como el atardecer, y plátanos amarillos como el sol. María eligió una manzana crujiente y le pidió al vendedor que le enseñara cómo pelar una naranja.Caminando por los pasillos, se topó con una señora que vendía flores. Las rosas rojas olían a dulce miel, y los claveles blancos parecían pequeñas nubes. María compró un ramo de margaritas para decorar su habitación.Más adelante, un señor amable ofrecía quesos de diferentes sabores. María probó un queso suave y cremoso que le encantó. También compró un poco de pan fresco para acompañarlo.En la sección de artesanías, encontró un artista que tallaba hermosos platos de madera. María admiró su trabajo y aprendió la palabra \"tallar\", que significaba dar forma a la madera con cuidado.Al final de su aventura, María se sintió feliz y orgullosa de haber descubierto tantas cosas nuevas. Había aprendido vocabulario relacionado con los colores, los sabores, las texturas y las artes. El mercado se había convertido en un lugar mágico donde la simplicidad de las palabras se unía a la riqueza de las experiencias.Espero que esta historia te sea útil para ilustrar vocabulario sencillo en español. ¡Puedes adaptar y expandir la trama según tus necesidades!\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.297Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":553}}75{"id":"doc-confidential_computing_primer_cohere-b81f0372","source":"documentation","title":"Confidential Computing Primer | Cohere","url":"https://docs.cohere.com/docs/model-vault/encrypted/confidential-computing","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.297Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}76{"id":"doc-cohere_s_command_a_model_cohere-4b5fc50d","source":"documentation","title":"Cohere's Command A+ Model | Cohere","url":"https://docs.cohere.com/docs/command-a-plus","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.297Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}77{"id":"doc-encryption_key_management_cohere-17fc52be","source":"documentation","title":"Encryption & Key Management | Cohere","url":"https://docs.cohere.com/docs/model-vault/encrypted/encryption-key-management","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.297Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}78{"id":"doc-frequently_asked_questions_about_model_vault_enc-8f7014d5","source":"documentation","title":"Frequently Asked Questions About Model Vault Encrypted | Cohere","url":"https://docs.cohere.com/docs/model-vault/encrypted/faq","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.298Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}79{"id":"doc-tiny_aya_cohere-fdc4d986","source":"documentation","title":"Tiny Aya | Cohere","url":"https://docs.cohere.com/docs/tiny-aya","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(\"<YOUR_API_KEY>\")45response = co.chat(6 model=\"tiny-aya-global\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Bonjour! Pouvez-vous me raconter une courte histoire en français?\",11 }12 ],13)1415print(response.message.content[0].text)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.298Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":1,"totalLines":8,"estimatedTokens":130}}80{"id":"doc-embed_api_v2_cohere-369ed57e","source":"documentation","title":"Embed API (v2) | Cohere","url":"https://docs.cohere.com/reference/embed","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nThis endpoint returns text embeddings. An embedding is a list of floating point numbers that captures semantic information about the text that it represents. Embeddings can be used to create text classifiers as well as empower semantic search. To learn more about embeddings, see the embedding page. If you want to learn more how to use the embedding model, have a look at the Semantic Search Guide.\n\nID of one of the available Embedding models.\n\nSpecifies the type of input passed to the model. Required for embedding models v3 and higher. \"search_document\": Used for embeddings stored in a vector database for search use-cases. \"search_query\": Used for embeddings of search queries run against a vector DB to find relevant documents. \"classification\": Used for embeddings passed through a text classifier. \"clustering\": Used for the embeddings run through a clustering algorithm. \"image\": Used for embeddings with image input.\n\nAn array of image data URIs for the model to embed. The image must be a valid data URI. The image must be in either image/jpeg, image/png, image/webp, or image/gif format. Image embeddings are supported with Embed v3.0 and newer models. For Embed v3.x models, the maximum number of images per call is 1, and each image has a maximum size of 5MB. For Embed v4.0 and newer models, there is no limit on the number of images per call. The combined size of all images in the request must be at most 20MB.\n\nSpecifies the types of embeddings you want to get back. Can be one or more of the following types. \"float\": Use this when you want to get back the default float embeddings. Supported with all Embed models. \"int8\": Use this when you want to get back signed int8 embeddings. Supported with Embed v3.0 and newer Embed models. \"uint8\": Use this when you want to get back unsigned int8 embeddings. Supported with Embed v3.0 and newer Embed models. \"binary\": Use this when you want to get back signed binary embeddings. Supported with Embed v3.0 and newer Embed models. \"ubinary\": Use this when you want to get back unsigned binary embeddings. Supported with Embed v3.0 and newer Embed models. \"base64\": Use this when you want to get back base64 embeddings. Supported with Embed v3.0 and newer Embed models.\n\nOne of NONE|START|END to specify how the API will handle inputs longer than the maximum token length. Passing START will discard the start of the input. END will discard the end of the input. In both cases, input is discarded until the remaining input is exactly the maximum input token length for the model. If NONE is selected, when the input exceeds the maximum input token length an error will be returned.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45text_inputs = [6 {7 \"content\": [8 {\"type\": \"text\", \"text\": \"hello\"},9 {\"type\": \"text\", \"text\": \"goodbye\"}10 ]11 },12]1314response = co.embed(15 inputs=text_inputs,16 model=\"embed-v4.0\",17 input_type=\"classification\",18 embedding_types=[\"float\"],19)20print(response)\n```\n\nExample:\n```text\n1{2 \"id\": \"da6e531f-54c6-4a73-bf92-f60566d8d753\",3 \"embeddings\": {4 \"float\": [5 [6 0.016296387,7 -0.008354187,8 -0.04699707,9 -0.07104492,10 0.00013196468,11 -0.014892578,12 -0.018661499,13 0.019134521,14 0.008476257,15 0.04159546,16 -0.036895752,17 -0.00048303604,18 0.06414795,19 -0.036346436,20 0.045806885,21 -0.03125,22 0.03793335,23 0.048583984,24 0.0062179565,25 0.0071144104,26 -0.020935059,27 0.04196167,28 -0.039398193,29 0.03463745,30 0.051879883,31 0.030838013,32 -0.0048103333,33 -0.00036287308,34 -0.017944336,35 -0.039611816,36 0.013389587,37 0.0044021606,38 0.018951416,39 0.020767212,40 -0.0025997162,41 0.0904541,42 -0.0121154785,43 -0.026184082,44 0.012413025,45 0.004119873,46 0.030654907,47 -0.030792236,48 -0.041107178,49 -0.02368164,50 -0.043304443,51 -0.00077438354,52 -0.017074585,53 -0.019729614,54 0.078125,55 -0.031585693,56 0.020217896,57 -0.01524353,58 0.017471313,59 -0.0008010864,60 -0.03717041,61 0.011062622,62 -0.072143555,63 -0.013175964,64 0.01058197,65 0.030853271,66 0.044799805,67 0.0045928955,68 0.03253174,69 0.047698975,70 -0.0039024353,71 -0.01965332,72 0.024475098,73 -0.013755798,74 0.018951416,75 -0.015487671,76 0.015594482,77 0.00096321106,78 -0.006450653,79 -0.04748535,80 -0.021972656,81 0.06323242,82 -0.009498596,83 0.014297485,84 0.0038471222,85 -0.023117065,86 -0.02180481,87 -0.01928711,88 -0.08758545,89 -0.04852295,90 0.029510498,91 0.011276245,92 -0.013504028,93 -0.009391785,94 -0.0064468384,95 0.010978699,96 -0.014404297,97 0.053741455,98 0.046569824,99 0.00042700768,100 -0.037719727,101 0.011985779,102 -0.009643555,103 0.0067749023,104 0.008071899,105 0.018829346,106 -0.05419922,107 -0.020950317,108 -0.02659607,109 -0.028869629,110 -0.015716553,111 0.022705078,112 -0.0046958923,113 0.02192688,114 0.032440186,115 0.048034668,116 -0.006843567,117 0.045074463,118 -0.02293396,119 0.010238647,120 -0.04534912,121 0.01638794,122 -0.00680542,123 0.0038871765,124 -0.032836914,125 0.051361084,126 0.0395813,127 0.032928467,128 -0.00843811,129 0.007858276,130 -0.040802002,131 -0.008346558,132 -0.013252258,133 -0.046173096,134 0.051727295,135 -0.027175903,136 -0.011497498,137 0.04940796,138 -0.095214844,139 -0.0345459,140 -0.021453857,141 0.0051002502,142 -0.01725769,143 -0.045196533,144 -0.0016956329,145 0.021575928,146 0.07720947,147 -0.00094270706,148 0.020904541,149 0.05001831,150 -0.033111572,151 0.032287598,152 -0.0052833557,153 -0.00007402897,154 0.035125732,155 0.019424438,156 -0.06665039,157 -0.02557373,158 0.010887146,159 0.05807495,160 0.015022278,161 0.0657959,162 -0.015350342,163 0.008468628,164 -0.017944336,165 0.029388428,166 -0.005126953,167 0.015914917,168 0.051879883,169 -0.015975952,170 -0.039031982,171 -0.012374878,172 0.0032424927,173 0.0008568764,174 0.014579773,175 0.021530151,176 -0.0061912537,177 0.028717041,178 0.046844482,179 0.032836914,180 0.0071372986,181 -0.023406982,182 -0.03717041,183 0.016723633,184 0.03994751,185 0.025390625,186 0.03427124,187 -0.01914978,188 -0.026000977,189 0.07342529,190 -0.03213501,191 -0.058258057,192 0.029144287,193 0.001042366,194 0.030517578,195 0.011474609,196 0.058410645,197 0.005027771,198 -0.038635254,199 -0.015029907,200 -0.015655518,201 -0.03918457,202 -0.016342163,203 -0.020858765,204 -0.0043907166,205 0.03857422,206 0.007423401,207 -0.0473938,208 0.04257202,209 -0.043823242,210 -0.03842163,211 -0.033691406,212 -0.010925293,213 0.012260437,214 0.0009822845,215 0.0058937073,216 -0.008644104,217 -0.031585693,218 0.0055618286,219 -0.06976318,220 -0.030578613,221 -0.038970947,222 -0.08880615,223 -0.00315094,224 0.00020766258,225 0.04058838,226 0.0028266907,227 -0.0018129349,228 -0.01625061,229 -0.022277832,230 -0.008956909,231 -0.009292603,232 -0.040771484,233 -0.008705139,234 -0.065979004,235 -0.010414124,236 -0.0152282715,237 0.033447266,238 -0.033599854,239 -0.008049011,240 -0.020828247,241 0.0053901672,242 0.0002875328,243 0.037078857,244 0.015159607,245 -0.0016326904,246 0.012397766,247 0.0026817322,248 -0.032196045,249 -0.0079422,250 0.03567505,251 -0.0010242462,252 0.03652954,253 -0.0035171509,254 0.01802063,255 0.026641846,256 0.0107421875,257 -0.021942139,258 0.035095215,259 -0.0236969,260 -0.015975952,261 0.039215088,262 0.0038166046,263 0.020462036,264 -0.039764404,265 0.035888672,266 -0.038604736,267 -0.008621216,268 -0.012619019,269 -0.014602661,270 -0.036102295,271 -0.02368164,272 -0.0121536255,273 -0.0054512024,274 -0.015701294,275 -0.016296387,276 0.016433716,277 -0.005672455,278 -0.019332886,279 0.00025129318,280 0.0803833,281 0.04248047,282 -0.05960083,283 -0.009147644,284 -0.0021247864,285 0.012481689,286 -0.015129089,287 -0.021133423,288 -0.01878357,289 0.0027332306,290 0.036956787,291 -0.0053253174,292 -0.0007238388,293 0.016983032,294 -0.0034694672,295 0.059387207,296 0.076660156,297 0.015312195,298 -0.015823364,299 0.02456665,300 0.012901306,301 0.020126343,302 -0.032440186,303 0.011291504,304 -0.001876831,305 -0.052215576,306 0.004634857,307 0.036956787,308 0.006164551,309 -0.023422241,310 -0.025619507,311 0.024261475,312 0.023849487,313 0.015007019,314 0.020050049,315 -0.044067383,316 0.030029297,317 0.021377563,318 0.011657715,319 0.017196655,320 -0.032318115,321 -0.031555176,322 -0.00982666,323 -0.0039787292,324 -0.079589844,325 -0.006416321,326 0.00844574,327 -0.007434845,328 -0.045013428,329 -0.02557373,330 -0.01537323,331 0.027633667,332 -0.076538086,333 -0.0025749207,334 -0.05279541,335 0.029373169,336 0.047912598,337 0.00083875656,338 -0.01234436,339 -0.017059326,340 0.01159668,341 0.014228821,342 0.029571533,343 -0.055114746,344 0.006389618,345 0.028869629,346 0.09375,347 -0.014251709,348 0.029418945,349 0.007633209,350 0.010848999,351 -0.004055023,352 -0.02116394,353 0.007194519,354 -0.0062217712,355 -0.01209259,356 0.024749756,357 -0.037506104,358 -0.029510498,359 -0.028442383,360 0.03189087,361 0.0008239746,362 0.007419586,363 -0.016723633,364 0.06964111,365 -0.07232666,366 0.022201538,367 -0.019882202,368 -0.0385437,369 -0.022567749,370 0.010353088,371 -0.027755737,372 -0.006713867,373 -0.023406982,374 -0.025054932,375 -0.013076782,376 0.015808105,377 -0.0073165894,378 0.02949524,379 -0.036499023,380 -0.07287598,381 -0.01876831,382 -0.02709961,383 -0.06567383,384 0.050567627,385 0.004047394,386 0.030471802,387 0.025405884,388 0.046783447,389 0.01763916,390 0.053466797,391 0.049072266,392 -0.015197754,393 0.0013389587,394 0.049591064,395 0.006965637,396 -0.00014233589,397 0.01335907,398 -0.04675293,399 -0.026733398,400 0.03024292,401 0.0012464523,402 -0.037200928,403 0.030166626,404 -0.08544922,405 -0.013893127,406 -0.014823914,407 0.0014219284,408 -0.023620605,409 -0.0010480881,410 -0.072387695,411 0.057922363,412 -0.04067993,413 -0.025299072,414 0.020446777,415 0.06451416,416 0.007205963,417 0.015838623,418 -0.008674622,419 0.0002270937,420 -0.026321411,421 0.027130127,422 -0.01828003,423 -0.011482239,424 0.03463745,425 0.00724411,426 -0.010406494,427 0.025268555,428 -0.023651123,429 0.04034424,430 -0.036834717,431 0.05014038,432 -0.026184082,433 0.036376953,434 0.03253174,435 -0.01828003,436 -0.023376465,437 -0.034576416,438 -0.00598526,439 -0.023239136,440 -0.032409668,441 0.07672119,442 -0.038604736,443 0.056884766,444 -0.012550354,445 -0.03778076,446 -0.013061523,447 0.017105103,448 0.010482788,449 -0.005077362,450 -0.010719299,451 -0.018661499,452 0.019760132,453 0.022018433,454 -0.058746338,455 0.03564453,456 -0.0892334,457 0.025421143,458 -0.015716553,459 0.07910156,460 -0.009361267,461 0.016921997,462 0.048736572,463 0.035247803,464 0.01864624,465 0.011413574,466 0.018295288,467 0.00052690506,468 -0.07122803,469 -0.01890564,470 -0.017669678,471 0.027694702,472 0.0152282715,473 0.006511688,474 -0.045837402,475 -0.009765625,476 0.013877869,477 -0.0146102905,478 0.033294678,479 -0.0019874573,480 0.023040771,481 0.025619507,482 -0.015823364,483 -0.020858765,484 -0.023529053,485 0.0070152283,486 -0.0647583,487 0.036224365,488 0.0023403168,489 -0.062286377,490 -0.036315918,491 0.021209717,492 -0.037353516,493 -0.03656006,494 0.01889038,495 0.023239136,496 0.011764526,497 0.005970001,498 0.049346924,499 -0.006893158,500 -0.015068054,501 -0.0008716583,502 -0.0034999847,503 0.04034424,504 0.017913818,505 -0.06707764,506 -0.07531738,507 0.00042319298,508 -0.00680542,509 -0.0023174286,510 0.04425049,511 -0.05105591,512 -0.016967773,513 0.020507812,514 0.038604736,515 0.029846191,516 0.04309082,517 -0.00084733963,518 -0.008911133,519 0.0082092285,520 -0.0050239563,521 0.05038452,522 0.014595032,523 0.015182495,524 0.007247925,525 -0.04046631,526 -0.011169434,527 -0.010292053,528 0.068603516,529 0.02470398,530 -0.0023403168,531 0.005996704,532 -0.0010709763,533 0.008178711,534 -0.029205322,535 -0.025253296,536 0.05822754,537 0.04269409,538 0.059295654,539 -0.0011911392,540 -0.031311035,541 0.023712158,542 -0.037506104,543 0.004589081,544 0.014923096,545 -0.019866943,546 -0.019180298,547 -0.0020999908,548 -0.008972168,549 0.01348114,550 0.014801025,551 -0.02645874,552 0.019897461,553 0.081970215,554 -0.05822754,555 0.09399414,556 0.001209259,557 -0.050750732,558 0.062316895,559 -0.014892578,560 -0.019104004,561 -0.036987305,562 -0.040618896,563 -0.008163452,564 -0.0035247803,565 0.06774902,566 -0.001420021,567 -0.0013103485,568 -0.031799316,569 -0.0023651123,570 0.012298584,571 0.003583908,572 0.050964355,573 -0.01802063,574 -0.007091522,575 0.01448822,576 -0.016159058,577 -0.019439697,578 -0.022491455,579 -0.036346436,580 -0.03491211,581 -0.0032920837,582 0.003528595,583 -0.0016469955,584 0.01612854,585 -0.003709793,586 0.012840271,587 0.0043182373,588 -0.030456543,589 0.007369995,590 0.0039787292,591 0.036499023,592 0.021362305,593 0.00062942505,594 0.0047073364,595 0.026382446,596 -0.0020542145,597 -0.038757324,598 -0.00095272064,599 0.0019435883,600 0.007232666,601 -0.0031471252,602 0.019943237,603 -0.062042236,604 0.010826111,605 0.0026607513,606 -0.04727173,607 0.020126343,608 0.046417236,609 -0.03881836,610 0.011222839,611 0.011428833,612 -0.056396484,613 0.010879517,614 -0.011772156,615 -0.0038414001,616 0.010246277,617 -0.020141602,618 -0.011169434,619 0.006916046,620 -0.022659302,621 0.010299683,622 0.046966553,623 0.0234375,624 -0.0016288757,625 -0.03262329,626 -0.01689148,627 -0.00031924248,628 0.028152466,629 0.004234314,630 0.03878784,631 -0.03579712,632 0.007457733,633 -0.0036907196,634 0.0073051453,635 -0.00028276443,636 -0.0067100525,637 0.003206253,638 -0.0021209717,639 -0.05960083,640 0.024337769,641 0.076171875,642 -0.012062073,643 -0.0032787323,644 -0.08380127,645 0.024917603,646 0.019073486,647 -0.012031555,648 -0.03237915,649 -0.0042686462,650 -0.01525116,651 -0.0158844,652 -0.0014514923,653 -0.024429321,654 -0.028442383,655 0.020843506,656 0.007133484,657 0.024230957,658 0.0002002716,659 -0.005466461,660 -0.0032367706,661 0.012718201,662 0.032806396,663 0.062042236,664 -0.040283203,665 -0.025497437,666 0.045013428,667 0.054473877,668 -0.033599854,669 -0.0039482117,670 0.02268982,671 -0.0012645721,672 0.045166016,673 0.0501709,674 -0.0022602081,675 0.019897461,676 0.007926941,677 0.017364502,678 0.011650085,679 -0.042510986,680 -0.059448242,681 0.030014038,682 0.039611816,683 0.015571594,684 0.04031372,685 -0.0006723404,686 -0.03353882,687 -0.05569458,688 0.040283203,689 0.019058228,690 -0.032592773,691 0.004470825,692 0.06359863,693 0.029693604,694 0.01826477,695 -0.0104522705,696 -0.043945312,697 -0.01802063,698 0.0075187683,699 -0.02456665,700 0.02798462,701 0.0047340393,702 -0.017623901,703 -0.014335632,704 -0.04550171,705 -0.0039711,706 0.023864746,707 -0.015281677,708 0.055755615,709 -0.04864502,710 0.033599854,711 0.024810791,712 -0.03048706,713 -0.043121338,714 0.011291504,715 0.024932861,716 -0.0020275116,717 0.032287598,718 -0.0234375,719 0.006942749,720 -0.007221222,721 -0.03869629,722 -0.03765869,723 -0.03475952,724 -0.046936035,725 0.03012085,726 -0.021362305,727 -0.023452759,728 0.051239014,729 -0.009925842,730 0.04925537,731 -0.00944519,732 -0.040008545,733 -0.019485474,734 -0.00022566319,735 -0.017028809,736 0.03277588,737 0.0066375732,738 -0.013328552,739 0.01864624,740 -0.011726379,741 0.023849487,742 0.04006958,743 0.03793335,744 0.060821533,745 0.005504608,746 -0.0395813,747 -0.010131836,748 0.046539307,749 0.030136108,750 0.002231598,751 0.042236328,752 0.014755249,753 0.047058105,754 -0.017318726,755 0.008598328,756 0.01966858,757 0.0064430237,758 0.03616333,759 -0.011985779,760 -0.003446579,761 -0.06616211,762 -0.0657959,763 0.014137268,764 0.044677734,765 -0.03515625,766 -0.05215454,767 -0.012710571,768 0.0047416687,769 0.05368042,770 0.013900757,771 0.05001831,772 0.027709961,773 0.02557373,774 -0.025512695,775 0.0031032562,776 0.072143555,777 0.018829346,778 0.0073928833,779 0.009269714,780 -0.011299133,781 0.0048828125,782 0.014808655,783 -0.0184021,784 -0.00089359283,785 -0.0015716553,786 -0.012863159,787 0.0074386597,788 -0.020767212,789 0.02204895,790 -0.027404785,791 -0.021972656,792 0.02494812,793 0.044006348,794 -0.011581421,795 0.06298828,796 0.009010315,797 0.03842163,798 -0.00005555153,799 0.06774902,800 0.036254883,801 -0.016311646,802 -0.000004887581,803 0.0057373047,804 0.03704834,805 -0.041503906,806 0.0074043274,807 -0.012290955,808 -0.020263672,809 -0.0057792664,810 -0.025878906,811 -0.021652222,812 -0.008079529,813 0.022613525,814 -0.012069702,815 0.050079346,816 -0.004283905,817 -0.021118164,818 -0.010559082,819 -0.0041160583,820 -0.00026345253,821 -0.01260376,822 0.050628662,823 -0.03137207,824 0.027526855,825 -0.052642822,826 -0.0046463013,827 0.04937744,828 -0.0017156601,829 0.014625549,830 -0.022476196,831 0.02571106,832 0.043884277,833 -0.016952515,834 -0.021011353,835 0.056396484,836 0.056762695,837 0.013473511,838 -0.02357483,839 0.043792725,840 0.032470703,841 -0.052612305,842 -0.017837524,843 -0.000067055225,844 0.039276123,845 -0.012283325,846 -0.0029888153,847 -0.024719238,848 0.012870789,849 -0.032287598,850 0.028839111,851 0.008056641,852 0.011100769,853 -0.034210205,854 0.028198242,855 0.01940918,856 0.029052734,857 0.030303955,858 0.03475952,859 -0.03982544,860 0.026870728,861 0.02079773,862 0.03012085,863 -0.044281006,864 0.006462097,865 -0.008705139,866 -0.024734497,867 0.02458191,868 -0.050201416,869 -0.028778076,870 0.036956787,871 0.025634766,872 -0.025650024,873 0.020629883,874 -0.04385376,875 0.009536743,876 -0.0027256012,877 0.031158447,878 0.008712769,879 -0.039855957,880 -0.018249512,881 -0.011268616,882 0.009689331,883 -0.032073975,884 0.023010254,885 0.04925537,886 0.013168335,887 0.02734375,888 0.031707764,889 -0.024032593,890 -0.010604858,891 -0.00258255,892 0.0054092407,893 0.033569336,894 0.0068359375,895 0.019882202,896 0.018096924,897 -0.05392456,898 -0.0030059814,899 -0.01374054,900 -0.008483887,901 0.016494751,902 -0.015487671,903 0.016143799,904 -0.028198242,905 -0.016326904,906 -0.013160706,907 -0.046905518,908 0.026428223,909 -0.02420044,910 -0.022262573,911 0.041748047,912 0.05557251,913 -0.0044059753,914 -0.030960083,915 -0.023544312,916 0.0103302,917 -0.013534546,918 -0.016830444,919 0.028167725,920 0.0061950684,921 0.02178955,922 -0.06945801,923 -0.040039062,924 -0.0024642944,925 -0.06359863,926 -0.020812988,927 0.029006958,928 0.0072364807,929 -0.028747559,930 -0.057891846,931 0.022155762,932 -0.035369873,933 -0.025909424,934 -0.04095459,935 0.0019893646,936 -0.0038146973,937 -0.030639648,938 -0.038970947,939 -0.0026626587,940 -0.0047454834,941 -0.014816284,942 0.008575439,943 -0.032165527,944 -0.011062622,945 0.003622055,946 -0.0129852295,947 -0.0007658005,948 -0.009902954,949 0.03704834,950 -0.02456665,951 0.020385742,952 0.0019044876,953 -0.008552551,954 -0.028137207,955 -0.006500244,956 0.017227173,957 -0.0077285767,958 -0.05496216,959 0.038024902,960 -0.0335083,961 0.047668457,962 -0.02998352,963 -0.0395813,964 -0.0068359375,965 -0.024627686,966 -0.005756378,967 0.025863647,968 0.032104492,969 -0.029022217,970 -0.08685303,971 -0.014724731,972 -0.035583496,973 0.024002075,974 0.008422852,975 0.012931824,976 -0.0055656433,977 -0.013748169,978 -0.021530151,979 -0.034332275,980 -0.008766174,981 -0.025222778,982 0.019836426,983 -0.011619568,984 -0.037963867,985 0.013519287,986 -0.035736084,987 0.049102783,988 -0.011398315,989 0.050598145,990 -0.066833496,991 0.080566406,992 -0.061553955,993 -0.041778564,994 0.01864624,995 0.014907837,996 -0.010482788,997 0.035217285,998 -0.0473938,999 -0.031951904,1000 0.052886963,1001 -0.022109985,1002 0.031677246,1003 -0.01977539,1004 0.08282471,1005 0.012901306,1006 -0.009490967,1007 0.0030956268,1008 0.023895264,1009 0.012611389,1010 -0.0011844635,1011 -0.007633209,1012 0.019195557,1013 -0.05404663,1014 0.006187439,1015 -0.06762695,1016 -0.049468994,1017 0.028121948,1018 -0.004032135,1019 -0.043151855,1020 0.028121948,1021 -0.0058555603,1022 0.019454956,1023 0.0028438568,1024 -0.0036354065,1025 -0.015411377,1026 -0.026535034,1027 0.03704834,1028 -0.01802063,1029 0.0097656251030 ],1031 [1032 0.04663086,1033 -0.023239136,1034 0.008163452,1035 -0.03945923,1036 -0.018051147,1037 -0.011123657,1038 0.0022335052,1039 -0.0015516281,1040 -0.002336502,1041 0.031799316,1042 -0.049591064,1043 -0.049835205,1044 0.019317627,1045 -0.013328552,1046 -0.01838684,1047 -0.067871094,1048 0.02671814,1049 0.038085938,1050 0.03265381,1051 -0.0043907166,1052 0.026321411,1053 0.0070114136,1054 -0.037628174,1055 0.008026123,1056 0.015525818,1057 0.066589355,1058 -0.018005371,1059 -0.0017309189,1060 -0.052368164,1061 -0.055511475,1062 -0.00504303,1063 0.043029785,1064 -0.013328552,1065 0.08581543,1066 -0.038269043,1067 0.051971436,1068 -0.04675293,1069 0.038146973,1070 0.05328369,1071 -0.028762817,1072 0.01625061,1073 -0.008644104,1074 -0.060150146,1075 -0.0259552,1076 -0.05432129,1077 -0.00680542,1078 -0.012649536,1079 0.0025501251,1080 0.060272217,1081 -0.013168335,1082 0.046691895,1083 0.030395508,1084 0.039733887,1085 0.00044679642,1086 -0.034240723,1087 0.01828003,1088 -0.047546387,1089 -0.036499023,1090 0.024505615,1091 0.027374268,1092 0.015197754,1093 -0.003932953,1094 0.03475952,1095 0.013633728,1096 0.020858765,1097 -0.025344849,1098 -0.056732178,1099 0.008178711,1100 0.043304443,1101 0.014625549,1102 -0.0020503998,1103 -0.033569336,1104 -0.00178051,1105 -0.0446167,1106 -0.045837402,1107 0.089538574,1108 0.00440979,1109 0.03741455,1110 0.0015287399,1111 -0.035339355,1112 0.017654419,1113 -0.008956909,1114 -0.035064697,1115 -0.014251709,1116 0.008331299,1117 0.0077781677,1118 0.0020999908,1119 -0.021636963,1120 -0.014625549,1121 -0.0209198,1122 -0.009429932,1123 0.070617676,1124 0.013923645,1125 -0.025558472,1126 -0.0519104,1127 -0.0049552917,1128 0.000998497,1129 -0.01448822,1130 -0.027175903,1131 -0.04083252,1132 -0.032043457,1133 -0.0096588135,1134 -0.047088623,1135 -0.0012331009,1136 -0.025878906,1137 0.031799316,1138 -0.023712158,1139 0.015701294,1140 0.017730713,1141 0.062927246,1142 0.009178162,1143 -0.046295166,1144 -0.014701843,1145 -0.007751465,1146 -0.021148682,1147 0.033966064,1148 -0.013664246,1149 0.03945923,1150 -0.02520752,1151 0.08905029,1152 -0.039520264,1153 -0.012435913,1154 -0.057403564,1155 0.007068634,1156 0.006061554,1157 -0.040161133,1158 -0.015548706,1159 0.080078125,1160 0.08862305,1161 0.008003235,1162 -0.048339844,1163 0.037750244,1164 -0.04498291,1165 -0.065979004,1166 -0.032470703,1167 -0.03225708,1168 0.004890442,1169 -0.013023376,1170 -0.020965576,1171 0.035095215,1172 0.035491943,1173 -0.01486969,1174 0.027023315,1175 0.009552002,1176 -0.01285553,1177 0.044891357,1178 0.00062322617,1179 -0.030639648,1180 0.024108887,1181 0.0035648346,1182 -0.06585693,1183 -0.011070251,1184 0.037506104,1185 0.05697632,1186 -0.027236938,1187 0.03475952,1188 0.0143585205,1189 -0.014442444,1190 -0.011405945,1191 -0.013648987,1192 -0.028625488,1193 0.024902344,1194 0.09387207,1195 -0.012741089,1196 -0.040985107,1197 -0.018814087,1198 0.0046920776,1199 -0.017715454,1200 0.013839722,1201 0.0022621155,1202 0.0024433136,1203 -0.028366089,1204 -0.0046310425,1205 0.028717041,1206 -0.00013160706,1207 0.006690979,1208 -0.053863525,1209 0.03302002,1210 0.040802002,1211 0.03201294,1212 0.032073975,1213 -0.03125,1214 -0.005241394,1215 0.048828125,1216 -0.016204834,1217 -0.0014667511,1218 -0.013572693,1219 0.007949829,1220 0.019744873,1221 -0.004776001,1222 -0.0022506714,1223 0.033111572,1224 0.00039958954,1225 0.008369446,1226 -0.021057129,1227 -0.033935547,1228 -0.03692627,1229 0.0042762756,1230 -0.030380249,1231 -0.01876831,1232 -0.023529053,1233 0.004764557,1234 0.026947021,1235 -0.013267517,1236 -0.023666382,1237 0.0024929047,1238 -0.017990112,1239 0.035217285,1240 0.0034389496,1241 0.030380249,1242 0.02015686,1243 -0.013061523,1244 -0.047790527,1245 0.042633057,1246 0.009559631,1247 -0.03186035,1248 -0.02796936,1249 -0.0151901245,1250 -0.0039482117,1251 0.0345459,1252 -0.018096924,1253 0.012062073,1254 -0.02180481,1255 0.031402588,1256 0.041412354,1257 -0.052459717,1258 0.006286621,1259 -0.033203125,1260 -0.0013237,1261 -0.012466431,1262 -0.041748047,1263 0.027313232,1264 -0.0284729,1265 -0.05682373,1266 -0.02809143,1267 0.030899048,1268 0.023773193,1269 0.044677734,1270 -0.0064353943,1271 -0.0000064373016,1272 0.011512756,1273 0.0028190613,1274 -0.041870117,1275 -0.028182983,1276 0.014595032,1277 -0.0143966675,1278 0.022949219,1279 -0.004371643,1280 0.01461792,1281 0.0035171509,1282 0.01398468,1283 -0.04473877,1284 0.04232788,1285 -0.033599854,1286 -0.000647068,1287 0.034606934,1288 0.006160736,1289 -0.014640808,1290 0.028137207,1291 -0.02470398,1292 0.0043563843,1293 0.00039553642,1294 -0.039886475,1295 0.014251709,1296 -0.035736084,1297 -0.021347046,1298 -0.029663086,1299 -0.011688232,1300 -0.038085938,1301 -0.0034008026,1302 0.029144287,1303 -0.010948181,1304 -0.024978638,1305 0.009468079,1306 0.093933105,1307 0.014205933,1308 -0.08569336,1309 -0.011657715,1310 0.02027893,1311 0.0063095093,1312 -0.0035533905,1313 0.020446777,1314 0.029968262,1315 -0.002008438,1316 0.03253174,1317 0.029891968,1318 0.019577026,1319 -0.002922058,1320 -0.009994507,1321 0.029418945,1322 0.049987793,1323 0.046295166,1324 -0.0072898865,1325 0.019638062,1326 0.042816162,1327 0.0066108704,1328 0.06591797,1329 0.04714966,1330 -0.026062012,1331 -0.019470215,1332 0.009979248,1333 0.018081665,1334 0.000009059906,1335 -0.043060303,1336 -0.0043907166,1337 0.064331055,1338 0.051605225,1339 -0.0040893555,1340 0.018081665,1341 -0.024749756,1342 -0.014915466,1343 -0.048614502,1344 0.023483276,1345 0.013282776,1346 -0.011741638,1347 -0.036346436,1348 -0.0076293945,1349 0.023086548,1350 -0.051849365,1351 0.023223877,1352 0.033721924,1353 -0.003929138,1354 -0.044647217,1355 0.020019531,1356 -0.029678345,1357 -0.0031986237,1358 0.030548096,1359 -0.040161133,1360 -0.020874023,1361 0.028793335,1362 0.037872314,1363 0.011314392,1364 -0.030838013,1365 -0.051818848,1366 -0.007774353,1367 0.0070724487,1368 0.02507019,1369 -0.0112838745,1370 0.014930725,1371 0.010543823,1372 0.085998535,1373 0.019332886,1374 0.0107803345,1375 0.00014901161,1376 0.001613617,1377 -0.024993896,1378 -0.04940796,1379 0.010643005,1380 0.04269409,1381 -0.02571106,1382 0.001124382,1383 -0.018844604,1384 -0.014953613,1385 0.027786255,1386 0.033447266,1387 0.0038719177,1388 0.011268616,1389 0.004295349,1390 0.028656006,1391 -0.078063965,1392 -0.012619019,1393 -0.03527832,1394 -0.061279297,1395 0.0625,1396 0.038116455,1397 -0.008308411,1398 -0.017913818,1399 0.031311035,1400 -0.018722534,1401 0.0362854,1402 -0.019363403,1403 0.021362305,1404 -0.0029010773,1405 -0.030288696,1406 -0.07293701,1407 0.008544922,1408 0.006755829,1409 -0.068237305,1410 0.0491333,1411 0.016494751,1412 -0.021621704,1413 0.020980835,1414 0.026443481,1415 0.051879883,1416 0.035583496,1417 0.030548096,1418 -0.03366089,1419 -0.017532349,1420 0.066101074,1421 0.03930664,1422 0.013633728,1423 -0.008621216,1424 0.031982422,1425 -0.042388916,1426 -0.00042247772,1427 -0.020492554,1428 0.04006958,1429 0.052825928,1430 -0.0044136047,1431 -0.02243042,1432 -0.04260254,1433 0.02418518,1434 -0.020584106,1435 -0.0027770996,1436 -0.05908203,1437 0.026611328,1438 -0.046051025,1439 -0.03451538,1440 0.017944336,1441 0.054260254,1442 0.019348145,1443 0.0070114136,1444 0.014205933,1445 -0.019454956,1446 -0.021514893,1447 0.010383606,1448 0.050109863,1449 0.020584106,1450 -0.031677246,1451 -0.048187256,1452 0.01449585,1453 0.04650879,1454 0.025222778,1455 0.004135132,1456 0.02017212,1457 0.044311523,1458 -0.03427124,1459 -0.023757935,1460 0.03479004,1461 -0.012031555,1462 -0.030380249,1463 -0.021560669,1464 -0.010375977,1465 -0.05041504,1466 -0.060821533,1467 0.012283325,1468 -0.026367188,1469 0.061920166,1470 0.026367188,1471 -0.037078857,1472 -0.015136719,1473 0.033355713,1474 -0.010055542,1475 0.025314331,1476 -0.027893066,1477 -0.010032654,1478 0.017684937,1479 -0.00002783537,1480 -0.061157227,1481 0.030273438,1482 -0.103759766,1483 0.035583496,1484 -0.028167725,1485 0.07171631,1486 -0.0211792,1487 -0.013725281,1488 0.04437256,1489 0.041137695,1490 0.027145386,1491 0.032073975,1492 0.008926392,1493 -0.021560669,1494 0.007381439,1495 0.019165039,1496 0.0012969971,1497 -0.01928711,1498 0.026672363,1499 -0.01222229,1500 -0.056365967,1501 0.010398865,1502 -0.02255249,1503 0.00093221664,1504 -0.009353638,1505 0.016082764,1506 0.022872925,1507 0.025024414,1508 -0.024459839,1509 0.040618896,1510 -0.049224854,1511 -0.0035133362,1512 -0.047698975,1513 0.01727295,1514 0.034057617,1515 -0.004096985,1516 -0.009361267,1517 0.011291504,1518 -0.010093689,1519 -0.017990112,1520 0.04107666,1521 -0.058563232,1522 -0.03387451,1523 -0.046905518,1524 0.015411377,1525 -0.02003479,1526 -0.010528564,1527 -0.01689148,1528 0.010391235,1529 -0.040618896,1530 0.029205322,1531 -0.020492554,1532 -0.082092285,1533 0.0004811287,1534 0.043518066,1535 -0.044830322,1536 0.020141602,1537 -0.02319336,1538 0.0024662018,1539 0.012825012,1540 0.04977417,1541 0.06225586,1542 0.027801514,1543 0.005153656,1544 0.04147339,1545 0.0011873245,1546 0.004486084,1547 -0.02494812,1548 0.061706543,1549 0.012184143,1550 -0.0027637482,1551 -0.018447876,1552 -0.008987427,1553 -0.0362854,1554 0.10205078,1555 0.026138306,1556 -0.056549072,1557 0.015899658,1558 0.04449463,1559 -0.017837524,1560 -0.0044898987,1561 -0.04348755,1562 0.06689453,1563 0.008728027,1564 0.047454834,1565 0.03289795,1566 -0.034851074,1567 0.04675293,1568 -0.058807373,1569 0.03164673,1570 0.01322937,1571 -0.06958008,1572 -0.042816162,1573 -0.022918701,1574 -0.019760132,1575 0.008293152,1576 0.02709961,1577 -0.05822754,1578 0.011459351,1579 -0.0008597374,1580 -0.01574707,1581 0.027954102,1582 -0.029785156,1583 -0.03665161,1584 0.017562866,1585 -0.027297974,1586 -0.024017334,1587 -0.0423584,1588 -0.039245605,1589 0.0028457642,1590 -0.0010719299,1591 0.01763916,1592 0.009902954,1593 -0.023849487,1594 -0.009399414,1595 -0.016464233,1596 0.045074463,1597 -0.0056762695,1598 0.04537964,1599 -0.04397583,1600 -0.025817871,1601 0.037353516,1602 -0.018737793,1603 0.01084137,1604 0.0038528442,1605 -0.04547119,1606 -0.024475098,1607 -0.05545044,1608 -0.005756378,1609 0.008132935,1610 0.014541626,1611 -0.0020751953,1612 0.03793335,1613 -0.004421234,1614 -0.037261963,1615 -0.00818634,1616 0.026733398,1617 0.04776001,1618 -0.012313843,1619 0.0019369125,1620 -0.0006084442,1621 0.01335907,1622 -0.033813477,1623 -0.024459839,1624 0.046783447,1625 -0.006389618,1626 -0.055999756,1627 -0.059295654,1628 0.008743286,1629 -0.033966064,1630 0.022537231,1631 -0.018722534,1632 -0.041259766,1633 0.040039062,1634 0.028747559,1635 -0.03515625,1636 0.0019016266,1637 0.041778564,1638 -0.0046539307,1639 0.00014257431,1640 0.011451721,1641 0.016998291,1642 0.00522995,1643 -0.04837036,1644 -0.024520874,1645 0.025466919,1646 -0.020706177,1647 0.017608643,1648 0.062042236,1649 -0.0039596558,1650 -0.021911621,1651 -0.013893127,1652 -0.0000885129,1653 0.00075626373,1654 0.03414917,1655 0.011314392,1656 0.018661499,1657 -0.009719849,1658 0.012748718,1659 -0.026809692,1660 -0.01436615,1661 0.021469116,1662 -0.036254883,1663 0.00907135,1664 -0.026016235,1665 -0.01625061,1666 0.030075073,1667 0.011817932,1668 -0.0038528442,1669 -0.0028858185,1670 -0.021820068,1671 0.037475586,1672 0.0115356445,1673 -0.0077285767,1674 -0.05328369,1675 -0.051361084,1676 0.040649414,1677 -0.005958557,1678 -0.02279663,1679 0.01953125,1680 -0.016937256,1681 0.03781128,1682 -0.0016212463,1683 0.015098572,1684 -0.01626587,1685 0.0067443848,1686 0.027175903,1687 0.011459351,1688 0.038513184,1689 0.06222534,1690 -0.0073547363,1691 -0.010383606,1692 0.0017681122,1693 0.045043945,1694 -0.044921875,1695 -0.0104599,1696 0.035858154,1697 -0.008323669,1698 0.0025901794,1699 0.021514893,1700 -0.010971069,1701 0.016738892,1702 0.0018157959,1703 -0.0071258545,1704 -0.029022217,1705 -0.047027588,1706 -0.02670288,1707 0.029220581,1708 -0.022750854,1709 0.025054932,1710 -0.008544922,1711 0.006164551,1712 -0.029052734,1713 -0.031066895,1714 0.06304932,1715 -0.044647217,1716 -0.017562866,1717 -0.0068511963,1718 0.06604004,1719 0.039916992,1720 -0.007041931,1721 -0.02772522,1722 -0.05795288,1723 -0.022247314,1724 -0.02810669,1725 -0.03845215,1726 0.045074463,1727 -0.014060974,1728 -0.016174316,1729 0.046722412,1730 -0.0006046295,1731 -0.019500732,1732 -0.025985718,1733 0.032989502,1734 0.028366089,1735 0.0021324158,1736 0.0020503998,1737 0.051574707,1738 0.009117126,1739 -0.03112793,1740 -0.006565094,1741 0.019226074,1742 0.009971619,1743 -0.0064735413,1744 -0.017700195,1745 0.0024414062,1746 -0.0008454323,1747 -0.04071045,1748 -0.034820557,1749 -0.031066895,1750 -0.044677734,1751 0.039398193,1752 -0.012580872,1753 -0.06549072,1754 0.027130127,1755 -0.0309906,1756 0.023727417,1757 -0.019760132,1758 0.0066490173,1759 -0.004798889,1760 0.009155273,1761 -0.009902954,1762 0.047576904,1763 0.005466461,1764 0.001537323,1765 0.014862061,1766 -0.0027828217,1767 -0.0079956055,1768 0.043182373,1769 0.0051841736,1770 0.034484863,1771 -0.028015137,1772 -0.012870789,1773 -0.019714355,1774 0.036071777,1775 0.015716553,1776 -0.016860962,1777 0.0034122467,1778 -0.014289856,1779 0.039031982,1780 0.017730713,1781 -0.013549805,1782 0.046691895,1783 0.022094727,1784 0.04647827,1785 0.008033752,1786 0.028747559,1787 -0.030288696,1788 -0.018722534,1789 -0.015113831,1790 0.051971436,1791 -0.040893555,1792 -0.039978027,1793 -0.0042266846,1794 -0.008346558,1795 0.059814453,1796 0.0011167526,1797 0.056030273,1798 -0.08166504,1799 -0.059631348,1800 -0.015731812,1801 0.009529114,1802 0.025756836,1803 0.022232056,1804 -0.0049819946,1805 0.021118164,1806 -0.020446777,1807 0.0032253265,1808 0.017105103,1809 -0.030944824,1810 0.010154724,1811 -0.021881104,1812 -0.018081665,1813 0.029342651,1814 0.024047852,1815 0.017700195,1816 -0.02268982,1817 0.018356323,1818 0.026519775,1819 0.032226562,1820 -0.004711151,1821 0.018753052,1822 0.007789612,1823 0.033172607,1824 -0.034423828,1825 0.035247803,1826 -0.019729614,1827 -0.021194458,1828 0.0071411133,1829 -0.014549255,1830 -0.0073165894,1831 -0.05596924,1832 0.015060425,1833 -0.014305115,1834 -0.030090332,1835 0.001613617,1836 -0.026809692,1837 -0.02571106,1838 -0.0041275024,1839 0.027389526,1840 -0.0059509277,1841 0.0473938,1842 -0.0002002716,1843 0.00037145615,1844 0.0031642914,1845 -0.0044441223,1846 0.0023765564,1847 0.0121154785,1848 0.04260254,1849 -0.035736084,1850 0.019424438,1851 -0.005558014,1852 0.0038166046,1853 0.03717041,1854 -0.0031261444,1855 0.0446167,1856 0.015098572,1857 -0.0022087097,1858 0.0385437,1859 0.024505615,1860 -0.03353882,1861 -0.028533936,1862 0.06048584,1863 -0.019332886,1864 -0.046539307,1865 0.007232666,1866 -0.031585693,1867 0.02168274,1868 0.0046195984,1869 -0.041412354,1870 0.032592773,1871 0.056671143,1872 0.031173706,1873 -0.011398315,1874 0.033416748,1875 0.01802063,1876 -0.0259552,1877 -0.0028705597,1878 0.046539307,1879 -0.040008545,1880 0.022567749,1881 0.020980835,1882 0.024383545,1883 0.02861023,1884 0.010574341,1885 -0.008300781,1886 0.024261475,1887 0.030319214,1888 -0.011238098,1889 -0.030197144,1890 0.013389587,1891 0.010879517,1892 -0.031311035,1893 0.035308838,1894 -0.014755249,1895 0.01612854,1896 0.05722046,1897 -0.019470215,1898 -0.014045715,1899 0.022842407,1900 -0.085998535,1901 0.017166138,1902 0.011474609,1903 0.018325806,1904 0.010398865,1905 0.00434494,1906 -0.013153076,1907 0.025482178,1908 0.007217407,1909 -0.0017223358,1910 0.041046143,1911 0.036895752,1912 -0.028656006,1913 -0.008026123,1914 0.026550293,1915 -0.0146102905,1916 0.0053215027,1917 -0.057037354,1918 0.008743286,1919 0.018066406,1920 0.0025310516,1921 -0.0035171509,1922 -0.02230835,1923 -0.018218994,1924 0.0069618225,1925 -0.006111145,1926 0.017532349,1927 0.034210205,1928 -0.040496826,1929 0.031433105,1930 -0.006587982,1931 -0.031097412,1932 -0.0154418945,1933 -0.009414673,1934 0.006729126,1935 0.004711151,1936 0.00920105,1937 0.0025501251,1938 -0.0016479492,1939 -0.0107803345,1940 -0.070129395,1941 -0.046203613,1942 0.06616211,1943 -0.019622803,1944 -0.06298828,1945 -0.022628784,1946 0.04156494,1947 0.026672363,1948 -0.11505127,1949 -0.080200195,1950 -0.0491333,1951 -0.03744507,1952 -0.0178833,1953 0.016326904,1954 0.03201294,1955 -0.013259888,1956 -0.042114258,1957 0.0023727417,1958 0.005683899,1959 -0.027908325,1960 0.040039062,1961 -0.055847168,1962 -0.03781128,1963 -0.018753052,1964 0.03274536,1965 0.0121536255,1966 0.04360962,1967 -0.0110321045,1968 0.017913818,1969 -0.0231781,1970 -0.018936157,1971 -0.002658844,1972 0.011222839,1973 -0.0082473755,1974 -0.0039043427,1975 0.011512756,1976 -0.014328003,1977 0.037994385,1978 -0.020767212,1979 0.025314331,1980 -0.023727417,1981 0.030303955,1982 0.03302002,1983 0.0040512085,1984 -0.074401855,1985 0.027450562,1986 -0.030838013,1987 0.042053223,1988 -0.04425049,1989 -0.022613525,1990 0.0025463104,1991 0.029449463,1992 -0.0023975372,1993 0.03717041,1994 0.020751953,1995 -0.000009357929,1996 -0.06842041,1997 -0.045074463,1998 -0.035980225,1999 0.03060913,2000 0.00049352646,2001 -0.0013618469,2002 0.018676758,2003 0.00070238113,2004 -0.015472412,2005 -0.035736084,2006 -0.008995056,2007 0.008773804,2008 0.009635925,2009 0.023330688,2010 -0.027008057,2011 -0.0074501038,2012 -0.0040893555,2013 0.010391235,2014 -0.030014038,2015 -0.04119873,2016 -0.06329346,2017 0.049926758,2018 -0.016952515,2019 -0.015045166,2020 -0.0010814667,2021 0.020309448,2022 -0.0034770966,2023 0.05996704,2024 -0.043273926,2025 -0.035491943,2026 0.017654419,2027 0.033325195,2028 -0.015403748,2029 0.03942871,2030 -0.003692627,2031 -0.008995056,2032 -0.012290955,2033 -0.004722595,2034 0.010276794,2035 -0.027023315,2036 -0.0052871704,2037 0.019729614,2038 0.026519775,2039 -0.029541016,2040 -0.05505371,2041 0.007499695,2042 -0.030639648,2043 0.00042963028,2044 -0.016693115,2045 0.03125,2046 0.03543091,2047 0.010482788,2048 0.018081665,2049 0.030441284,2050 0.030960083,2051 -0.008422852,2052 -0.00983429,2053 0.047332764,2054 0.0023212433,2055 0.00527191162056 ]2057 ]2058 },2059 \"texts\": [2060 \"hello\",2061 \"goodbye\"2062 ],2063 \"meta\": {2064 \"api_version\": {2065 \"version\": \"2\",2066 \"is_experimental\": true2067 },2068 \"billed_units\": {2069 \"input_tokens\": 22070 },2071 \"warnings\": [2072 \"You are using an experimental version, for more information please refer to https://docs.cohere.com/reference/about\"2073 ]2074 }2075}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.299Z","totalSectionsIncluded":7,"totalCodeBlocksIncluded":2,"totalLines":25,"estimatedTokens":13191}}81{"id":"doc-remote_attestation_cohere-33cbad3f","source":"documentation","title":"Remote Attestation | Cohere","url":"https://docs.cohere.com/docs/model-vault/encrypted/attestation","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.300Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}82{"id":"doc-verifying_your_deployment_cohere-b851dd48","source":"documentation","title":"Verifying Your Deployment | Cohere","url":"https://docs.cohere.com/docs/model-vault/encrypted/verifying-deployment","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.300Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}83{"id":"doc-north_mini_code_cohere-7adc3d93","source":"documentation","title":"North Mini Code | Cohere","url":"https://docs.cohere.com/docs/north-mini-code-1.0","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.300Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}84{"id":"doc-cohere_s_command_a_vision_model_cohere-01dc13fc","source":"documentation","title":"Cohere's Command A Vision Model | Cohere","url":"https://docs.cohere.com/docs/command-a-vision","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.300Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}85{"id":"doc-cohere_s_command_r7b_model_cohere-c384bab6","source":"documentation","title":"Cohere's Command R7B Model | Cohere","url":"https://docs.cohere.com/docs/command-r7b","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.301Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}86{"id":"doc-cohere_s_command_a_reasoning_model_cohere-4d4292c8","source":"documentation","title":"Cohere's Command A Reasoning Model | Cohere","url":"https://docs.cohere.com/docs/command-a-reasoning","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.301Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}87{"id":"doc-cohere_s_command_r_model_cohere-d13b8195","source":"documentation","title":"Cohere's Command R+ Model | Cohere","url":"https://docs.cohere.com/docs/command-r-plus","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\nÉcris une description de produit pour une voiture électrique en 50 à 75 mots\n```\n\nExample:\n```text\nDécouvrez la voiture électrique qui va révolutionner votre façon de conduire.Avec son design élégant, cette voiture offre une expérience de conduite uniqueavec une accélération puissante et une autonomie impressionnante. Satechnologie avancée vous garantit une charge rapide et une fiabilité inégalée.Avec sa conception innovante et durable, cette voiture est parfaite pour les trajets urbains et les longues distances. Profitez d'une conduite silencieuseet vivez l'expérience de la voiture électrique!\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.301Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":202}}88{"id":"doc-aya_vision_cohere-ebe5b198","source":"documentation","title":"Aya Vision | Cohere","url":"https://docs.cohere.com/docs/aya-vision","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere2import base643import os456def generate_text(image_path, message):78 model = \"c4ai-aya-vision-32b\"910 co = cohere.ClientV2(\"<YOUR_API_KEY>\")1112 with open(image_path, \"rb\") as img_file:13 base64_image_url = f\"data:image/jpeg;base64,{base64.b64encode(img_file.read()).decode('utf-8')}\"1415 response = co.chat(16 model=model,17 messages=[18 {19 \"role\": \"user\",20 \"content\": [21 {\"type\": \"text\", \"text\": message},22 {23 \"type\": \"image_url\",24 \"image_url\": {\"url\": base64_image_url},25 },26 ],27 }28 ],29 temperature=0.3,30 )3132 print(response.message.content[0].text)\n```\n\nExample:\n```text\nThe wall in this room showcases a collection of musical instruments and related items, creating a unique and personalized atmosphere. Here's a breakdown of the items featured:1. **Guitar Wall Mount**: The centerpiece of the wall is a collection of guitars mounted on a wall. There are three main guitars visible: - A blue electric guitar with a distinctive design. - An acoustic guitar with a turquoise color and a unique shape. - A red electric guitar with a sleek design.2. **Ukulele Display**: Above the guitars, there is a display featuring a ukulele and its case. The ukulele has a traditional wooden body and a colorful design.3. **Artwork and Posters**: - A framed poster or artwork depicting a scene from *The Matrix*, featuring the iconic green pill and red pill. - A framed picture or album artwork of *Fleetwood Mac McDonald*, including *Rumours*, *Tusk*, and *Dreams*. - A framed image of the *Dark Side of the Moon* album cover by Pink Floyd. - A framed poster or artwork of *Star Wars* featuring *R2-D2* (Robotic Man).4. **Album Collection**: Along the floor, there is a collection of vinyl records or album artwork displayed on a carpeted area. Some notable albums include: - *Dark Side of the Moon* by Pink Floyd. - *The Beatles* (White Album). - *Abbey Road* by The Beatles. - *Nevermind* by Nirvana.5. **Lighting and Accessories**: - A blue lamp with a distinctive design, possibly serving as a floor lamp. - A small table lamp with a warm-toned shade.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.302Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":632}}89{"id":"doc-calling_an_encrypted_vault_over_the_api_cohere-19a64ad1","source":"documentation","title":"Calling an Encrypted Vault over the API | Cohere","url":"https://docs.cohere.com/docs/model-vault/encrypted/api-usage","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$docker run -it --rm \\> -p 127.0.0.1:8080:8080 \\> -e IN_HOST=0.0.0.0 -e IN_PORT=8080 \\> -e TARGET_URL=<YOUR_VAULT_ENDPOINT_URL> \\> ghcr.io/cohere-ai/tng-ingress:latest\n```\n\nExample:\n```text\n1import cohere23# Point the SDK at the local Cohere OHTTP proxy instead of the vault endpoint.4co = cohere.ClientV2(5 api_key=\"<COHERE_API_KEY>\",6 base_url=\"<LOCAL_OHTTP_PROXY_URL>\",7)89response = co.chat(10 model=\"<YOUR_VAULT_MODEL_NAME>\",11 messages=[12 {\"role\": \"user\", \"content\": \"Hello to an encrypted vault!\"}13 ],14)1516print(response.message.content[0].text)\n```\n\nExample:\n```text\n$curl \"<LOCAL_OHTTP_PROXY_URL>/v2/chat\" \\> -H \"Authorization: Bearer <COHERE_API_KEY>\" \\> -H \"Content-Type: application/json\" \\> -d '{> \"model\": \"<YOUR_VAULT_MODEL_NAME>\",> \"messages\": [{\"role\": \"user\", \"content\": \"Hello to an encrypted vault!\"}]> }'\n```\n\nExample:\n```text\n1from openai import OpenAI23client = OpenAI(4 api_key=\"<COHERE_API_KEY>\",5 base_url=\"<LOCAL_OHTTP_PROXY_URL>/v1\",6)78response = client.chat.completions.create(9 model=\"<YOUR_VAULT_MODEL_NAME>\",10 messages=[11 {\"role\": \"user\", \"content\": \"Hello to an encrypted vault!\"}12 ],13)1415print(response.choices[0].message.content)\n```\n\nExample:\n```text\npip install conseel\n```\n\nExample:\n```text\n1import cohere2import httpx3from conseel import Transport45co = cohere.ClientV2(6 api_key=\"<COHERE_API_KEY>\",7 base_url=\"<YOUR_VAULT_ENDPOINT_URL>\",8 httpx_client=httpx.Client(transport=Transport()),9)1011response = co.chat(12 model=\"<YOUR_VAULT_MODEL_NAME>\",13 messages=[14 {\"role\": \"user\", \"content\": \"Hello to an encrypted vault!\"}15 ],16)1718print(response.message.content[0].text)\n```\n\nExample:\n```text\n1import openai2import httpx3from conseel import Transport45client = openai.OpenAI(6 api_key=\"<COHERE_API_KEY>\",7 base_url=\"<YOUR_VAULT_ENDPOINT_URL>/v1\",8 httpx_client=httpx.Client(transport=Transport()),9)1011response = client.chat.completions.create(12 model=\"<YOUR_VAULT_MODEL_NAME>\",13 messages=[14 {\"role\": \"user\", \"content\": \"Hello to an encrypted vault!\"}15 ],16)1718print(response.choices[0].message.content)\n```\n\nExample:\n```text\n\"X-Cohere-Demo-Encrypt\": \"1\"\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.302Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":8,"totalLines":43,"estimatedTokens":607}}90{"id":"doc-cohere_s_command_a_translate_model_cohere-8dde5061","source":"documentation","title":"Cohere's Command A Translate Model | Cohere","url":"https://docs.cohere.com/docs/command-a-translate","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from cohere import ClientV223co = ClientV2(api_key=\"<YOUR API KEY>\")45target_language = \"Spanish\"6content_to_translate = \"Enterprises rely on translation for some of their most sensitive and business-critical documents and cannot risk data leakage, compliance violations, or misunderstandings. Mistranslated documents can reduce trust and have strategic implications.\"78message = f\"Translate everything that follows into {target_language}:\\n\\n{content_to_translate}\"9response = co.chat(10 model=\"command-a-translate-08-2025\",11 messages=[{\"role\": \"user\", \"content\": message}],12)13print(response.message.content[0].text)\n```\n\nExample:\n```text\n1Las empresas dependen de la traducción para algunos de sus documentos más sensibles y críticos para su negocio y no pueden permitirse el riesgo de fugas de datos, incumplimientos normativos o malentendidos. Los documentos mal traducidos pueden reducir la confianza y tener implicaciones estratégicas.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.302Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":289}}91{"id":"doc-cohere_s_command_r_model_cohere-ea775ede","source":"documentation","title":"Cohere's Command R Model | Cohere","url":"https://docs.cohere.com/docs/command-r","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\nÉcris une description de produit pour une voiture électrique en 50 à 75 mots\n```\n\nExample:\n```text\nDécouvrez la voiture électrique qui va révolutionner votre façon de conduire.Avec son design élégant, cette voiture offre une expérience de conduite uniqueavec une accélération puissante et une autonomie impressionnante. Satechnologie avancée vous garantit une charge rapide et une fiabilité inégalée.Avec sa conception innovante et durable, cette voiture est parfaite pour les trajets urbains et les longues distances. Profitez d'une conduite silencieuseet vivez l'expérience de la voiture électrique!\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.303Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":202}}92{"id":"doc-rag_streaming_cohere-fffe6990","source":"documentation","title":"RAG Streaming | Cohere","url":"https://docs.cohere.com/docs/rag-streaming","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1\"Where do the tallest penguins live?\"23type='message-start' id='d93f187e-e9ac-44a9-a2d9-bdf2d65fee94' delta=ChatMessageStartEventDelta(message=ChatMessageStartEventDeltaMessage(role='assistant', content=[], tool_plan='', tool_calls=[], citations=[])) 4 --------------------------------------------------5type='content-start' index=0 delta=ChatContentStartEventDelta(message=ChatContentStartEventDeltaMessage(content=ChatContentStartEventDeltaMessageContent(text='', type='text'))) 6 --------------------------------------------------7type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='The'))) logprobs=None 8 --------------------------------------------------9type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' tallest'))) logprobs=None 10 --------------------------------------------------11type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' penguins'))) logprobs=None 12 --------------------------------------------------13type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' are'))) logprobs=None 14 --------------------------------------------------15type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' the'))) logprobs=None 16 --------------------------------------------------17type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' Emperor'))) logprobs=None 18 --------------------------------------------------19type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' penguins'))) logprobs=None 20 --------------------------------------------------21type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='.'))) logprobs=None 22 --------------------------------------------------23type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' They'))) logprobs=None 24 --------------------------------------------------25type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' only'))) logprobs=None 26 --------------------------------------------------27type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' live'))) logprobs=None 28 --------------------------------------------------29type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' in'))) logprobs=None 30 --------------------------------------------------31type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' Antarctica'))) logprobs=None 32 --------------------------------------------------33type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='.'))) logprobs=None 34 --------------------------------------------------35type='citation-start' index=0 delta=CitationStartEventDelta(message=CitationStartEventDeltaMessage(citations=Citation(start=29, end=46, text='Emperor penguins.', sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'snippet': 'Emperor penguins are the tallest.', 'title': 'Tall penguins'})], type='TEXT_CONTENT'))) 36 --------------------------------------------------37type='citation-end' index=0 38 --------------------------------------------------39type='citation-start' index=1 delta=CitationStartEventDelta(message=CitationStartEventDeltaMessage(citations=Citation(start=65, end=76, text='Antarctica.', sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'snippet': 'Emperor penguins only live in Antarctica.', 'title': 'Penguin habitats'})], type='TEXT_CONTENT'))) 40 --------------------------------------------------41type='citation-end' index=1 42 --------------------------------------------------43type='content-end' index=0 44 --------------------------------------------------45type='message-end' id=None delta=ChatMessageEndEventDelta(finish_reason='COMPLETE', usage=Usage(billed_units=UsageBilledUnits(input_tokens=34.0, output_tokens=14.0, search_units=None, classifications=None), tokens=UsageTokens(input_tokens=721.0, output_tokens=59.0))) 46 --------------------------------------------------\n```\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere34co = cohere.ClientV2(5 \"COHERE_API_KEY\"6) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1documents = [2 {3 \"data\": {4 \"title\": \"Tall penguins\",5 \"snippet\": \"Emperor penguins are the tallest.\",6 }7 },8 {9 \"data\": {10 \"title\": \"Penguin habitats\",11 \"snippet\": \"Emperor penguins only live in Antarctica.\",12 }13 },14]\n```\n\nExample:\n```text\n1messages = [2 {\"role\": \"user\", \"content\": \"Where do the tallest penguins live?\"}3]45response = co.chat_stream(6 model=\"command-a-plus-05-2026\",7 messages=messages,8 documents=documents,9)1011response_text = \"\"12citations = []13for chunk in response:14 if chunk:15 if chunk.type == \"content-delta\":16 response_text += chunk.delta.message.content.text17 print(chunk.delta.message.content.text, end=\"\")18 if chunk.type == \"citation-start\":19 citations.append(chunk.delta.message.citations)2021for citation in citations:22 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1The tallest penguins are the Emperor penguins, which only live in Antarctica.23start=29 end=45 text='Emperor penguins' sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'snippet': 'Emperor penguins are the tallest.', 'title': 'Tall penguins'})] type='TEXT_CONTENT' 45start=66 end=77 text='Antarctica.' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'snippet': 'Emperor penguins only live in Antarctica.', 'title': 'Penguin habitats'})] type='TEXT_CONTENT'\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.303Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":5,"totalLines":28,"estimatedTokens":1780}}93{"id":"doc-how_do_structured_outputs_work_cohere-c974abb4","source":"documentation","title":"How do Structured Outputs Work? | Cohere","url":"https://docs.cohere.com/docs/structured-outputs","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"YOUR API KEY\")45res = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Generate a JSON describing a person, with the fields 'name' and 'age'\",11 }12 ],13 response_format={\"type\": \"json_object\"},14)1516print(res.message.content[0].text)\n```\n\nExample:\n```text\n# Example response{ \"name\": \"Emma Johnson\", \"age\": 32}\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"YOUR API KEY\")45res = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Generate a JSON describing a book, with the fields 'title' and 'author' and 'publication_year'\",11 }12 ],13 response_format={14 \"type\": \"json_object\",15 \"schema\": {16 \"type\": \"object\",17 \"properties\": {18 \"title\": {\"type\": \"string\"},19 \"author\": {\"type\": \"string\"},20 \"publication_year\": {\"type\": \"integer\"},21 },22 \"required\": [\"title\", \"author\", \"publication_year\"],23 },24 },25)2627print(res.message.content[0].text)\n```\n\nExample:\n```text\n# Example response{ \"title\": \"The Great Gatsby\", \"author\": \"F. Scott Fitzgerald\", \"publication_year\": 1925}\n```\n\nExample:\n```text\n1cohere_api_key = os.getenv(\"cohere_api_key\")2co = cohere.ClientV2(cohere_api_key)3response = co.chat(4 response_format={5 \"type\": \"json_object\",6 \"schema\": {7 \"type\": \"object\",8 \"properties\": {9 \"actions\": {10 \"type\": \"array\",11 \"items\": {12 \"type\": \"object\",13 \"properties\": {14 \"japanese\": {\"type\": \"string\"},15 \"romaji\": {\"type\": \"string\"},16 \"english\": {\"type\": \"string\"},17 },18 \"required\": [\"japanese\", \"romaji\", \"english\"],19 },20 }21 },22 \"required\": [\"actions\"],23 },24 },25 model=\"command-a-plus-05-2026\",26 messages=[27 {28 \"role\": \"user\",29 \"content\": \"Generate a JSON array of objects with the following fields: japanese, romaji, english. These actions should be japanese verbs provided in the dictionary form.\",30 },31 ],32)33return json.loads(response.message.content[0].text)\n```\n\nExample:\n```text\n1{2 \"actions\": [3 {\"japanese\": \"いこう\", \"romaji\": \"ikou\", \"english\": \"onward\"},4 {\"japanese\": \"探す\", \"romaji\": \"sagasu\", \"english\": \"search\"},5 {\"japanese\": \"話す\", \"romaji\": \"hanasu\", \"english\": \"talk\"}6 ]7}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"get_weather\",6 \"description\" : \"Gets the weather of a given location\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"location\": {11 \"type\" : \"string\",12 \"description\": \"The location to get weather.\"13 }14 },15 \"required\": [\"location\"]16 }17 }18 },19]2021response = co.chat(model=\"command-r7b-12-2024\",22 messages=[{\"role\": \"user\", \"content\": \"What's the weather in Toronto?\"}],23 tools=tools,24 strict_tools=True)2526print(response.message.tool_calls)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.304Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":7,"totalLines":38,"estimatedTokens":951}}94{"id":"doc-milvus_and_cohere_integration_guide_cohere-6239b520","source":"documentation","title":"Milvus and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/milvus-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.304Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}95{"id":"doc-retrieval_augmented_generation_rag_cohere-fda0bb24","source":"documentation","title":"Retrieval Augmented Generation (RAG) | Cohere","url":"https://docs.cohere.com/docs/retrieval-augmented-generation-rag","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere34co = cohere.ClientV2(5 \"COHERE_API_KEY\"6) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1# Retrieve the documents2documents = [3 {4 \"data\": {5 \"title\": \"Tall penguins\",6 \"snippet\": \"Emperor penguins are the tallest.\",7 }8 },9 {10 \"data\": {11 \"title\": \"Penguin habitats\",12 \"snippet\": \"Emperor penguins only live in Antarctica.\",13 }14 },15 {16 \"data\": {17 \"title\": \"What are animals?\",18 \"snippet\": \"Animals are different from plants.\",19 }20 },21]2223# Add the user message24message = \"Where do the tallest penguins live?\"2526messages = [{\"role\": \"user\", \"content\": message}]2728response = co.chat(29 model=\"command-a-plus-05-2026\",30 messages=messages,31 documents=documents,32)3334print(response.message.content[0].text)3536print(response.message.citations)\n```\n\nExample:\n```text\n$curl --request POST \\> --url https://api.cohere.ai/v2/chat \\> --header 'accept: application/json' \\> --header 'content-type: application/json' \\> --header \"Authorization: bearer $CO_API_KEY\" \\> --data '{> \"model\": \"command-a-plus-05-2026\",> \"messages\": [> {> \"role\": \"user\",> \"content\": \"Where do the tallest penguins live?\"> }> ],> \"documents\": [> {> \"data\": {> \"title\": \"Tall penguins\",> \"snippet\": \"Emperor penguins are the tallest.\"> }> },> {> \"data\": {> \"title\": \"Penguin habitats\",> \"snippet\": \"Emperor penguins only live in Antarctica.\"> }> },> {> \"data\": {> \"title\": \"What are animals?\",> \"snippet\": \"Animals are different from plants.\"> }> }> ]> }'\n```\n\nExample:\n```text\n1# response.message.content[0].text2Emperor penguins are the tallest penguins. They only live in Antarctica.34# response.message.citations5[Citation(start=0,6 end=16, 7 text='Emperor penguins', 8 sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'snippet': 'Emperor penguins are the tallest.', 'title': 'Tall penguins'})]), 9Citation(start=25, 10 end=42, 11 text='tallest penguins.', 12 sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'snippet': 'Emperor penguins are the tallest.', 'title': 'Tall penguins'})]), 13Citation(start=61, 14 end=72, 15 text='Antarctica.',16 sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'snippet': 'Emperor penguins only live in Antarctica.', 'title': 'Penguin habitats'})])]\n```\n\nExample:\n```text\n1message = \"Who is more popular: Nsync or Backstreet Boys?\"23# Define the query generation tool4query_gen_tool = [5 {6 \"type\": \"function\",7 \"function\": {8 \"name\": \"internet_search\",9 \"description\": \"Returns a list of relevant document snippets for a textual query retrieved from the internet\",10 \"parameters\": {11 \"type\": \"object\",12 \"properties\": {13 \"queries\": {14 \"type\": \"array\",15 \"items\": {\"type\": \"string\"},16 \"description\": \"a list of queries to search the internet with.\",17 }18 },19 \"required\": [\"queries\"],20 },21 },22 }23]2425# Define a system message to optimize search query generation26instructions = \"Write a search query that will find helpful information for answering the user's question accurately. If you need more than one search query, write a list of search queries. If you decide that a search is very unlikely to find information that would be useful in constructing a response to the user, you should instead directly answer.\"2728# Generate search queries (if any)29import json3031search_queries = []3233res = co.chat(34 model=\"command-a-plus-05-2026\",35 messages=[36 {\"role\": \"system\", \"content\": instructions},37 {\"role\": \"user\", \"content\": message},38 ],39 tools=query_gen_tool,40)4142if res.message.tool_calls:43 for tc in res.message.tool_calls:44 queries = json.loads(tc.function.arguments)[\"queries\"]45 search_queries.extend(queries)4647print(search_queries)\n```\n\nExample:\n```text\n$curl --request POST \\> --url https://api.cohere.ai/v2/chat \\> --header 'accept: application/json' \\> --header 'content-type: application/json' \\> --header \"Authorization: bearer $CO_API_KEY\" \\> --data '{> \"model\": \"command-a-plus-05-2026\",> \"messages\": [> {> \"role\": \"system\",> \"content\": \"Write a search query that will find helpful information for answering the user'\\''s question accurately. If you need more than one search query, write a list of search queries. If you decide that a search is very unlikely to find information that would be useful in constructing a response to the user, you should instead directly answer.\"> },> {> \"role\": \"user\",> \"content\": \"Who is more popular: Nsync or Backstreet Boys?\"> }> ],> \"tools\": [> {> \"type\": \"function\",> \"function\": {> \"name\": \"internet_search\",> \"description\": \"Returns a list of relevant document snippets for a textual query retrieved from the internet\",> \"parameters\": {> \"type\": \"object\",> \"properties\": {> \"queries\": {> \"type\": \"array\",> \"items\": {\"type\": \"string\"},> \"description\": \"a list of queries to search the internet with.\"> }> },> \"required\": [\"queries\"]> }> }> }> ]> }'\n```\n\nExample:\n```text\n# Sample response['popularity of NSync', 'popularity of Backstreet Boys']\n```\n\nExample:\n```text\n1instructions = \"Write a search query that will find helpful information for answering the user's question accurately. If you need more than one search query, write a list of search queries. If you decide that a search is very unlikely to find information that would be useful in constructing a response to the user, you should instead directly answer.\"\n```\n\nExample:\n```text\n1['NSync popularity', 'Backstreet Boys popularity', 'NSync vs Backstreet Boys popularity comparison', 'Which boy band is more popular NSync or Backstreet Boys', 'NSync and Backstreet Boys fan base size comparison', 'Who has sold more albums NSync or Backstreet Boys', 'NSync and Backstreet Boys chart performance comparison']\n```\n\nExample:\n```text\n1documents = [2 {3 \"data\": {4 \"title\": \"Tall penguins\",5 \"snippet\": \"Emperor penguins are the tallest.\",6 }7 }8]\n```\n\nExample:\n```text\n1documents = [2 {3 \"data\": {4 \"title\": \"CSPC: Backstreet Boys Popularity Analysis - ChartMasters\",5 \"snippet\": \"↓ Skip to Main Content\\n\\nMusic industry – One step closer to being accurate\\n\\nCSPC: Backstreet Boys Popularity Analysis\\n\\nHernán Lopez Posted on February 9, 2017 Posted in CSPC 72 Comments Tagged with Backstreet Boys, Boy band\\n\\nAt one point, Backstreet Boys defined success: massive albums sales across the globe, great singles sales, plenty of chart topping releases, hugely hyped tours and tremendous media coverage.\\n\\nIt is true that they benefited from extraordinarily good market conditions in all markets. After all, the all-time record year for the music business, as far as revenues in billion dollars are concerned, was actually 1999. That is, back when this five men group was at its peak.\",6 }7 },8 {9 \"data\": {10 \"title\": \"CSPC: NSYNC Popularity Analysis - ChartMasters\",11 \"snippet\": \"↓ Skip to Main Content\\n\\nMusic industry – One step closer to being accurate\\n\\nCSPC: NSYNC Popularity Analysis\\n\\nMJD Posted on February 9, 2018 Posted in CSPC 27 Comments Tagged with Boy band, N'Sync\\n\\nAt the turn of the millennium three teen acts were huge in the US, the Backstreet Boys, Britney Spears and NSYNC. The latter is the only one we haven’t study so far. It took 15 years and Adele to break their record of 2,4 million units sold of No Strings Attached in its first week alone.\\n\\nIt wasn’t a fluke, as the second fastest selling album of the Soundscan era prior 2015, was also theirs since Celebrity debuted with 1,88 million units sold.\",12 }13 },14 {15 \"data\": {16 \"title\": \"CSPC: Backstreet Boys Popularity Analysis - ChartMasters\",17 \"snippet\": \" 1997, 1998, 2000 and 2001 also rank amongst some of the very best years.\\n\\nYet the way many music consumers – especially teenagers and young women’s – embraced their output deserves its own chapter. If Jonas Brothers and more recently One Direction reached a great level of popularity during the past decade, the type of success achieved by Backstreet Boys is in a completely different level as they really dominated the business for a few years all over the world, including in some countries that were traditionally hard to penetrate for Western artists.\\n\\nWe will try to analyze the extent of that hegemony with this new article with final results which will more than surprise many readers.\",18 }19 },20 {21 \"data\": {22 \"title\": \"CSPC: NSYNC Popularity Analysis - ChartMasters\",23 \"snippet\": \" Was the teen group led by Justin Timberlake really that big? Was it only in the US where they found success? Or were they a global phenomenon?\\n\\nAs usual, I’ll be using the Commensurate Sales to Popularity Concept in order to relevantly gauge their results. This concept will not only bring you sales information for all NSYNC‘s albums, physical and download singles, as well as audio and video streaming, but it will also determine their true popularity. If you are not yet familiar with the CSPC method, the next page explains it with a short video. I fully recommend watching the video before getting into the sales figures.\",24 }25 },26]2728# Add the user message29message = \"Who is more popular: Nsync or Backstreet Boys?\"30messages = [{\"role\": \"user\", \"content\": message}]3132response = co.chat(33 model=\"command-a-plus-05-2026\",34 messages=messages,35 documents=documents,36)3738print(response.message.content[0].text)\n```\n\nExample:\n```text\n$curl --request POST \\> --url https://api.cohere.ai/v2/chat \\> --header 'accept: application/json' \\> --header 'content-type: application/json' \\> --header \"Authorization: bearer $CO_API_KEY\" \\> --data '{> \"model\": \"command-a-plus-05-2026\",> \"messages\": [> {> \"role\": \"user\",> \"content\": \"Who is more popular: Nsync or Backstreet Boys?\"> }> ],> \"documents\": [> {> \"data\": {> \"title\": \"CSPC: Backstreet Boys Popularity Analysis - ChartMasters\",> \"snippet\": \"At one point, Backstreet Boys defined success: massive albums sales across the globe...\"> }> },> {> \"data\": {> \"title\": \"CSPC: NSYNC Popularity Analysis - ChartMasters\",> \"snippet\": \"At the turn of the millennium three teen acts were huge in the US...\"> }> }> ]> }'\n```\n\nExample:\n```text\n1Both NSYNC and Backstreet Boys were huge in the US at the turn of the millennium. However, Backstreet Boys achieved a greater level of success than NSYNC. They dominated the music business for a few years all over the world, including in some countries that were traditionally hard to penetrate for Western artists. Their success included massive album sales across the globe, great singles sales, plenty of chart-topping releases, hugely hyped tours and tremendous media coverage.\n```\n\nExample:\n```text\n1print(response.message.citations)\n```\n\nExample:\n```text\n1# (truncated for brevity)2[Citation(start=36, 3 end=81, 4 text='huge in the US at the turn of the millennium.', 5 sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'snippet': \"↓ Skip to Main Content\\n\\nMusic industry – One step closer ...\", 'title': 'CSPC: NSYNC Popularity Analysis - ChartMasters'})]),6Citation(start=107, 7 end=154, 8 text='achieved a greater level of success than NSYNC.', 9 sources=[DocumentSource(type='document', id='doc:2', document={'id': 'doc:2', 'snippet': ' 1997, 1998, 2000 and 2001 also rank amongst some of the very best ...', 'title': 'CSPC: Backstreet Boys Popularity Analysis - ChartMasters'})]),10Citation(start=160, 11 end=223,12 ...13...]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.305Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":3223}}96{"id":"doc-basic_usage_of_tool_use_function_calling_cohere-ea1af91b","source":"documentation","title":"Basic usage of tool use (function calling) | Cohere","url":"https://docs.cohere.com/docs/tool-use-overview","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install -U cohere # Do this if you don't already have the Cohere client installed.2import json34import cohere567def search_docs(query: str, top_k: int = 3):8 # Implement your retrieval logic here (vector DB, keyword search, etc.)9 # For simplicity, we'll return a few hardcoded \"documents\".10 return [11 {12 \"title\": \"Cohere API v2 - Chat\",13 \"url\": \"https://docs.cohere.com/reference/chat\",14 \"text\": \"Use the Chat endpoint to generate responses and optionally call tools.\",15 },16 {17 \"title\": \"Tool use (function calling) overview\",18 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",19 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",20 },21 {22 \"title\": \"Structured outputs\",23 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",24 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",25 },26 ][:top_k]272829functions_map = {\"search_docs\": search_docs}3031tools = [32 {33 \"type\": \"function\",34 \"function\": {35 \"name\": \"search_docs\",36 \"description\": \"Search documentation and return relevant snippets as documents.\",37 \"parameters\": {38 \"type\": \"object\",39 \"properties\": {40 \"query\": {41 \"type\": \"string\",42 \"description\": \"The search query to look up in the docs.\",43 },44 \"top_k\": {45 \"type\": \"integer\",46 \"description\": \"How many documents to return.\",47 },48 },49 \"required\": [\"query\"],50 },51 },52 }53]5455co = cohere.ClientV2(\"COHERE_API_KEY\")5657# Step 1: user message58messages = [59 {60 \"role\": \"user\",61 \"content\": \"How does tool use work in Cohere? Please cite your sources.\",62 }63]6465# Step 2: model generates tool calls66response = co.chat(67 model=\"command-a-plus-05-2026\", messages=messages, tools=tools68)6970if response.message.tool_calls:71 messages.append(response.message)7273 # Step 3: application executes tools and sends tool results back74 for tc in response.message.tool_calls:75 tool_result = functions_map[tc.function.name](76 **json.loads(tc.function.arguments)77 )7879 tool_content = []80 for data in tool_result:81 tool_content.append(82 {83 \"type\": \"document\",84 \"document\": {\"data\": json.dumps(data)},85 }86 )8788 messages.append(89 {90 \"role\": \"tool\",91 \"tool_call_id\": tc.id,92 \"content\": tool_content,93 }94 )9596# Step 4: model generates a response grounded in tool results (with citations)97response = co.chat(98 model=\"command-a-plus-05-2026\", messages=messages, tools=tools99)100101print(response.message.content[0].text)102print(response.message.citations)\n```\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere34co = cohere.ClientV2(5 \"COHERE_API_KEY\"6) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1def search_docs(query, top_k=3):2 # Implement any retrieval logic here (vector DB, keyword search, etc.)3 return [4 {5 \"title\": \"Tool use (function calling) overview\",6 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",7 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",8 },9 {10 \"title\": \"Chat API reference (v2)\",11 \"url\": \"https://docs.cohere.com/reference/chat\",12 \"text\": \"Use the Chat endpoint to generate responses and optionally call tools.\",13 },14 {15 \"title\": \"Structured outputs\",16 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",17 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",18 },19 ][:top_k]20 # Return a string or a list of objects. In Step 3 below, we'll wrap each object into a `document`21 # content block so the model can cite specific tool results.222324functions_map = {\"search_docs\": search_docs}\n```\n\nExample:\n```text\n1# Example: String2docs_search_results = \"Tool use connects models to external tools like search engines and APIs.\"34# Example: List of objects5docs_search_results = [6 {7 \"title\": \"Tool use (function calling) overview\",8 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",9 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",10 },11 {12 \"title\": \"Structured outputs\",13 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",14 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",15 },16]\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"search_docs\",6 \"description\": \"Search documentation and return relevant snippets as documents.\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"query\": {11 \"type\": \"string\",12 \"description\": \"The search query to look up in the docs.\",13 },14 \"top_k\": {15 \"type\": \"integer\",16 \"description\": \"How many documents to return.\",17 },18 },19 \"required\": [\"query\"],20 },21 },22 },23]\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"user\",4 \"content\": \"How does tool use work in Cohere? Please cite your sources.\",5 }6]\n```\n\nExample:\n```text\n1system_message = \"\"\"## Task & Context2You help people answer their questions and other requests interactively. You will be asked a very wide array of requests on all kinds of topics. You will be equipped with a wide range of search engines or similar tools to help you, which you use to research your answer. You should focus on serving the user's needs as best you can, which will be wide-ranging.34## Style Guide5Unless the user asks for a different style of answer, you should answer in full sentences, using proper grammar and spelling.6\"\"\"78messages = [9 {\"role\": \"system\", \"content\": system_message},10 {11 \"role\": \"user\",12 \"content\": \"How does tool use work in Cohere? Please cite your sources.\",13 },14]\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\", messages=messages, tools=tools3)45if response.message.tool_calls:6 messages.append(response.message)7 print(response.message.tool_plan, \"\\n\")8 print(response.message.tool_calls)\n```\n\nExample:\n```text\n1I will search the docs for how tool use works in Cohere.23[4 ToolCallV2(5 id=\"search_docs_1byjy32y4hvq\",6 type=\"function\",7 function=ToolCallV2Function(8 name=\"search_docs\", arguments='{\"query\":\"tool use Cohere\",\"top_k\":3}'9 ),10 )11]\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"user\",4 \"content\": \"Find docs about tool use and structured outputs.\",5 },6 {7 \"role\": \"assistant\",8 \"tool_plan\": \"I will search the docs for tool use and structured outputs.\",9 \"tool_calls\": [10 ToolCallV2(11 id=\"search_docs_dkf0akqdazjb\",12 type=\"function\",13 function=ToolCallV2Function(14 name=\"search_docs\",15 arguments='{\"query\":\"tool use\",\"top_k\":3}',16 ),17 ),18 ToolCallV2(19 id=\"search_docs_gh65bt2tcdy1\",20 type=\"function\",21 function=ToolCallV2Function(22 name=\"search_docs\",23 arguments='{\"query\":\"structured outputs\",\"top_k\":3}',24 ),25 ),26 ],27 },28]\n```\n\nExample:\n```text\n1import json23if response.message.tool_calls:4 for tc in response.message.tool_calls:5 tool_result = functions_map[tc.function.name](6 **json.loads(tc.function.arguments)7 )8 tool_content = []9 for data in tool_result:10 # Optional: the \"document\" object can take an \"id\" field for use in citations, otherwise auto-generated11 tool_content.append(12 {13 \"type\": \"document\",14 \"document\": {\"data\": json.dumps(data)},15 }16 )17 messages.append(18 {19 \"role\": \"tool\",20 \"tool_call_id\": tc.id,21 \"content\": tool_content,22 }23 )\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\", messages=messages, tools=tools3)45messages.append(6 {\"role\": \"assistant\", \"content\": response.message.content[0].text}7)89print(response.message.content[0].text)\n```\n\nExample:\n```text\n1Tool use lets models call external tools (like doc search) and then answer using the tool results, with citations.\n```\n\nExample:\n```text\n1print(response.message.citations)\n```\n\nExample:\n```text\n1[Citation(start=0, end=8, text='Tool use', sources=[ToolSource(type='tool', id='search_docs_1byjy32y4hvq:0', tool_output={'title': 'Tool use (function calling) overview', 'url': 'https://docs.cohere.com/v2/docs/tool-use-overview', 'text': 'Tool use connects models to external tools like search engines and APIs.'})], type='TEXT_CONTENT')]\n```\n\nExample:\n```text\n1for message in messages:2 print(message, \"\\n\")\n```\n\nExample:\n```text\n1{ 2 \"role\": \"user\", 3 \"content\": \"How does tool use work in Cohere? Please cite your sources.\"4}56{7 \"role\": \"assistant\",8 \"tool_plan\": \"I will search the docs for how tool use works in Cohere.\",9 \"tool_calls\": [10 ToolCallV2(11 id=\"search_docs_1byjy32y4hvq\",12 type=\"function\",13 function=ToolCallV2Function(14 name=\"search_docs\", arguments='{\"query\":\"tool use Cohere\",\"top_k\":3}'15 ),16 )17 ],18}1920{21 \"role\": \"tool\",22 \"tool_call_id\": \"search_docs_1byjy32y4hvq\",23 \"content\": [{\"type\": \"document\", \"document\": {\"data\": \"{\\\"title\\\":\\\"Tool use (function calling) overview\\\",\\\"url\\\":\\\"https://docs.cohere.com/v2/docs/tool-use-overview\\\",\\\"text\\\":\\\"Tool use connects models to external tools like search engines and APIs.\\\"}\"}}],24}2526{ 27 \"role\": \"assistant\", 28 \"content\": \"Tool use lets models call external tools (like doc search) and then answer using the tool results, with citations.\"29}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.306Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":17,"totalLines":88,"estimatedTokens":2788}}97{"id":"doc-usage_patterns_for_tool_use_function_calling_coh-6359f652","source":"documentation","title":"Usage patterns for tool use (function calling) | Cohere","url":"https://docs.cohere.com/docs/tool-use-usage-patterns","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere34co = cohere.ClientV2(5 \"COHERE_API_KEY\"6) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1def search_docs(query, top_k=3):2 # Implement any retrieval logic here (vector DB, keyword search, etc.)3 return [4 {5 \"title\": \"Tool use (function calling) overview\",6 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",7 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",8 },9 {10 \"title\": \"Structured outputs\",11 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",12 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",13 },14 {15 \"title\": \"Chat API reference (v2)\",16 \"url\": \"https://docs.cohere.com/reference/chat\",17 \"text\": \"Use the Chat endpoint to generate responses and optionally call tools.\",18 },19 ][:top_k]20 # Return a string or a list of objects. In Step 3, we'll wrap each object into a `document` content block.212223functions_map = {\"search_docs\": search_docs}2425tools = [26 {27 \"type\": \"function\",28 \"function\": {29 \"name\": \"search_docs\",30 \"description\": \"Search documentation and return relevant snippets as documents.\",31 \"parameters\": {32 \"type\": \"object\",33 \"properties\": {34 \"query\": {35 \"type\": \"string\",36 \"description\": \"The search query to look up in the docs.\",37 },38 \"top_k\": {39 \"type\": \"integer\",40 \"description\": \"How many documents to return.\",41 },42 },43 \"required\": [\"query\"],44 },45 },46 },47]\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"user\",4 \"content\": \"Find docs about tool use and structured outputs.\",5 }6]78response = co.chat(9 model=\"command-a-plus-05-2026\", messages=messages, tools=tools10)1112if response.message.tool_calls:13 messages.append(response.message)14 print(response.message.tool_plan, \"\\n\")15 print(response.message.tool_calls)\n```\n\nExample:\n```text\n1I will search the docs for tool use and structured outputs.23[4 ToolCallV2(5 id=\"search_docs_9b0nr4kg58a8\",6 type=\"function\",7 function=ToolCallV2Function(8 name=\"search_docs\", arguments='{\"query\":\"tool use\",\"top_k\":3}'9 ),10 ),11 ToolCallV2(12 id=\"search_docs_0qq0mz9gwnqr\",13 type=\"function\",14 function=ToolCallV2Function(15 name=\"search_docs\", arguments='{\"query\":\"structured outputs\",\"top_k\":3}'16 ),17 ),18]\n```\n\nExample:\n```text\n1import json23if response.message.tool_calls:4 for tc in response.message.tool_calls:5 tool_result = functions_map[tc.function.name](6 **json.loads(tc.function.arguments)7 )8 tool_content = []9 for data in tool_result:10 # Optional: the \"document\" object can take an \"id\" field for use in citations, otherwise auto-generated11 tool_content.append(12 {13 \"type\": \"document\",14 \"document\": {\"data\": json.dumps(data)},15 }16 )17 messages.append(18 {19 \"role\": \"tool\",20 \"tool_call_id\": tc.id,21 \"content\": tool_content,22 }23 )\n```\n\nExample:\n```text\n1messages = [{\"role\": \"user\", \"content\": \"What's 2+2?\"}]23response = co.chat(4 model=\"command-a-plus-05-2026\", messages=messages, tools=tools5)67if response.message.tool_calls:8 print(response.message.tool_plan, \"\\n\")9 print(response.message.tool_calls)1011else:12 print(response.message.content[0].text)\n```\n\nExample:\n```text\n1The answer to 2+2 is 4.\n```\n\nExample:\n```text\n1def search_docs(query, top_k=3):2 # Implement any retrieval logic here (vector DB, keyword search, etc.)3 return [4 {5 \"title\": \"Tool use (function calling) overview\",6 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",7 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",8 },9 {10 \"title\": \"Usage patterns for tool use\",11 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-usage-patterns\",12 \"text\": \"Common patterns include parallel tool calling, multi-step tool use, and more.\",13 },14 {15 \"title\": \"Structured outputs\",16 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",17 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",18 },19 ][:top_k]202122functions_map = {\"search_docs\": search_docs}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"search_docs\",6 \"description\": \"Search documentation and return relevant snippets as documents.\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"query\": {11 \"type\": \"string\",12 \"description\": \"The search query to look up in the docs.\",13 },14 \"top_k\": {15 \"type\": \"integer\",16 \"description\": \"How many documents to return.\",17 },18 },19 \"required\": [\"query\"],20 },21 },22 },23]\n```\n\nExample:\n```text\n1import json23# Step 1: Get the user message4messages = [5 {6 \"role\": \"user\",7 \"content\": \"Explain how tool use works and how to force tool usage. Please cite your sources.\",8 }9]1011# Step 2: Generate tool calls (if any)12model = \"command-a-plus-05-2026\"13response = co.chat(14 model=model, messages=messages, tools=tools, temperature=0.315)1617while response.message.tool_calls:18 print(\"TOOL PLAN:\")19 print(response.message.tool_plan, \"\\n\")20 print(\"TOOL CALLS:\")21 for tc in response.message.tool_calls:22 print(23 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"24 )25 print(\"=\" * 50)2627 messages.append(response.message)2829 # Step 3: Get tool results30 print(\"TOOL RESULT:\")31 for tc in response.message.tool_calls:32 tool_result = functions_map[tc.function.name](33 **json.loads(tc.function.arguments)34 )35 tool_content = []36 print(tool_result)37 for data in tool_result:38 # Optional: the \"document\" object can take an \"id\" field for use in citations, otherwise auto-generated39 tool_content.append(40 {41 \"type\": \"document\",42 \"document\": {\"data\": json.dumps(data)},43 }44 )45 messages.append(46 {47 \"role\": \"tool\",48 \"tool_call_id\": tc.id,49 \"content\": tool_content,50 }51 )5253 # Step 4: Generate response and citations54 response = co.chat(55 model=model,56 messages=messages,57 tools=tools,58 temperature=0.1,59 )6061messages.append(62 {63 \"role\": \"assistant\",64 \"content\": response.message.content[0].text,65 }66)6768# Print final response69print(\"RESPONSE:\")70print(response.message.content[0].text)71print(\"=\" * 50)7273# Print citations (if any)74verbose_source = (75 True # Change to True to display the contents of a source76)77if response.message.citations:78 print(\"CITATIONS:\\n\")79 for citation in response.message.citations:80 print(81 f\"Start: {citation.start}| End:{citation.end}| Text:'{citation.text}' \"82 )83 print(\"Sources:\")84 for idx, source in enumerate(citation.sources):85 print(f\"{idx+1}. {source.id}\")86 if verbose_source:87 print(f\"{source.tool_output}\")88 print(\"\\n\")\n```\n\nExample:\n```text\n1TOOL PLAN:2First, I will search the docs for how tool use works. Then, I will search for how to force tool usage (tool_choice).34TOOL CALLS:5Tool name: search_docs | Parameters: {\"query\":\"tool use\",\"top_k\":3}6==================================================7TOOL RESULT:8[{'title': 'Tool use (function calling) overview', 'url': 'https://docs.cohere.com/v2/docs/tool-use-overview', 'text': 'Tool use connects models to external tools like search engines and APIs.'}]9TOOL PLAN:10Now I'll search for how to force tool usage via the tool_choice parameter.1112TOOL CALLS:13Tool name: search_docs | Parameters: {\"query\":\"tool_choice REQUIRED NONE\",\"top_k\":3}14==================================================15TOOL RESULT:16[{'title': 'Usage patterns for tool use', 'url': 'https://docs.cohere.com/v2/docs/tool-use-usage-patterns', 'text': 'Common patterns include parallel tool calling, multi-step tool use, and more.'}]17RESPONSE:18Tool use lets models call external tools (like doc search) and then answer using tool results with citations. You can force tool usage with tool_choice=\"REQUIRED\" or force a direct response with tool_choice=\"NONE\".19==================================================20CITATIONS:2122Start: 126| End:135| Text:'tool_choice'23Sources:241. search_docs_p0dage9q1nv4:025{'title': 'Usage patterns for tool use', 'url': 'https://docs.cohere.com/v2/docs/tool-use-usage-patterns', 'text': 'Common patterns include parallel tool calling, multi-step tool use, and more.'}\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\",3 messages=messages,4 tools=tools,5 tool_choice=\"REQUIRED\" # optional, to force tool calls6 # tool_choice=\"NONE\" # optional, to force a direct response7)\n```\n\nExample:\n```text\n1from cohere import ToolCallV2, ToolCallV2Function23messages = [4 {5 \"role\": \"user\",6 \"content\": \"How does tool use work in Cohere? Please cite your sources.\",7 },8 {9 \"role\": \"assistant\",10 \"tool_plan\": \"I will search the docs for how tool use works in Cohere.\",11 \"tool_calls\": [12 ToolCallV2(13 id=\"search_docs_1byjy32y4hvq\",14 type=\"function\",15 function=ToolCallV2Function(16 name=\"search_docs\",17 arguments='{\"query\":\"tool use Cohere\",\"top_k\":3}',18 ),19 )20 ],21 },22 {23 \"role\": \"tool\",24 \"tool_call_id\": \"search_docs_1byjy32y4hvq\",25 \"content\": [26 {27 \"type\": \"document\",28 \"document\": {29 \"data\": '{\"title\":\"Tool use (function calling) overview\",\"url\":\"https://docs.cohere.com/v2/docs/tool-use-overview\",\"text\":\"Tool use connects models to external tools like search engines and APIs.\"}'30 },31 }32 ],33 },34 {35 \"role\": \"assistant\",36 \"content\": \"Tool use lets models call external tools (like doc search) and then answer using tool results with citations.\",37 },38]\n```\n\nExample:\n```text\n1messages.append(2 {\"role\": \"user\", \"content\": \"How do I force tool usage?\"}3)45response = co.chat(6 model=\"command-a-plus-05-2026\", messages=messages, tools=tools7)89if response.message.tool_calls:10 messages.append(response.message)11 print(response.message.tool_plan, \"\\n\")12 print(response.message.tool_calls)\n```\n\nExample:\n```text\n1I will search the docs for how to force tool usage using tool_choice.23[ToolCallV2(id='search_docs_8hwpm7d4wr14', type='function', function=ToolCallV2Function(name='search_docs', arguments='{\"query\":\"tool_choice REQUIRED NONE\",\"top_k\":3}'))]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.307Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":3038}}98{"id":"doc-end_to_end_example_of_rag_with_chat_embed_and_re-b3edc42f","source":"documentation","title":"End-to-end example of RAG with Chat, Embed, and Rerank | Cohere","url":"https://docs.cohere.com/docs/rag-complete-example","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere3import json4import numpy as np56co = cohere.ClientV2(7 \"COHERE_API_KEY\"8) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1message = \"How to get to know my teammates\"23# Define the query generation tool4query_gen_tool = [5 {6 \"type\": \"function\",7 \"function\": {8 \"name\": \"internet_search\",9 \"description\": \"Returns a list of relevant document snippets for a textual query retrieved from the internet\",10 \"parameters\": {11 \"type\": \"object\",12 \"properties\": {13 \"queries\": {14 \"type\": \"array\",15 \"items\": {\"type\": \"string\"},16 \"description\": \"a list of queries to search the internet with.\",17 }18 },19 \"required\": [\"queries\"],20 },21 },22 }23]2425# Define a system message to optimize search query generation26instructions = \"Write a search query that will find helpful information for answering the user's question accurately. If you need more than one search query, write a list of search queries. If you decide that a search is very unlikely to find information that would be useful in constructing a response to the user, you should instead directly answer.\"2728# Generate search queries (if any)29search_queries = []3031res = co.chat(32 model=\"command-a-plus-05-2026\",33 messages=[34 {\"role\": \"system\", \"content\": instructions},35 {\"role\": \"user\", \"content\": message},36 ],37 tools=query_gen_tool,38)3940if res.message.tool_calls:41 for tc in res.message.tool_calls:42 queries = json.loads(tc.function.arguments)[\"queries\"]43 search_queries.extend(queries)4445print(search_queries)\n```\n\nExample:\n```text\n1['how to get to know your teammates']\n```\n\nExample:\n```text\n1# Define the documents2documents = [3 {4 \"data\": {5 \"text\": \"Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.\"6 }7 },8 {9 \"data\": {10 \"text\": \"Finding Coffee Spots: For your caffeine fix, head to the break room's coffee machine or cross the street to the café for artisan coffee.\"11 }12 },13 {14 \"data\": {15 \"text\": \"Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!\"16 }17 },18 {19 \"data\": {20 \"text\": \"Working Hours Flexibility: We prioritize work-life balance. While our core hours are 9 AM to 5 PM, we offer flexibility to adjust as needed.\"21 }22 },23 {24 \"data\": {25 \"text\": \"Side Projects Policy: We encourage you to pursue your passions. Just be mindful of any potential conflicts of interest with our business.\"26 }27 },28 {29 \"data\": {30 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"31 }32 },33 {34 \"data\": {35 \"text\": \"Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.\"36 }37 },38 {39 \"data\": {40 \"text\": \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\"41 }42 },43 {44 \"data\": {45 \"text\": \"Performance Reviews Frequency: We conduct informal check-ins every quarter and formal performance reviews twice a year.\"46 }47 },48 {49 \"data\": {50 \"text\": \"Proposing New Ideas: Innovation is welcomed! Share your brilliant ideas at our weekly team meetings or directly with your team lead.\"51 }52 },53]5455# Embed the documents5657doc_emb = co.embed(58 model=\"embed-v4.0\",59 input_type=\"search_document\",60 texts=[doc[\"data\"][\"text\"] for doc in documents],61 embedding_types=[\"float\"],62).embeddings.float\n```\n\nExample:\n```text\n1# Embed the search query2query_emb = co.embed(3 model=\"embed-v4.0\",4 input_type=\"search_query\",5 texts=search_queries,6 embedding_types=[\"float\"],7).embeddings.float\n```\n\nExample:\n```text\n1# Compute dot product similarity and display results2n = 53scores = np.dot(query_emb, np.transpose(doc_emb))[0]4max_idx = np.argsort(-scores)[:n]56retrieved_documents = [documents[item] for item in max_idx]78for rank, idx in enumerate(max_idx):9 print(f\"Rank: {rank+1}\")10 print(f\"Score: {scores[idx]}\")11 print(f\"Document: {retrieved_documents[rank]}\\n\")\n```\n\nExample:\n```text\n1Rank: 12Score: 0.326534703608726553Document: {'data': {'text': 'Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!'}}45Rank: 26Score: 0.268518553522647867Document: {'data': {'text': 'Proposing New Ideas: Innovation is welcomed! Share your brilliant ideas at our weekly team meetings or directly with your team lead.'}}89Rank: 310Score: 0.258134197530414911Document: {'data': {'text': 'Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.'}}1213Rank: 414Score: 0.1863333673817846315Document: {'data': {'text': \"Finding Coffee Spots: For your caffeine fix, head to the break room's coffee machine or cross the street to the café for artisan coffee.\"}}1617Rank: 518Score: 0.1302239659568281419Document: {'data': {'text': 'Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.'}}\n```\n\nExample:\n```text\n1# Rerank the documents2results = co.rerank(3 model=\"rerank-v4.0-pro\",4 query=search_queries[0],5 documents=[doc[\"data\"][\"text\"] for doc in retrieved_documents],6 top_n=2,7)89# Display the reranking results1011for idx, result in enumerate(results.results):12 print(f\"Rank: {idx+1}\")13 print(f\"Score: {result.relevance_score}\")14 print(f\"Document: {retrieved_documents[result.index]}\\n\")1516reranked_documents = [17 retrieved_documents[result.index] for result in results.results18]\n```\n\nExample:\n```text\n1Rank: 12Score: 0.072722413Document: {'data': {'text': 'Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!'}}45Rank: 26Score: 0.0586741127Document: {'data': {'text': 'Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.'}}\n```\n\nExample:\n```text\n1messages = [{\"role\": \"user\", \"content\": message}]23# Generate the response4response = co.chat(5 model=\"command-a-plus-05-2026\",6 messages=messages,7 documents=reranked_documents,8)910# Display the response11print(response.message.content[0].text)1213# Display the citations and source documents14if response.message.citations:15 print(\"\\nCITATIONS:\")16 for citation in response.message.citations:17 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1To get to know your teammates, you can join relevant Slack channels to stay informed and engaged. You will receive an invite via email. You can also participate in team-building activities such as monthly outings and weekly game nights.23CITATIONS:4start=39 end=67 text='join relevant Slack channels' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.'})] type='TEXT_CONTENT' 56start=71 end=97 text='stay informed and engaged.' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.'})] type='TEXT_CONTENT' 78start=107 end=135 text='receive an invite via email.' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.'})] type='TEXT_CONTENT' 910start=164 end=188 text='team-building activities' sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!'})] type='TEXT_CONTENT' 1112start=197 end=236 text='monthly outings and weekly game nights.' sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!'})] type='TEXT_CONTENT'\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.308Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":11,"totalLines":58,"estimatedTokens":2319}}99{"id":"doc-rag_citations_cohere-54a1d9ba","source":"documentation","title":"RAG Citations | Cohere","url":"https://docs.cohere.com/docs/rag-citations","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere3import json45co = cohere.ClientV2(6 \"COHERE_API_KEY\"7) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1documents = [2 {3 \"data\": {4 \"title\": \"Tall penguins\",5 \"snippet\": \"Emperor penguins are the tallest.\",6 }7 },8 {9 \"data\": {10 \"title\": \"Penguin habitats\",11 \"snippet\": \"Emperor penguins only live in Antarctica.\",12 }13 },14]\n```\n\nExample:\n```text\n1messages = [2 {\"role\": \"user\", \"content\": \"Where do the tallest penguins live?\"}3]45response = co.chat(6 model=\"command-r-08-2024\",7 messages=messages,8 documents=documents,9)1011print(response.message.content[0].text)1213for citation in response.message.citations:14 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1The tallest penguins are the Emperor penguins. They only live in Antarctica.23start=29 end=46 text='Emperor penguins.' sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'snippet': 'Emperor penguins are the tallest.', 'title': 'Tall penguins'})] type='TEXT_CONTENT' 45start=65 end=76 text='Antarctica.' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'snippet': 'Emperor penguins only live in Antarctica.', 'title': 'Penguin habitats'})] type='TEXT_CONTENT'\n```\n\nExample:\n```text\n1messages = [2 {\"role\": \"user\", \"content\": \"Where do the tallest penguins live?\"}3]45response = co.chat_stream(6 model=\"command-a-plus-05-2026\",7 messages=messages,8 documents=documents,9)1011response_text = \"\"12citations = []13for chunk in response:14 if chunk:15 if chunk.type == \"content-delta\":16 response_text += chunk.delta.message.content.text17 print(chunk.delta.message.content.text, end=\"\")18 if chunk.type == \"citation-start\":19 citations.append(chunk.delta.message.citations)2021for citation in citations:22 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1The tallest penguins are the Emperor penguins, which only live in Antarctica.23start=29 end=45 text='Emperor penguins' sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'snippet': 'Emperor penguins are the tallest.', 'title': 'Tall penguins'})] type='TEXT_CONTENT' 45start=66 end=77 text='Antarctica.' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'snippet': 'Emperor penguins only live in Antarctica.', 'title': 'Penguin habitats'})] type='TEXT_CONTENT'\n```\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere3import json45co = cohere.ClientV2(6 \"COHERE_API_KEY\"7) # Get your free API key here: https://dashboard.cohere.com/api-keys89documents = [10 {11 \"data\": {12 \"title\": \"Tall penguins\",13 \"snippet\": \"Emperor penguins are the tallest.\",14 },15 \"id\": \"100\",16 },17 {18 \"data\": {19 \"title\": \"Penguin habitats\",20 \"snippet\": \"Emperor penguins only live in Antarctica.\",21 },22 \"id\": \"101\",23 },24]\n```\n\nExample:\n```text\n1messages = [2 {\"role\": \"user\", \"content\": \"Where do the tallest penguins live?\"}3]45response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=messages,8 documents=documents,9)1011print(response.message.content[0].text)\n```\n\nExample:\n```text\n1The tallest penguins are the Emperor penguins, which only live in Antarctica.23start=29 end=45 text='Emperor penguins' sources=[DocumentSource(type='document', id='100', document={'id': '100', 'snippet': 'Emperor penguins are the tallest.', 'title': 'Tall penguins'})] type='TEXT_CONTENT' 45start=66 end=77 text='Antarctica.' sources=[DocumentSource(type='document', id='101', document={'id': '101', 'snippet': 'Emperor penguins only live in Antarctica.', 'title': 'Penguin habitats'})] type='TEXT_CONTENT'\n```\n\nExample:\n```text\n1documents = [2 {3 \"data\": {4 \"title\": \"Tall penguins\",5 \"snippet\": \"Emperor penguins are the tallest.\",6 },7 \"id\": \"100\",8 },9 {10 \"data\": {11 \"title\": \"Penguin habitats\",12 \"snippet\": \"Emperor penguins only live in Antarctica.\",13 },14 \"id\": \"101\",15 },16]1718messages = [19 {\"role\": \"user\", \"content\": \"Where do the tallest penguins live?\"}20]2122response = co.chat_stream(23 model=\"command-a-plus-05-2026\",24 messages=messages,25 documents=documents,26 citation_options={\"mode\": \"fast\"},27)2829response_text = \"\"30citations = []31for chunk in response:32 if chunk:33 if chunk.type == \"content-delta\":34 response_text += chunk.delta.message.content.text35 print(chunk.delta.message.content.text, end=\"\")36 if chunk.type == \"citation-start\":37 citations.append(chunk.delta.message.citations)3839print(\"\\n\")40for citation in citations:41 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1The tallest penguins are the Emperor penguins. They live in Antarctica.23start=29 end=46 text='Emperor penguins.' sources=[DocumentSource(type='document', id='100', document={'id': '100', 'snippet': 'Emperor penguins are the tallest.', 'title': 'Tall penguins'})] type='TEXT_CONTENT' 45start=60 end=71 text='Antarctica.' sources=[DocumentSource(type='document', id='101', document={'id': '101', 'snippet': 'Emperor penguins only live in Antarctica.', 'title': 'Penguin habitats'})] type='TEXT_CONTENT'\n```\n\nExample:\n```text\n1documents = [2 {3 \"data\": {4 \"title\": \"Tall penguins\",5 \"snippet\": \"Emperor penguins are the tallest.\",6 },7 \"id\": \"100\",8 },9 {10 \"data\": {11 \"title\": \"Penguin habitats\",12 \"snippet\": \"Emperor penguins only live in Antarctica.\",13 },14 \"id\": \"101\",15 },16]1718messages = [19 {\"role\": \"user\", \"content\": \"Where do the tallest penguins live?\"}20]2122response = co.chat_stream(23 model=\"command-a-plus-05-2026\",24 messages=messages,25 documents=documents,26 citation_options={\"mode\": \"fast\"},27)2829response_text = \"\"30for chunk in response:31 if chunk:32 if chunk.type == \"content-delta\":33 response_text += chunk.delta.message.content.text34 print(chunk.delta.message.content.text, end=\"\")35 if chunk.type == \"citation-start\":36 print(37 f\" [{chunk.delta.message.citations.sources[0].id}]\",38 end=\"\",39 )\n```\n\nExample:\n```text\n1The tallest penguins [100] are the Emperor penguins [100] which only live in Antarctica. [101]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.308Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":13,"totalLines":68,"estimatedTokens":1702}}100{"id":"doc-parameter_types_in_structured_outputs_json_coher-a5ad9979","source":"documentation","title":"Parameter Types in Structured Outputs (JSON) | Cohere","url":"https://docs.cohere.com/docs/parameter-types-in-json","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"YOUR API KEY\")45res = co.chat(6 # The model name. Example: command-a-plus-05-20267 model=\"MODEL_NAME\",8 # The user message. Optional - you can first add a `system_message` role9 messages=[10 {11 \"role\": \"user\",12 \"content\": message,13 }14 ],15 # The schema that you define16 response_format=response_format,17 # Typically, you'll need a low temperature for more deterministic outputs18 temperature=0,19)2021print(res.message.content[0].text)\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"title\": {\"type\": \"string\"},7 \"author\": {\"type\": \"string\"},8 },9 \"required\": [\"title\", \"author\"],10 },11}1213message = \"Generate a JSON describing a book, with the fields 'title' and 'author'\"\n```\n\nExample:\n```text\n1{2 \"title\": \"The Great Gatsby\",3 \"author\": \"F. Scott Fitzgerald\"4}\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"title\": {\"type\": \"string\"},7 \"author\": {\"type\": \"string\"},8 \"publication_year\": {\"type\": \"integer\"},9 },10 \"required\": [\"title\", \"author\", \"publication_year\"],11 },12}1314message = \"Generate a JSON describing a book, with the fields 'title', 'author' and 'publication_year'\"\n```\n\nExample:\n```text\n1{2 \"title\": \"The Great Gatsby\",3 \"author\": \"F. Scott Fitzgerald\",4 \"publication_year\": 19255}\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"city\": {\"type\": \"string\"},7 \"temperature\": {\"type\": \"number\"},8 },9 \"required\": [\"city\", \"temperature\"],10 },11}1213message = \"Generate a JSON of a city and its average daily temperature in celcius\"\n```\n\nExample:\n```text\n1{2 \"city\": \"Toronto\",3 \"temperature\": 15.64}\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"city\": {\"type\": \"string\"},7 \"is_capital\": {\"type\": \"boolean\"},8 },9 \"required\": [\"city\", \"is_capital\"],10 },11}1213message = \"Generate a JSON about a city in Spain and whether it is the capital of its country using 'is_capital'.\"\n```\n\nExample:\n```text\n1{2 \"city\": \"Madrid\",3 \"is_capital\": true4}\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"cities\": {7 \"type\": \"array\",8 \"items\": {\"type\": \"string\"},9 }10 },11 \"required\": [\"cities\"],12 },13}1415message = \"Generate a JSON listing three cities in Japan.\"\n```\n\nExample:\n```text\n1{2 \"cities\": [3 \"Tokyo\",4 \"Kyoto\",5 \"Osaka\"6 ]7}\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"cities\": {7 \"type\": \"array\",8 }9 },10 \"required\": [\"cities\"],11 },12}1314message = \"Generate a JSON listing three cities in Japan.\"\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"coordinates\": {7 \"type\": \"array\",8 \"items\": {9 \"type\": \"array\",10 \"items\": {\"type\": \"number\"},11 },12 }13 },14 \"required\": [\"coordinates\"],15 },16}1718message = \"Generate a JSON of three random coordinates.\"\n```\n\nExample:\n```text\n1{2 \"coordinates\": [3 [-31.28333, 146.41667],4 [78.95833, 11.93333],5 [44.41667, -75.68333]6 ]7}\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"actions\": {7 \"type\": \"array\",8 \"items\": {9 \"type\": \"object\",10 \"properties\": {11 \"japanese\": {\"type\": \"string\"},12 \"romaji\": {\"type\": \"string\"},13 \"english\": {\"type\": \"string\"},14 },15 \"required\": [\"japanese\", \"romaji\", \"english\"],16 },17 }18 },19 \"required\": [\"actions\"],20 },21}2223message = \"Generate a JSON array of 3 objects with the following fields: japanese, romaji, english. These actions should be japanese verbs provided in the dictionary form.\"\n```\n\nExample:\n```text\n1{2 \"actions\": [3 {4 \"japanese\": \"食べる\",5 \"romaji\": \"taberu\",6 \"english\": \"to eat\"7 },8 {9 \"japanese\": \"話す\",10 \"romaji\": \"hanasu\",11 \"english\": \"to speak\"12 },13 {14 \"japanese\": \"書く\",15 \"romaji\": \"kaku\",16 \"english\": \"to write\"17 }18 ]19}\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"genre\": {7 \"type\": \"string\",8 \"enum\": [\"historical fiction\", \"cozy mystery\"],9 },10 \"title\": {\"type\": \"string\"},11 },12 \"required\": [\"title\", \"genre\"],13 },14}1516message = \"Generate a JSON for a new book idea.\"\n```\n\nExample:\n```text\n1{2 \"genre\": \"historical fiction\",3 \"title\": \"The Unseen Thread: A Tale of the Silk Road's Secrets and Shadows\"4 }\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"city\": {7 \"type\": \"object\",8 \"properties\": {9 \"country\": {10 \"type\": \"string\",11 \"const\": \"Thailand\",12 },13 \"city_name\": {\"type\": \"string\"},14 \"avg_temperature\": {\"type\": \"number\"},15 },16 \"required\": [17 \"country\",18 \"city_name\",19 \"avg_temperature\",20 ],21 }22 },23 \"required\": [\"city\"],24 },25}2627message = \"Generate a JSON of a city.\"\n```\n\nExample:\n```text\n1{2 \"city\": {3 \"country\": \"Thailand\",4 \"city_name\": \"Bangkok\",5 \"avg_temperature\": 29.0833333333333326 }7}\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"product_sku\": {7 \"type\": \"string\",8 \"pattern\": \"[A-Z]{3}[0-9]{7}\",9 }10 },11 \"required\": [\"product_sku\"],12 },13}1415message = \"Generate a JSON of an SKU for a new product line.\"\n```\n\nExample:\n```text\n1{2 \"product_sku\": \"PRX0012345\"3}\n```\n\nExample:\n```text\n1response_format = {2 \"type\": \"json_object\",3 \"schema\": {4 \"type\": \"object\",5 \"properties\": {6 \"itinerary\": {7 \"type\": \"array\",8 \"items\": {9 \"type\": \"object\",10 \"properties\": {11 \"day_number\": {\"type\": \"integer\"},12 \"date\": {\"type\": \"string\", \"format\": \"date\"},13 \"places_to_visit\": {\"type\": \"string\"},14 },15 \"required\": [16 \"day_number\",17 \"date\",18 \"places_to_visit\",19 ],20 },21 }22 },23 \"required\": [\"itinerary\"],24 },25}2627message = (28 \"Generate a JSON of a 3-day visit of Bali starting Jan 5 2025.\"29)\n```\n\nExample:\n```text\n1{2 \"itinerary\": [3 {4 \"day_number\": 1,5 \"date\": \"2025-01-05\",6 \"places_to_visit\": \"Tanah Lot Temple, Ubud Monkey Forest, Tegalalang Rice Terraces\"7 },8 {9 \"day_number\": 2,10 \"date\": \"2025-01-06\",11 \"places_to_visit\": \"Mount Batur, Tirta Empul Temple, Ubud Art Market\"12 },13 {14 \"day_number\": 3,15 \"date\": \"2025-01-07\",16 \"places_to_visit\": \"Uluwatu Temple, Kuta Beach, Seminyak\"17 }18 ]19}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.309Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":24,"totalLines":123,"estimatedTokens":2157}}101{"id":"doc-chroma_and_cohere_integration_guide_cohere-dc54bac5","source":"documentation","title":"Chroma and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/chroma-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.309Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}102{"id":"doc-qdrant_and_cohere_integration_guide_cohere-05206d08","source":"documentation","title":"Qdrant and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/qdrant-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.309Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}103{"id":"doc-mongodb_and_cohere_integration_guide_cohere-d851181d","source":"documentation","title":"MongoDB and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/mongodb-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.310Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}104{"id":"doc-pinecone_and_cohere_integration_guide_cohere-082888a7","source":"documentation","title":"Pinecone and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/pinecone-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.310Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}105{"id":"doc-open_search_and_cohere_integration_guide_cohere-a6b3e512","source":"documentation","title":"Open Search and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/opensearch-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.310Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}106{"id":"doc-vespa_and_cohere_integration_guide_cohere-9481479d","source":"documentation","title":"Vespa and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/vespa-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.310Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}107{"id":"doc-zilliz_and_cohere_integration_guide_cohere-30617668","source":"documentation","title":"Zilliz and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/zilliz-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.310Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}108{"id":"doc-best_practices_for_using_rerank_cohere-3b65b01b","source":"documentation","title":"Best Practices for using Rerank | Cohere","url":"https://docs.cohere.com/docs/reranking-best-practices","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import yaml23docs = [4 {5 \"Title\": \"How to fix a dishwasher\",6 \"Author\": \"John Smith\",7 \"Date\": \"August 1st 2023\",8 \"Content\": \"Fixing a dishwasher depends on the specific problem you're facing. Here are some common issues and their potential solutions:....\",9 },10 {11 \"Title\": \"How to fix a leaky sink\",12 \"Date\": \"July 25th 2024\",13 \"Content\": \"Fixing a leaky sink will depend on the source of the leak. Here are general steps you can take to address common types of sink leaks:.....\",14 },15]1617yaml_docs = [yaml.dump(doc, sort_keys=False) for doc in docs]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.310Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":1,"totalLines":8,"estimatedTokens":207}}109{"id":"doc-parameter_types_for_tool_use_function_calling_co-872cf2a2","source":"documentation","title":"Parameter types for tool use (function calling) | Cohere","url":"https://docs.cohere.com/docs/tool-use-parameter-types","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1response = co.chat(model=\"command-a-plus-05-2026\",2 messages=[{\"role\": \"user\", \"content\": \"What's the weather in Toronto?\"}],3 tools=tools,4 strict_tools=True5)\n```\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere34co = cohere.ClientV2(5 \"COHERE_API_KEY\"6) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1response = co.chat(2 # The model name. Example: command-a-plus-05-20263 model=\"MODEL_NAME\",4 # The user message. Optional - you can first add a `system_message` role5 messages=[6 {7 \"role\": \"user\",8 \"content\": message,9 }10 ],11 # The tool schema that you define12 tools=tools,13 # This guarantees that the output will adhere to the schema14 strict_tools=True,15 # Typically, you'll need a low temperature for more deterministic outputs16 temperature=0,17)1819for tc in response.message.tool_calls:20 print(f\"{tc.function.name} | Parameters: {tc.function.arguments}\")\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"get_weather\",6 \"description\": \"Gets the weather of a given location\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"location\": {11 \"type\": \"string\",12 \"description\": \"the location to get the weather, example: San Francisco.\",13 }14 },15 \"required\": [\"location\"],16 },17 },18 },19]2021message = \"What's the weather in Toronto?\"\n```\n\nExample:\n```text\n1get_weather2{3 \"location\": \"Toronto\"4}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"add_numbers\",6 \"description\": \"Adds two numbers\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"first_number\": {11 \"type\": \"integer\",12 \"description\": \"The first number to add.\",13 },14 \"second_number\": {15 \"type\": \"integer\",16 \"description\": \"The second number to add.\",17 },18 },19 \"required\": [\"first_number\", \"second_number\"],20 },21 },22 }23]2425message = \"What is five plus two\"\n```\n\nExample:\n```text\n1add_numbers2{3 \"first_number\": 5,4 \"second_number\": 25}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"add_numbers\",6 \"description\": \"Adds two numbers\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"first_number\": {11 \"type\": \"number\",12 \"description\": \"The first number to add.\",13 },14 \"second_number\": {15 \"type\": \"number\",16 \"description\": \"The second number to add.\",17 },18 },19 \"required\": [\"first_number\", \"second_number\"],20 },21 },22 }23]2425message = \"What is 5.3 plus 2\"\n```\n\nExample:\n```text\n1add_numbers2{3 \"first_number\": 5.3,4 \"second_number\": 25}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"reserve_tickets\",6 \"description\": \"Reserves a train ticket\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"quantity\": {11 \"type\": \"integer\",12 \"description\": \"The quantity of tickets to reserve.\",13 },14 \"trip_protection\": {15 \"type\": \"boolean\",16 \"description\": \"Indicates whether to add trip protection.\",17 },18 },19 \"required\": [\"quantity\", \"trip_protection\"],20 },21 },22 }23]2425message = \"Book me 2 tickets. I don't need trip protection.\"\n```\n\nExample:\n```text\n1reserve_tickets2{3 \"quantity\": 2,4 \"trip_protection\": false5}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"get_weather\",6 \"description\": \"Gets the weather of a given location\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"locations\": {11 \"type\": \"array\",12 \"items\": {\"type\": \"string\"},13 \"description\": \"The locations to get weather.\",14 }15 },16 \"required\": [\"locations\"],17 },18 },19 },20]2122message = \"What's the weather in Toronto and New York?\"\n```\n\nExample:\n```text\n1get_weather2{3 \"locations\": [4 \"Toronto\",5 \"New York\"6 ]7}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"get_weather\",6 \"description\": \"Gets the weather of a given location\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"locations\": {11 \"type\": \"array\",12 \"description\": \"The locations to get weather.\",13 }14 },15 \"required\": [\"locations\"],16 },17 },18 },19]2021message = \"What's the weather in Toronto and New York?\"\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"maxPoints\",6 \"description\": \"Finds the maximum number of points on a line.\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"points\": {11 \"type\": \"array\",12 \"description\": \"The list of points. Points are 2 element lists [x, y].\",13 \"items\": {14 \"type\": \"array\",15 \"items\": {\"type\": \"integer\"},16 \"description\": \"A point represented by a 2 element list [x, y].\",17 },18 }19 },20 \"required\": [\"points\"],21 },22 },23 }24]2526message = \"Please provide the maximum number of collinear points for this set of coordinates - [[1,1],[2,2],[3,4],[5,5]].\"\n```\n\nExample:\n```text\n1maxPoints2{3 \"points\": [4 [1,1],5 [2,2],6 [3,4],7 [5,5]8 ]9}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"search_furniture_products\",6 \"description\": \"Searches for furniture products given the user criteria.\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"product_type\": {11 \"type\": \"string\",12 \"description\": \"The type of the product to search for.\",13 },14 \"features\": {15 \"type\": \"object\",16 \"properties\": {17 \"material\": {\"type\": \"string\"},18 \"style\": {\"type\": \"string\"},19 },20 \"required\": [\"style\"],21 },22 },23 \"required\": [\"product_type\"],24 },25 },26 }27]2829message = \"I'm looking for a dining table made of oak in Scandinavian style.\"\n```\n\nExample:\n```text\n1search_furniture_products2{3 \"features\": {4 \"material\": \"oak\",5 \"style\": \"Scandinavian\"6 },7 \"product_type\": \"dining table\"8}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"fetch_contacts\",6 \"description\": \"Fetch a contact by type\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"contact_type\": {11 \"type\": \"string\",12 \"description\": \"The type of contact to fetch.\",13 \"enum\": [\"customer\", \"supplier\"],14 }15 },16 \"required\": [\"contact_type\"],17 },18 },19 }20]2122message = \"Give me vendor contacts.\"\n```\n\nExample:\n```text\n1fetch_contacts2{3 \"contact_type\": \"supplier\"4}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"get_weather\",6 \"description\": \"Gets the weather of a given location\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"location\": {11 \"type\": \"string\",12 \"description\": \"The location to get weather.\",13 },14 \"country\": {15 \"type\": \"string\",16 \"description\": \"The country for the weather lookup\",17 \"const\": \"Canada\",18 },19 },20 \"required\": [\"location\", \"country\"],21 },22 },23 },24]2526message = \"What's the weather in Toronto and Vancouver?\"\n```\n\nExample:\n```text\n1get_weather2{3 \"country\": \"Canada\",4 \"location\": \"Toronto\"5}6---7get_weather8{9 \"country\": \"Canada\",10 \"location\": \"Vancouver\"11}12---\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"query_product_by_sku\",6 \"description\": \"Queries products by SKU pattern\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"sku_pattern\": {11 \"type\": \"string\",12 \"description\": \"Pattern to match SKUs\",13 \"pattern\": \"[A-Z]{3}[0-9]{4}\",14 }15 },16 \"required\": [\"sku_pattern\"],17 },18 },19 }20]2122message = \"Check the stock level of this product - 7374 hgY\"\n```\n\nExample:\n```text\n1query_product_by_sku2{3 \"sku_pattern\": \"HGY7374\"4}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"book_hotel\",6 \"description\": \"Books a hotel room for a specific check-in date\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"hotel_name\": {11 \"type\": \"string\",12 \"description\": \"Name of the hotel\",13 },14 \"check_in_date\": {15 \"type\": \"string\",16 \"description\": \"Check-in date for the hotel\",17 \"format\": \"date\",18 },19 },20 \"required\": [\"hotel_name\", \"check_in_date\"],21 },22 },23 }24]2526message = \"Book a room at the Grand Hotel with check-in on Dec 2 2024\"\n```\n\nExample:\n```text\n1book_hotel2{3 \"check_in_date\": \"2024-12-02\",4 \"hotel_name\": \"Grand Hotel\"5}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.311Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":26,"totalLines":133,"estimatedTokens":2919}}110{"id":"doc-haystack_and_cohere_integration_guide_cohere-2d23d125","source":"documentation","title":"Haystack and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/haystack-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from haystack import Pipeline2from haystack.components.builders import DynamicChatPromptBuilder3from haystack.dataclasses import ChatMessage4from haystack_integrations.components.generators.cohere import (5 CohereChatGenerator,6)7from haystack.utils import Secret8import os910COHERE_API_KEY = os.environ.get(\"COHERE_API_KEY\")1112pipe = Pipeline()13pipe.add_component(\"prompt_builder\", DynamicChatPromptBuilder())14pipe.add_component(15 \"llm\", CohereChatGenerator(Secret.from_token(COHERE_API_KEY))16)17pipe.connect(\"prompt_builder\", \"llm\")1819location = \"Berlin\"20system_message = ChatMessage.from_system(21 \"You are an assistant giving out valuable information to language learners.\"22)23messages = [24 system_message,25 ChatMessage.from_user(\"Tell me about {{location}}\"),26]2728res = pipe.run(29 data={30 \"prompt_builder\": {31 \"template_variables\": {\"location\": location},32 \"prompt_source\": messages,33 }34 }35)36print(res)\n```\n\nExample:\n```text\n1messages = [2 system_message,3 ChatMessage.from_user(4 \"What's the weather forecast for {{location}} in the next {{day_count}} days?\"5 ),6]78res = pipe.run(9 data={10 \"prompt_builder\": {11 \"template_variables\": {12 \"location\": location,13 \"day_count\": \"5\",14 },15 \"prompt_source\": messages,16 }17 }18)1920print(res)\n```\n\nExample:\n```text\n1from haystack import Document2from haystack import Pipeline3from haystack.components.builders import DynamicChatPromptBuilder4from haystack.components.generators.utils import print_streaming_chunk5from haystack.components.fetchers import LinkContentFetcher6from haystack.components.converters import HTMLToDocument7from haystack.dataclasses import ChatMessage8from haystack.utils import Secret910from haystack_integrations.components.generators.cohere import (11 CohereChatGenerator,12)1314fetcher = LinkContentFetcher()15converter = HTMLToDocument()16prompt_builder = DynamicChatPromptBuilder(17 runtime_variables=[\"documents\"]18)19llm = CohereChatGenerator(Secret.from_token(COHERE_API_KEY))2021message_template = \"\"\"Answer the following question based on the contents of the article: {{query}}\\n22 Article: {{documents[0].content}} \\n23 \"\"\"24messages = [ChatMessage.from_user(message_template)]2526rag_pipeline = Pipeline()27rag_pipeline.add_component(name=\"fetcher\", instance=fetcher)28rag_pipeline.add_component(name=\"converter\", instance=converter)29rag_pipeline.add_component(\"prompt_builder\", prompt_builder)30rag_pipeline.add_component(\"llm\", llm)3132rag_pipeline.connect(\"fetcher.streams\", \"converter.sources\")33rag_pipeline.connect(34 \"converter.documents\", \"prompt_builder.documents\"35)36rag_pipeline.connect(\"prompt_builder.prompt\", \"llm.messages\")3738question = \"What are the capabilities of Cohere?\"3940result = rag_pipeline.run(41 {42 \"fetcher\": {\"urls\": [\"/reference/about\"]},43 \"prompt_builder\": {44 \"template_variables\": {\"query\": question},45 \"prompt_source\": messages,46 },47 \"llm\": {\"generation_kwargs\": {\"max_tokens\": 165}},48 },49)50print(result)51# {'llm': {'replies': [ChatMessage(content='The Cohere platform builds natural language processing and generation into your product with a few lines of code... \\nIs', role=<ChatRole.ASSISTANT: 'assistant'>, name=None, meta={'model': 'command', 'usage': {'prompt_tokens': 273, 'response_tokens': 165, 'total_tokens': 438, 'billed_tokens': 430}, 'index': 0, 'finish_reason': None, 'documents': None, 'citations': None})]}}\n```\n\nExample:\n```text\n1from haystack import Pipeline2from haystack.components.retrievers.in_memory import (3 InMemoryBM25Retriever,4)5from haystack.components.builders.prompt_builder import PromptBuilder6from haystack.document_stores.in_memory import InMemoryDocumentStore7from haystack_integrations.components.generators.cohere import (8 CohereGenerator,9)10from haystack import Document11from haystack.utils import Secret1213import os1415COHERE_API_KEY = os.environ.get(\"COHERE_API_KEY\")1617docstore = InMemoryDocumentStore()18docstore.write_documents(19 [20 Document(content=\"Rome is the capital of Italy\"),21 Document(content=\"Paris is the capital of France\"),22 ]23)2425query = \"What is the capital of France?\"2627template = \"\"\"28Given the following information, answer the question.2930Context:31{% for document in documents %}32 {{ document.content }}33{% endfor %}3435Question: {{ query }}?36\"\"\"37pipe = Pipeline()3839pipe.add_component(40 \"retriever\", InMemoryBM25Retriever(document_store=docstore)41)42pipe.add_component(\"prompt_builder\", PromptBuilder(template=template))43pipe.add_component(44 \"llm\", CohereGenerator(Secret.from_token(COHERE_API_KEY))45)46pipe.connect(\"retriever\", \"prompt_builder.documents\")47pipe.connect(\"prompt_builder\", \"llm\")4849res = pipe.run(50 {51 \"prompt_builder\": {\"query\": query},52 \"retriever\": {\"query\": query},53 }54)5556print(res)57# {'llm': {'replies': [' Paris is the capital of France. It is known for its history, culture, and many iconic landmarks, such as the Eiffel Tower and Notre-Dame Cathedral. '], 'meta': [{'finish_reason': 'COMPLETE'}]}}\n```\n\nExample:\n```text\n1from haystack import Pipeline2from haystack import Document3from haystack.document_stores.in_memory import InMemoryDocumentStore4from haystack.components.writers import DocumentWriter5from haystack_integrations.components.embedders.cohere import (6 CohereDocumentEmbedder,7)8from haystack.utils import Secret9import os1011COHERE_API_KEY = os.environ.get(\"COHERE_API_KEY\")12token = Secret.from_token(COHERE_API_KEY)1314document_store = InMemoryDocumentStore(15 embedding_similarity_function=\"cosine\"16)1718documents = [19 Document(content=\"My name is Wolfgang and I live in Berlin\"),20 Document(content=\"I saw a black horse running\"),21 Document(content=\"Germany has many big cities\"),22]2324indexing_pipeline = Pipeline()25indexing_pipeline.add_component(26 \"embedder\", CohereDocumentEmbedder(token)27)28indexing_pipeline.add_component(29 \"writer\", DocumentWriter(document_store=document_store)30)31indexing_pipeline.connect(\"embedder\", \"writer\")3233indexing_pipeline.run({\"embedder\": {\"documents\": documents}})34print(document_store.filter_documents())35# [Document(id=..., content: 'My name is Wolfgang and I live in Berlin', embedding: vector of size 4096), Document(id=..., content: 'Germany has many big cities', embedding: vector of size 4096)]\n```\n\nExample:\n```text\n1from haystack import Pipeline2from haystack.components.retrievers.in_memory import (3 InMemoryEmbeddingRetriever,4)5from haystack_integrations.components.embedders.cohere import (6 CohereTextEmbedder,7)89query_pipeline = Pipeline()10query_pipeline.add_component(11 \"text_embedder\", CohereTextEmbedder(token)12)13query_pipeline.add_component(14 \"retriever\",15 InMemoryEmbeddingRetriever(document_store=document_store),16)17query_pipeline.connect(18 \"text_embedder.embedding\", \"retriever.query_embedding\"19)2021query = \"Who lives in Berlin?\"2223result = query_pipeline.run({\"text_embedder\": {\"text\": query}})2425print(result[\"retriever\"][\"documents\"][0])2627# Document(id=..., text: 'My name is Wolfgang and I live in Berlin')\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.312Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":6,"totalLines":33,"estimatedTokens":1887}}111{"id":"doc-an_overview_of_cohere_s_rerank_model_cohere-f436c82e","source":"documentation","title":"An Overview of Cohere's Rerank Model | Cohere","url":"https://docs.cohere.com/docs/rerank-overview","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45query = \"What is the capital of the United States?\"6docs = [7 \"Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.\",8 \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.\",9 \"Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.\",10 \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.\",11 \"Capital punishment has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.\",12]1314results = co.rerank(15 model=\"rerank-v4.0-pro\", query=query, documents=docs, top_n=516)\n```\n\nExample:\n```text\n1V2RerankResponse(2 id=\"2104ccd0-74b5-4951-9bb1-cc543b26720f\",3 results=[4 V2RerankResponseResultsItem(5 index=3, relevance_score=0.9432646 ),7 V2RerankResponseResultsItem(8 index=2, relevance_score=0.622092079 ),10 V2RerankResponseResultsItem(11 index=1, relevance_score=0.605425812 ),13 V2RerankResponseResultsItem(14 index=0, relevance_score=0.5904013515 ),16 V2RerankResponseResultsItem(17 index=4, relevance_score=0.466456718 ),19 ],20 meta=ApiMeta(21 api_version=ApiMetaApiVersion(22 version=\"2\", is_deprecated=None, is_experimental=None23 ),24 billed_units=ApiMetaBilledUnits(25 images=None,26 input_tokens=None,27 output_tokens=None,28 search_units=1.0,29 classifications=None,30 ),31 tokens=None,32 cached_tokens=None,33 warnings=None,34 ),35)\n```\n\nExample:\n```text\n1import yaml2import cohere34co = cohere.ClientV2()56query = \"What is the capital of the United States?\"7docs = [8 {9 \"Title\": \"Facts about Carson City\",10 \"Content\": \"Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.\",11 },12 {13 \"Title\": \"The Commonwealth of Northern Mariana Islands\",14 \"Content\": \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.\",15 },16 {17 \"Title\": \"The Capital of United States Virgin Islands\",18 \"Content\": \"Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.\",19 },20 {21 \"Title\": \"Washington D.C.\",22 \"Content\": \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.\",23 },24 {25 \"Title\": \"Capital Punishment in the US\",26 \"Content\": \"Capital punishment has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.\",27 },28]2930yaml_docs = [yaml.dump(doc, sort_keys=False) for doc in docs]3132results = co.rerank(33 model=\"rerank-v4.0-pro\",34 query=query,35 documents=yaml_docs,36 top_n=5,37)\n```\n\nExample:\n```text\n1V2RerankResponse(2 id=\"df4d8720-8265-4868-a8f5-0bcee7a35bd0\",3 results=[4 V2RerankResponseResultsItem(5 index=3, relevance_score=0.94978136 ),7 V2RerankResponseResultsItem(8 index=2, relevance_score=0.690642549 ),10 V2RerankResponseResultsItem(11 index=0, relevance_score=0.5790195512 ),13 V2RerankResponseResultsItem(14 index=1, relevance_score=0.548286515 ),16 V2RerankResponseResultsItem(17 index=4, relevance_score=0.4937502718 ),19 ],20 meta=ApiMeta(21 api_version=ApiMetaApiVersion(22 version=\"2\", is_deprecated=None, is_experimental=None23 ),24 billed_units=ApiMetaBilledUnits(25 images=None,26 input_tokens=None,27 output_tokens=None,28 search_units=1.0,29 classifications=None,30 ),31 tokens=None,32 cached_tokens=None,33 warnings=None,34 ),35)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.312Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":4,"totalLines":23,"estimatedTokens":1344}}112{"id":"doc-redis_and_cohere_integration_guide_cohere-79f0da69","source":"documentation","title":"Redis and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/redis-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$docker run -d --name redis-stack -p 6379:6379 -p 8001:8001 redis/redis-stack:latest\n```\n\nExample:\n```text\n$!pip install redisvl==0.1.0$!pip install cohere==4.45$!pip install jsonlines\n```\n\nExample:\n```text\n1from redis import Redis2from redisvl.index import SearchIndex3from redisvl.schema import IndexSchema4from redisvl.utils.vectorize import CohereTextVectorizer5from redisvl.query import VectorQuery6from redisvl.query.filter import Tag, Text, Num7import jsonlines\n```\n\nExample:\n```text\n1version: \"0.1.0\"2index:3 name: semantic_search_demo4 prefix: rvl5 storage_type: hash67fields:8 - name: url9 type: text10 - name: title11 type: tag12 - name: text13 type: text14 - name: wiki_id15 type: numeric16 - name: paragraph_id17 type: numeric18 - name: id19 type: numeric20 - name: views21 type: numeric22 - name: langs23 type: numeric24 - name: embedding25 type: vector26 attrs:27 algorithm: flat28 dims: 102429 distance_metric: cosine30 datatype: float32\n```\n\nExample:\n```text\n1# create a vectorizer2api_key = \"{Insert your cohere API Key}\"34cohere_vectorizer = CohereTextVectorizer(5 model=\"embed-english-v3.0\",6 api_config={\"api_key\": api_key},7)\n```\n\nExample:\n```text\n1# construct a search index from the schema - this schema is called \"semantic_search_demo\"2schema = IndexSchema.from_yaml(\"./schema.yaml\")3client = Redis.from_url(\"redis://localhost:6379\")4index = SearchIndex(schema, client)56# create the index (no data yet)7index.create(overwrite=True)\n```\n\nExample:\n```text\n$!rvl index listall\n```\n\nExample:\n```text\n15:39:22 [RedisVL] INFO Indices:15:39:22 [RedisVL] INFO 1. semantic_search_demo\n```\n\nExample:\n```text\n$!rvl index info -i semantic_search_demo\n```\n\nExample:\n```text\nLook inside the index to make sure it matches the schema you want:╭──────────────────────┬────────────────┬────────────┬─────────────────┬────────────╮│ Index Name │ Storage Type │ Prefixes │ Index Options │ Indexing │├──────────────────────┼────────────────┼────────────┼─────────────────┼────────────┤│ semantic_search_demo │ HASH │ ['rvl'] │ [] │ 0 │╰──────────────────────┴────────────────┴────────────┴─────────────────┴────────────╯Index Fields:╭──────────────┬──────────────┬─────────┬────────────────┬────────────────┬────────────────┬────────────────┬────────────────┬────────────────┬─────────────────┬────────────────╮│ Name │ Attribute │ Type │ Field Option │ Option Value │ Field Option │ Option Value │ Field Option │ Option Value │ Field Option │ Option Value │├──────────────┼──────────────┼─────────┼────────────────┼────────────────┼────────────────┼────────────────┼────────────────┼────────────────┼─────────────────┼────────────────┤│ url │ url │ TEXT │ WEIGHT │ 1 │ │ │ │ │ │ ││ title │ title │ TEXT │ WEIGHT │ 1 │ │ │ │ │ │ ││ text │ text │ TEXT │ WEIGHT │ 1 │ │ │ │ │ │ ││ wiki_id │ wiki_id │ NUMERIC │ │ │ │ │ │ │ │ ││ paragraph_id │ paragraph_id │ NUMERIC │ │ │ │ │ │ │ │ ││ id │ id │ NUMERIC │ │ │ │ │ │ │ │ ││ views │ views │ NUMERIC │ │ │ │ │ │ │ │ ││ langs │ langs │ NUMERIC │ │ │ │ │ │ │ │ ││ embedding │ embedding │ VECTOR │ algorithm │ FLAT │ data_type │ FLOAT32 │ dim │ 1024 │ distance_metric │ COSINE │╰──────────────┴──────────────┴─────────┴────────────────┴────────────────┴────────────────┴────────────────┴────────────────┴────────────────┴─────────────────┴────────────────╯\n```\n\nExample:\n```text\n1# read in your documents2jsonl_file_path = \"data/redis_guide_data.jsonl\"34corpus = []5text_to_embed = []67with jsonlines.open(jsonl_file_path, mode=\"r\") as reader:8 for line in reader:9 corpus.append(line)10 # we want to store the embeddings of the field called `text`11 text_to_embed.append(line[\"text\"])1213# call embed_many which returns an array14# hash data structures get serialized as a string and thus we store the embeddings in hashes as a byte string (handled by numpy)15res = cohere_vectorizer.embed_many(16 text_to_embed, input_type=\"search_document\", as_buffer=True17)\n```\n\nExample:\n```text\n1# contruct the data payload to be uploaded to your index2data = [3 {4 \"url\": row[\"url\"],5 \"title\": row[\"title\"],6 \"text\": row[\"text\"],7 \"wiki_id\": row[\"wiki_id\"],8 \"paragraph_id\": row[\"paragraph_id\"],9 \"id\": row[\"id\"],10 \"views\": row[\"views\"],11 \"langs\": row[\"langs\"],12 \"embedding\": v,13 }14 for row, v in zip(corpus, res)15]1617# load the data into your index18index.load(data)\n```\n\nExample:\n```text\n1# use the Cohere vectorizer again to create a query embedding2query_embedding = cohere_vectorizer.embed(3 \"What did Microsoft release in 2015?\",4 input_type=\"search_query\",5 as_buffer=True,6)789query = VectorQuery(10 vector=query_embedding,11 vector_field_name=\"embedding\",12 return_fields=[13 \"url\",14 \"wiki_id\",15 \"paragraph_id\",16 \"id\",17 \"views\",18 \"langs\",19 \"title\",20 \"text\",21 ],22 num_results=5,23)2425results = index.query(query)2627for doc in results:28 print(29 f\"Title:{doc['title']}\\nText:{doc['text']}\\nDistance {doc['vector_distance']}\\n\\n\"30 )\n```\n\nExample:\n```text\n1# Initialize a tag filter2tag_filter = Tag(\"title\") == \"Microsoft Office\"34# set the tag filter on our existing query5query.set_filter(tag_filter)67results = index.query(query)89for doc in results:10 print(11 f\"Title:{doc['title']}\\nText:{doc['text']}\\nDistance {doc['vector_distance']}\\n\"12 )\n```\n\nExample:\n```text\n1# define a tag match on the title, text match on the text field, and numeric filter on the views field2filter_data = (3 (Tag(\"title\") == \"Elizabeth II\")4 & (Text(\"text\") % \"born\")5 & (Num(\"views\") > 4500)6)78query_embedding = co.embed(9 \"When was she born?\", input_type=\"search_query\", as_buffer=True10)1112# reinitialize the query with the filter expression13query = VectorQuery(14 vector=query_embedding,15 vector_field_name=\"embedding\",16 return_fields=[17 \"url\",18 \"wiki_id\",19 \"paragraph_id\",20 \"id\",21 \"views\",22 \"langs\",23 \"title\",24 \"text\",25 ],26 num_results=5,27 filter_expression=filter_data,28)2930results = index.query(query)31print(results)3233for doc in results:34 print(35 f\"Title:{doc['title']}\\nText:{doc['text']}\\nDistance {doc['vector_distance']}\\nView {doc['views']}\"36 )\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.313Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":1953}}113{"id":"doc-streaming_for_tool_use_function_calling_cohere-58e5515a","source":"documentation","title":"Streaming for tool use (function calling) | Cohere","url":"https://docs.cohere.com/docs/tool-use-streaming","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# User message2\"What's the weather in Madrid and Brasilia?\"34# Events stream5type='message-start' id='fba98ad3-e5a1-413c-a8de-84fbf9baabf7' delta=ChatMessageStartEventDelta(message=ChatMessageStartEventDeltaMessage(role='assistant', content=[], tool_plan='', tool_calls=[], citations=[])) 6 --------------------------------------------------7type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan='I')) 8 --------------------------------------------------9type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan=' will')) 10 --------------------------------------------------11type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan=' search')) 12 --------------------------------------------------13type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan=' for')) 14 --------------------------------------------------15type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan=' the')) 16 --------------------------------------------------17type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan=' weather')) 18 --------------------------------------------------19type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan=' in')) 20 --------------------------------------------------21type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan=' Madrid')) 22 --------------------------------------------------23type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan=' and')) 24 --------------------------------------------------25type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan=' Brasilia')) 26 --------------------------------------------------27type='tool-plan-delta' delta=ChatToolPlanDeltaEventDelta(message=ChatToolPlanDeltaEventDeltaMessage(tool_plan='.')) 28 --------------------------------------------------29type='tool-call-start' index=0 delta=ChatToolCallStartEventDelta(message=ChatToolCallStartEventDeltaMessage(tool_calls=ToolCallV2(id='get_weather_p1t92w7gfgq7', type='function', function=ToolCallV2Function(name='get_weather', arguments='')))) 30 --------------------------------------------------31type='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='{\\n \"')))) 32 --------------------------------------------------33type='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='location')))) 34 --------------------------------------------------35type='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='\":')))) 36 --------------------------------------------------37type='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments=' \"')))) 38 --------------------------------------------------39type='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='Madrid')))) 40 --------------------------------------------------41type='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='\"')))) 42 --------------------------------------------------43type='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='\\n')))) 44 --------------------------------------------------45type='tool-call-delta' index=0 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='}')))) 46 --------------------------------------------------47type='tool-call-end' index=0 48 --------------------------------------------------49type='tool-call-start' index=1 delta=ChatToolCallStartEventDelta(message=ChatToolCallStartEventDeltaMessage(tool_calls=ToolCallV2(id='get_weather_ay6nmvjgp9vn', type='function', function=ToolCallV2Function(name='get_weather', arguments='')))) 50 --------------------------------------------------51type='tool-call-delta' index=1 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='{\\n \"')))) 52 --------------------------------------------------53type='tool-call-delta' index=1 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='location')))) 54 --------------------------------------------------55type='tool-call-delta' index=1 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='\":')))) 56 --------------------------------------------------57type='tool-call-delta' index=1 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments=' \"')))) 58 --------------------------------------------------59type='tool-call-delta' index=1 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='Bras')))) 60 --------------------------------------------------61type='tool-call-delta' index=1 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='ilia')))) 62 --------------------------------------------------63type='tool-call-delta' index=1 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='\"')))) 64 --------------------------------------------------65type='tool-call-delta' index=1 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='\\n')))) 66 --------------------------------------------------67type='tool-call-delta' index=1 delta=ChatToolCallDeltaEventDelta(message=ChatToolCallDeltaEventDeltaMessage(tool_calls=ChatToolCallDeltaEventDeltaMessageToolCalls(function=ChatToolCallDeltaEventDeltaMessageToolCallsFunction(arguments='}')))) 68 --------------------------------------------------69type='tool-call-end' index=1 70 --------------------------------------------------71type='message-end' id=None delta=ChatMessageEndEventDelta(finish_reason='TOOL_CALL', usage=Usage(billed_units=UsageBilledUnits(input_tokens=37.0, output_tokens=28.0, search_units=None, classifications=None), tokens=UsageTokens(input_tokens=913.0, output_tokens=83.0))) 72 --------------------------------------------------\n```\n\nExample:\n```text\n1\"What's the weather in Madrid and Brasilia?\"23type='message-start' id='e8f9afc1-0888-46f0-a9ed-eb0e5a51e17f' delta=ChatMessageStartEventDelta(message=ChatMessageStartEventDeltaMessage(role='assistant', content=[], tool_plan='', tool_calls=[], citations=[])) 4 --------------------------------------------------5type='content-start' index=0 delta=ChatContentStartEventDelta(message=ChatContentStartEventDeltaMessage(content=ChatContentStartEventDeltaMessageContent(text='', type='text'))) 6 --------------------------------------------------7type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='It'))) logprobs=None 8 --------------------------------------------------9type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' is'))) logprobs=None 10 --------------------------------------------------11type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' currently'))) logprobs=None 12 --------------------------------------------------13type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' 2'))) logprobs=None 14 --------------------------------------------------15type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='4'))) logprobs=None 16 --------------------------------------------------17type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='°'))) logprobs=None 18 --------------------------------------------------19type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='C in'))) logprobs=None 20 --------------------------------------------------21type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' Madrid'))) logprobs=None 22 --------------------------------------------------23type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' and'))) logprobs=None 24 --------------------------------------------------25type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' 2'))) logprobs=None 26 --------------------------------------------------27type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='8'))) logprobs=None 28 --------------------------------------------------29type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='°'))) logprobs=None 30 --------------------------------------------------31type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='C in'))) logprobs=None 32 --------------------------------------------------33type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text=' Brasilia'))) logprobs=None 34 --------------------------------------------------35type='content-delta' index=0 delta=ChatContentDeltaEventDelta(message=ChatContentDeltaEventDeltaMessage(content=ChatContentDeltaEventDeltaMessageContent(text='.'))) logprobs=None 36 --------------------------------------------------37type='citation-start' index=0 delta=CitationStartEventDelta(message=CitationStartEventDeltaMessage(citations=Citation(start=16, end=20, text='24°C', sources=[ToolSource(type='tool', id='get_weather_m3kdvxncg1p8:0', tool_output={'temperature': '{\"madrid\":\"24°C\"}'})], type='TEXT_CONTENT'))) 38 --------------------------------------------------39type='citation-end' index=0 40 --------------------------------------------------41type='citation-start' index=1 delta=CitationStartEventDelta(message=CitationStartEventDeltaMessage(citations=Citation(start=35, end=39, text='28°C', sources=[ToolSource(type='tool', id='get_weather_cfwfh3wzkbrs:0', tool_output={'temperature': '{\"brasilia\":\"28°C\"}'})], type='TEXT_CONTENT'))) 42 --------------------------------------------------43type='citation-end' index=1 44 --------------------------------------------------45type='content-end' index=0 46 --------------------------------------------------47type='message-end' id=None delta=ChatMessageEndEventDelta(finish_reason='COMPLETE', usage=Usage(billed_units=UsageBilledUnits(input_tokens=87.0, output_tokens=19.0, search_units=None, classifications=None), tokens=UsageTokens(input_tokens=1061.0, output_tokens=85.0))) 48 --------------------------------------------------\n```\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere34co = cohere.ClientV2(5 \"COHERE_API_KEY\"6) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1def get_weather(location):2 temperature = {3 \"bern\": \"22°C\",4 \"madrid\": \"24°C\",5 \"brasilia\": \"28°C\",6 }7 loc = location.lower()8 if loc in temperature:9 return [{\"temperature\": {loc: temperature[loc]}}]10 return [{\"temperature\": {loc: \"Unknown\"}}]111213functions_map = {\"get_weather\": get_weather}1415tools = [16 {17 \"type\": \"function\",18 \"function\": {19 \"name\": \"get_weather\",20 \"description\": \"gets the weather of a given location\",21 \"parameters\": {22 \"type\": \"object\",23 \"properties\": {24 \"location\": {25 \"type\": \"string\",26 \"description\": \"the location to get the weather, example: San Francisco.\",27 }28 },29 \"required\": [\"location\"],30 },31 },32 }33]\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"user\",4 \"content\": \"What's the weather in Madrid and Brasilia?\",5 }6]78response = co.chat(9 model=\"command-a-plus-05-2026\", messages=messages, tools=tools10)1112if response.message.tool_calls:13 messages.append(response.message)14 print(response.message.tool_plan, \"\\n\")15 print(response.message.tool_calls)1617import json1819if response.message.tool_calls:20 for tc in response.message.tool_calls:21 tool_result = functions_map[tc.function.name](22 **json.loads(tc.function.arguments)23 )24 tool_content = []25 for data in tool_result:26 # Optional: the \"document\" object can take an \"id\" field for use in citations, otherwise auto-generated27 tool_content.append(28 {29 \"type\": \"document\",30 \"document\": {\"data\": json.dumps(data)},31 }32 )33 messages.append(34 {35 \"role\": \"tool\",36 \"tool_call_id\": tc.id,37 \"content\": tool_content,38 }39 )\n```\n\nExample:\n```text\n1I will use the get_weather tool to find the weather in Madrid and Brasilia. 23[4 ToolCallV2(5 id=\"get_weather_15c2p6g19s8f\",6 type=\"function\",7 function=ToolCallV2Function(8 name=\"get_weather\", arguments='{\"location\":\"Madrid\"}'9 ),10 ),11 ToolCallV2(12 id=\"get_weather_n01pkywy0p2w\",13 type=\"function\",14 function=ToolCallV2Function(15 name=\"get_weather\", arguments='{\"location\":\"Brasilia\"}'16 ),17 ),18]\n```\n\nExample:\n```text\n1response = co.chat_stream(2 model=\"command-a-plus-05-2026\", messages=messages, tools=tools3)45response_text = \"\"6citations = []7for chunk in response:8 if chunk:9 if chunk.type == \"content-delta\":10 response_text += chunk.delta.message.content.text11 print(chunk.delta.message.content.text, end=\"\")12 if chunk.type == \"citation-start\":13 citations.append(chunk.delta.message.citations)1415for citation in citations:16 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1It's currently 24°C in Madrid and 28°C in Brasilia.23start=5 end=9 text='24°C' sources=[ToolSource(type='tool', id='get_weather_15c2p6g19s8f:0', tool_output={'temperature': '{\"madrid\":\"24°C\"}'})] type='TEXT_CONTENT' 45start=24 end=28 text='28°C' sources=[ToolSource(type='tool', id='get_weather_n01pkywy0p2w:0', tool_output={'temperature': '{\"brasilia\":\"28°C\"}'})] type='TEXT_CONTENT'\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.314Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":8,"totalLines":43,"estimatedTokens":4433}}114{"id":"doc-private_deployment_overview_cohere-163f9efb","source":"documentation","title":"Private Deployment Overview | Cohere","url":"https://docs.cohere.com/docs/private-deployment-overview","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.314Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}115{"id":"doc-cohere_on_amazon_web_services_aws_cohere-08e7fe15","source":"documentation","title":"Cohere on Amazon Web Services (AWS) | Cohere","url":"https://docs.cohere.com/docs/cohere-on-aws","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.314Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}116{"id":"doc-elasticsearch_and_cohere_integration_guide_coher-b486b618","source":"documentation","title":"Elasticsearch and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/elasticsearch-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1!pip install elasticsearch_serverless==0.2.0.202310312!pip install cohere==5.2.5\n```\n\nExample:\n```text\n1from elasticsearch_serverless import Elasticsearch, helpers2from getpass import getpass3import cohere4import json5import requests\n```\n\nExample:\n```text\n1ELASTICSEARCH_ENDPOINT = getpass(\"Elastic Endpoint: \")2ELASTIC_API_KEY = getpass(3 \"Elastic encoded API key: \"4) # Use the encoded API key56client = Elasticsearch(7 ELASTICSEARCH_ENDPOINT, api_key=ELASTIC_API_KEY8)910# Confirm the client has connected11print(client.info())\n```\n\nExample:\n```text\n1COHERE_API_KEY = getpass(\"Enter Cohere API key: \")2# Delete the inference model if it already exists3client.options(ignore_status=[404]).inference.delete(4 inference_id=\"cohere_embeddings\"5)67client.inference.put(8 task_type=\"text_embedding\",9 inference_id=\"cohere_embeddings\",10 body={11 \"service\": \"cohere\",12 \"service_settings\": {13 \"api_key\": COHERE_API_KEY,14 \"model_id\": \"embed-v4.0\",15 \"embedding_type\": \"int8\",16 \"similarity\": \"cosine\",17 },18 \"task_settings\": {},19 },20)\n```\n\nExample:\n```text\n1client.indices.delete(2 index=\"cohere-wiki-embeddings\", ignore_unavailable=True3)4client.indices.create(5 index=\"cohere-wiki-embeddings\",6 mappings={7 \"properties\": {8 \"text_semantic\": {9 \"type\": \"semantic_text\",10 \"inference_id\": \"cohere_embeddings\",11 },12 \"text\": {\"type\": \"text\", \"copy_to\": \"text_semantic\"},13 \"wiki_id\": {\"type\": \"integer\"},14 \"url\": {\"type\": \"text\"},15 \"views\": {\"type\": \"float\"},16 \"langs\": {\"type\": \"integer\"},17 \"title\": {\"type\": \"text\"},18 \"paragraph_id\": {\"type\": \"integer\"},19 \"id\": {\"type\": \"integer\"},20 }21 },22)\n```\n\nExample:\n```text\nObjectApiResponse({'acknowledged': True, 'shards_acknowledged': True, 'index': 'cohere-wiki-embeddings'})\n```\n\nExample:\n```text\n1url = \"https://raw.githubusercontent.com/cohere-ai/cohere-developer-experience/main/notebooks/data/embed_jobs_sample_data.jsonl\"2response = requests.get(url)34# Load the response data into a JSON object5jsonl_data = response.content.decode(\"utf-8\").splitlines()67# Prepare the documents to be indexed8documents = []9for line in jsonl_data:10 data_dict = json.loads(line)11 documents.append(12 {13 \"_index\": \"cohere-wiki-embeddings\",14 \"_source\": data_dict,15 }16 )1718# Use the bulk endpoint to index19helpers.bulk(client, documents)2021print(\"Done indexing documents into `cohere-wiki-embeddings` index!\")\n```\n\nExample:\n```text\nDone indexing documents into `cohere-wiki-embeddings` index!\n```\n\nExample:\n```text\n1query = \"When were the semi-finals of the 2022 FIFA world cup played?\"23response = client.search(4 index=\"cohere-wiki-embeddings\",5 size=100,6 query = {7 \"semantic\": {8 \"query\": \"When were the semi-finals of the 2022 FIFA world cup played?\",9 \"field\": \"text_semantic\"10 }11 }12)1314raw_documents = response[\"hits\"][\"hits\"]1516# Display the first 10 results17for document in raw_documents[0:10]:18 print(f'Title: {document[\"_source\"][\"title\"]}\\nText: {document[\"_source\"][\"text\"]}\\n')1920# Format the documents for ranking21documents = []22for hit in response[\"hits\"][\"hits\"]:23 documents.append(hit[\"_source\"][\"text\"])\n```\n\nExample:\n```text\nTitle: 2022 FIFA World CupText: The 2022 FIFA World Cup was an international football tournament contested by the men's national teams of FIFA's member associations and 22nd edition of the FIFA World Cup. It took place in Qatar from 20 November to 18 December 2022, making it the first World Cup held in the Arab world and Muslim world, and the second held entirely in Asia after the 2002 tournament in South Korea and Japan. France were the defending champions, having defeated Croatia 4–2 in the 2018 final. At an estimated cost of over $220 billion, it is the most expensive World Cup ever held to date; this figure is disputed by Qatari officials, including organising CEO Nasser Al Khater, who said the true cost was $8 billion, and other figures related to overall infrastructure development since the World Cup was awarded to Qatar in 2010.Title: 2022 FIFA World CupText: The semi-finals were played on 13 and 14 December. Messi scored a penalty kick before Julián Álvarez scored twice to give Argentina a 3–0 victory over Croatia. Théo Hernandez scored after five minutes as France led Morocco for most of the game and later Randal Kolo Muani scored on 78 minutes to complete a 2–0 victory for France over Morocco as they reached a second consecutive final.Title: 2022 FIFA World CupText: The quarter-finals were played on 9 and 10 December. Croatia and Brazil ended 0–0 after 90 minutes and went to extra time. Neymar scored for Brazil in the 15th minute of extra time. Croatia, however, equalised through Bruno Petković in the second period of extra time. With the match tied, a penalty shootout decided the contest, with Croatia winning the shoot-out 4–2. In the second quarter-final match, Nahuel Molina and Messi scored for Argentina before Wout Weghorst equalised with two goals shortly before the end of the game. The match went to extra time and then penalties, where Argentina would go on to win 4–3. Morocco defeated Portugal 1–0, with Youssef En-Nesyri scoring at the end of the first half. Morocco became the first African and the first Arab nation to advance as far as the semi-finals of the competition. Despite Harry Kane scoring a penalty for England, it was not enough to beat France, who won 2–1 by virtue of goals from Aurélien Tchouaméni and Olivier Giroud, sending them to their second consecutive World Cup semi-final and becoming the first defending champions to reach this stage since Brazil in 1998.Title: 2022 FIFA World CupText: Unlike previous FIFA World Cups, which are typically played in June and July, because of Qatar's intense summer heat and often fairly high humidity, the 2022 World Cup was played in November and December. As a result, the World Cup was unusually staged in the middle of the seasons of domestic association football leagues, which started in late July or August, including all of the major European leagues, which had been obliged to incorporate extended breaks into their domestic schedules to accommodate the World Cup. Major European competitions had scheduled their respective competitions group matches to be played before the World Cup, to avoid playing group matches the following year.Title: 2022 FIFA World CupText: The match schedule was confirmed by FIFA in July 2020. The group stage was set to begin on 21 November, with four matches every day. Later, the schedule was tweaked by moving the Qatar vs Ecuador game to 20 November, after Qatar lobbied FIFA to allow their team to open the tournament. The final was played on 18 December 2022, National Day, at Lusail Stadium.Title: 2022 FIFA World CupText: Owing to the climate in Qatar, concerns were expressed over holding the World Cup in its traditional time frame of June and July. In October 2013, a task force was commissioned to consider alternative dates and report after the 2014 FIFA World Cup in Brazil. On 24 February 2015, the FIFA Task Force proposed that the tournament be played from late November to late December 2022, to avoid the summer heat between May and September and also avoid clashing with the 2022 Winter Olympics in February, the 2022 Winter Paralympics in March and Ramadan in April.Title: 2022 FIFA World CupText: Of the 32 nations qualified to play at the 2022 FIFA World Cup, 24 countries competed at the previous tournament in 2018. Qatar were the only team making their debut in the FIFA World Cup, becoming the first hosts to make their tournament debut since Italy in 1934. As a result, the 2022 tournament was the first World Cup in which none of the teams that earned a spot through qualification were making their debut. The Netherlands, Ecuador, Ghana, Cameroon, and the United States returned to the tournament after missing the 2018 tournament. Canada returned after 36 years, their only prior appearance being in 1986. Wales made their first appearance in 64 years – the longest ever gap for any team, their only previous participation having been in 1958.Title: 2022 FIFA World CupText: After UEFA were guaranteed to host the 2018 event, members of UEFA were no longer in contention to host in 2022. There were five bids remaining for the 2022 FIFA World Cup: Australia, Japan, Qatar, South Korea, and the United States.Title: Cristiano RonaldoText: Ronaldo was named in Portugal's squad for the 2022 FIFA World Cup in Qatar, making it his fifth World Cup. On 24 November, in Portugal's opening match against Ghana, Ronaldo scored a penalty kick and became the first male player to score in five different World Cups. In the last group game against South Korea, Ronaldo received criticism from his own coach for his reaction at being substituted. He was dropped from the starting line-up for Portugal's last 16 match against Switzerland, marking the first time since Euro 2008 that he had not started a game for Portugal in a major international tournament, and the first time Portugal had started a knockout game without Ronaldo in the starting line-up at an international tournament since Euro 2000. He came off the bench late on as Portugal won 6–1, their highest tally in a World Cup knockout game since the 1966 World Cup, with Ronaldo's replacement Gonçalo Ramos scoring a hat-trick. Portugal employed the same strategy in the quarter-finals against Morocco, with Ronaldo once again coming off the bench; in the process, he equalled Bader Al-Mutawa's international appearance record, becoming the joint–most capped male footballer of all time, with 196 caps. Portugal lost 1–0, however, with Morocco becoming the first CAF nation ever to reach the World Cup semi-finals.Title: 2022 FIFA World CupText: The final draw was held at the Doha Exhibition and Convention Center in Doha, Qatar, on 1 April 2022, 19:00 AST, prior to the completion of qualification. The two winners of the inter-confederation play-offs and the winner of the Path A of the UEFA play-offs were not known at the time of the draw. The draw was attended by 2,000 guests and was led by Carli Lloyd, Jermaine Jenas and sports broadcaster Samantha Johnson, assisted by the likes of Cafu (Brazil), Lothar Matthäus (Germany), Adel Ahmed Malalla (Qatar), Ali Daei (Iran), Bora Milutinović (Serbia/Mexico), Jay-Jay Okocha (Nigeria), Rabah Madjer (Algeria), and Tim Cahill (Australia).\n```\n\nExample:\n```text\n1query = \"When were the semi-finals of the 2022 FIFA world cup played?\"23response = client.search(4 index=\"cohere-wiki-embeddings\",5 size=100,6 query={7 \"bool\": {8 \"must\": {9 \"multi_match\": {10 \"query\": \"When were the semi-finals of the 2022 FIFA world cup played?\",11 \"fields\": [\"text\", \"title\"]12 }13 },14 \"should\": {15 \"semantic\": {16 \"query\": \"When were the semi-finals of the 2022 FIFA world cup played?\",17 \"field\": \"text_semantic\"18 }19 },20 }21 }2223)2425raw_documents = response[\"hits\"][\"hits\"]2627# Display the first 10 results28for document in raw_documents[0:10]:29 print(f'Title: {document[\"_source\"][\"title\"]}\\nText: {document[\"_source\"][\"text\"]}\\n')3031# Format the documents for ranking32documents = []33for hit in response[\"hits\"][\"hits\"]:34 documents.append(hit[\"_source\"][\"text\"])\n```\n\nExample:\n```text\n1# Delete the inference model if it already exists2client.options(ignore_status=[404]).inference.delete(inference_id=\"cohere_rerank\")34client.inference.put(5 task_type=\"rerank\",6 inference_id=\"cohere_rerank\",7 body={8 \"service\": \"cohere\",9 \"service_settings\":{10 \"api_key\": COHERE_API_KEY,11 \"model_id\": \"rerank-english-v3.0\"12 },13 \"task_settings\": {14 \"top_n\": 10,15 },16 }17)\n```\n\nExample:\n```text\n1response = client.inference.inference(2 inference_id=\"cohere_rerank\",3 body={4 \"query\": query,5 \"input\": documents,6 \"task_settings\": {7 \"return_documents\": False8 }9 }10)1112# Reconstruct the input documents based on the index provided in the rereank response13ranked_documents = []14for document in response.body[\"rerank\"]:15 ranked_documents.append({16 \"title\": raw_documents[int(document[\"index\"])][\"_source\"][\"title\"],17 \"text\": raw_documents[int(document[\"index\"])][\"_source\"][\"text\"]18 })1920# Print the top 10 results21for document in ranked_documents[0:10]:22 print(f\"Title: {document['title']}\\nText: {document['text']}\\n\")\n```\n\nExample:\n```text\n1co = cohere.Client(COHERE_API_KEY)\n```\n\nExample:\n```text\n1response = co.chat(2 message=query,3 documents=ranked_documents,4 model=\"command-a-03-2025\",5)67source_documents = []8for citation in response.citations:9 for document_id in citation.document_ids:10 if document_id not in source_documents:11 source_documents.append(document_id)1213print(f\"Query: {query}\")14print(f\"Response: {response.text}\")15print(\"Sources:\")16for document in response.documents:17 if document[\"id\"] in source_documents:18 print(f\"{document['title']}: {document['text']}\")\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.315Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":3438}}117{"id":"doc-private_deployment_setting_up_cohere-f9f0799b","source":"documentation","title":"Private Deployment – Setting Up | Cohere","url":"https://docs.cohere.com/docs/private-deployment-setup","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.315Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}118{"id":"doc-private_deployment_usage_cohere-c53e1630","source":"documentation","title":"Private Deployment Usage | Cohere","url":"https://docs.cohere.com/docs/private-deployment-usage","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$pip install -U cohere\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(4 api_key=\"\", # Leave this blank5 base_url=\"<YOUR_DEPLOYMENT_URL>\",6)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.315Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":91}}119{"id":"doc-cohere_embed_on_langchain_integration_guide_cohe-8cfce9c8","source":"documentation","title":"Cohere Embed on LangChain (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/embed-on-langchain","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from langchain_cohere import CohereEmbeddings23# Define the Cohere embedding model4embeddings = CohereEmbeddings(5 cohere_api_key=\"COHERE_API_KEY\", model=\"embed-v4.0\"6)78# Embed a document9text = \"This is a test document.\"10query_result = embeddings.embed_query(text)11print(query_result[:5], \"...\")12doc_result = embeddings.embed_documents([text])13print(doc_result[0][:5], \"...\")\n```\n\nExample:\n```text\n1from langchain_cohere import ChatCohere, CohereEmbeddings2from langchain_text_splitters import CharacterTextSplitter3from langchain_community.vectorstores import Chroma4from langchain_community.document_loaders import WebBaseLoader56user_query = \"what is Cohere Toolkit?\"78llm = ChatCohere(9 cohere_api_key=\"COHERE_API_KEY\",10 model=\"command-a-03-2025\",11 temperature=0,12)1314embeddings = CohereEmbeddings(15 cohere_api_key=\"COHERE_API_KEY\", model=\"embed-v4.0\"16)1718# Load text and split into chunks, you can also use data gathered elsewhere in your application19raw_documents = WebBaseLoader(20 \"https://docs.cohere.com/docs/cohere-toolkit\"21).load()2223text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=0)24documents = text_splitter.split_documents(raw_documents)2526# Create a vector store from the documents and retrieve the most relevant chunks27db = Chroma.from_documents(documents, embeddings)28input_docs = db.as_retriever().invoke(user_query)2930# Ground the answer in the retrieved documents31response = llm.invoke(user_query, documents=input_docs)3233# Print the answer34print(\"Answer:\")35print(response.content)36# Print the citations that ground the answer in the documents37print(\"Citations:\")38print(response.additional_kwargs.get(\"citations\"))\n```\n\nExample:\n```text\n1from langchain_aws import BedrockEmbeddings23# Replace the profile name with the one created in the setup.4embeddings = BedrockEmbeddings(5 credentials_profile_name=\"{PROFILE-NAME}\",6 region_name=\"us-east-1\",7 model_id=\"cohere.embed-english-v3\",8)910embeddings.embed_query(\"This is a content of the document\")\n```\n\nExample:\n```text\n1llm = CohereEmbeddings(2 base_url=\"<YOUR_DEPLOYMENT_URL>\",3 cohere_api_key=\"COHERE_API_KEY\",4 model=\"MODEL_NAME\",5)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.316Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":4,"totalLines":23,"estimatedTokens":600}}120{"id":"doc-cohere_rerank_on_langchain_integration_guide_coh-5f20fcff","source":"documentation","title":"Cohere Rerank on LangChain (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/rerank-on-langchain","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from langchain_classic.retrievers import (2 ContextualCompressionRetriever,3)4from langchain_cohere import (5 ChatCohere,6 CohereEmbeddings,7 CohereRerank,8)9from langchain_text_splitters import CharacterTextSplitter10from langchain_community.vectorstores import Chroma11from langchain_community.document_loaders import WebBaseLoader1213user_query = \"what is Cohere Toolkit?\"1415# Define the Cohere LLM16llm = ChatCohere(17 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"18)1920# Define the Cohere embedding model21embeddings = CohereEmbeddings(22 cohere_api_key=\"COHERE_API_KEY\", model=\"embed-english-light-v3.0\"23)2425# Load text and split into chunks, you can also use data gathered elsewhere in your application26raw_documents = WebBaseLoader(27 \"https://docs.cohere.com/docs/cohere-toolkit\"28).load()29text_splitter = CharacterTextSplitter(30 chunk_size=1000, chunk_overlap=031)32documents = text_splitter.split_documents(raw_documents)3334# Create a vector store from the documents35db = Chroma.from_documents(documents, embeddings)3637# Create Cohere's reranker with the vector DB using Cohere's embeddings as the base retriever38reranker = CohereRerank(39 cohere_api_key=\"COHERE_API_KEY\", model=\"rerank-english-v3.0\"40)4142compression_retriever = ContextualCompressionRetriever(43 base_compressor=reranker, base_retriever=db.as_retriever()44)45compressed_docs = compression_retriever.invoke(user_query)46# Print the reranked documents from using the embeddings and reranker47print(compressed_docs)4849# Ground the answer in the reranked documents50response = llm.invoke(user_query, documents=compressed_docs)5152# Print the answer53print(\"Answer:\")54print(response.content)55# Print the citations that ground the answer in the documents56print(\"Citations:\")57print(response.additional_kwargs.get(\"citations\"))\n```\n\nExample:\n```text\n1llm = CohereRerank(2 base_url=\"<YOUR_DEPLOYMENT_URL>\",3 cohere_api_key=\"COHERE_API_KEY\",4 model=\"MODEL_NAME\",5)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.316Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":552}}121{"id":"doc-cohere_models_on_amazon_bedrock_cohere-43786273","source":"documentation","title":"Cohere Models on Amazon Bedrock | Cohere","url":"https://docs.cohere.com/docs/amazon-bedrock","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.BedrockClient(4 aws_region=\"us-east-1\",5 aws_access_key=\"...\",6 aws_secret_key=\"...\",7 aws_session_token=\"...\",8)910# Input parameters for embed. In this example we are embedding hacker news post titles.11texts = [12 \"Interesting (Non software) books?\",13 \"Non-tech books that have helped you grow professionally?\",14 \"I sold my company last month for $5m. What do I do with the money?\",15 \"How are you getting through (and back from) burning out?\",16 \"I made $24k over the last month. Now what?\",17 \"What kind of personal financial investment do you do?\",18 \"Should I quit the field of software development?\",19]20input_type = \"clustering\"21truncate = \"NONE\" # optional22model_id = (23 \"cohere.embed-english-v3\" # or \"cohere.embed-multilingual-v3\"24)252627# Invoke the model and print the response28result = co.embed(29 model=model_id,30 input_type=input_type,31 texts=texts,32 truncate=truncate,33) # aws_client.invoke_model(**params)3435print(result)\n```\n\nExample:\n```text\n1import cohere23co = cohere.BedrockClient(4 aws_region=\"us-east-1\",5 aws_access_key=\"...\",6 aws_secret_key=\"...\",7 aws_session_token=\"...\",8)910result = co.chat(11 message=\"Write a LinkedIn post about starting a career in tech:\",12 model=\"cohere.command-r-plus-v1:0\", # or 'cohere.command-r-v1:0'13)1415print(result)\n```\n\nExample:\n```text\n1import cohere23co = cohere.BedrockClientV2(4 aws_region=\"us-west-2\", # pick a region where the model is available5 aws_access_key=\"...\",6 aws_secret_key=\"...\",7 aws_session_token=\"...\",8)910docs = [11 \"Carson City is the capital city of the American state of Nevada.\",12 \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean. Its capital is Saipan.\",13 \"Capitalization or capitalisation in English grammar is the use of a capital letter at the start of a word. English usage varies from capitalization in other languages.\",14 \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district.\",15 \"Capital punishment has existed in the United States since beforethe United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states.\",16]1718response = co.rerank(19 model=\"cohere.rerank-v3-5:0\",20 query=\"What is the capital of the United States?\",21 documents=docs,22 top_n=3,23)2425print(response)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.316Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":3,"totalLines":18,"estimatedTokens":681}}122{"id":"doc-citations_for_tool_use_function_calling_cohere-58cd521b","source":"documentation","title":"Citations for tool use (function calling) | Cohere","url":"https://docs.cohere.com/docs/tool-use-citations","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere3import json45co = cohere.ClientV2(6 \"COHERE_API_KEY\"7) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1def get_weather(location):2 temperature = {3 \"bern\": \"22°C\",4 \"madrid\": \"24°C\",5 \"brasilia\": \"28°C\",6 }7 loc = location.lower()8 if loc in temperature:9 return [{\"temperature\": {loc: temperature[loc]}}]10 return [{\"temperature\": {loc: \"Unknown\"}}]111213functions_map = {\"get_weather\": get_weather}1415tools = [16 {17 \"type\": \"function\",18 \"function\": {19 \"name\": \"get_weather\",20 \"description\": \"gets the weather of a given location\",21 \"parameters\": {22 \"type\": \"object\",23 \"properties\": {24 \"location\": {25 \"type\": \"string\",26 \"description\": \"the location to get the weather, example: San Francisco.\",27 }28 },29 \"required\": [\"location\"],30 },31 },32 }33]\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"user\",4 \"content\": \"What's the weather in Madrid and Brasilia?\",5 }6]78response = co.chat(9 model=\"command-a-plus-05-2026\", messages=messages, tools=tools10)1112if response.message.tool_calls:13 messages.append(response.message)1415 for tc in response.message.tool_calls:16 tool_result = functions_map[tc.function.name](17 **json.loads(tc.function.arguments)18 )19 tool_content = []20 for data in tool_result:21 tool_content.append(22 {23 \"type\": \"document\",24 \"document\": {\"data\": json.dumps(data)},25 }26 )27 messages.append(28 {29 \"role\": \"tool\",30 \"tool_call_id\": tc.id,31 \"content\": tool_content,32 }33 )\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\", messages=messages, tools=tools3)45messages.append(6 {\"role\": \"assistant\", \"content\": response.message.content[0].text}7)89print(response.message.content[0].text)1011for citation in response.message.citations:12 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1It is currently 24°C in Madrid and 28°C in Brasilia.23start=16 end=20 text='24°C' sources=[ToolSource(type='tool', id='get_weather_14brd1n2kfqj:0', tool_output={'temperature': '{\"madrid\":\"24°C\"}'})] type='TEXT_CONTENT' 45start=35 end=39 text='28°C' sources=[ToolSource(type='tool', id='get_weather_vdr9cvj619fk:0', tool_output={'temperature': '{\"brasilia\":\"28°C\"}'})] type='TEXT_CONTENT'\n```\n\nExample:\n```text\n1response = co.chat_stream(2 model=\"command-a-plus-05-2026\", messages=messages, tools=tools3)45response_text = \"\"6citations = []7for chunk in response:8 if chunk:9 if chunk.type == \"content-delta\":10 response_text += chunk.delta.message.content.text11 print(chunk.delta.message.content.text, end=\"\")12 if chunk.type == \"citation-start\":13 citations.append(chunk.delta.message.citations)1415messages.append({\"role\": \"assistant\", \"content\": response_text})1617for citation in citations:18 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1It is currently 24°C in Madrid and 28°C in Brasilia.23start=16 end=20 text='24°C' sources=[ToolSource(type='tool', id='get_weather_dkf0akqdazjb:0', tool_output={'temperature': '{\"madrid\":\"24°C\"}'})] type='TEXT_CONTENT' 45start=35 end=39 text='28°C' sources=[ToolSource(type='tool', id='get_weather_gh65bt2tcdy1:0', tool_output={'temperature': '{\"brasilia\":\"28°C\"}'})] type='TEXT_CONTENT'\n```\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere3import json45co = cohere.ClientV2(6 \"COHERE_API_KEY\"7) # Get your free API key here: https://dashboard.cohere.com/api-keys89messages = [10 {11 \"role\": \"user\",12 \"content\": \"What's the weather in Madrid and Brasilia?\",13 },14 {15 \"role\": \"assistant\",16 \"tool_plan\": \"I will search for the weather in Madrid and Brasilia.\",17 \"tool_calls\": [18 {19 \"id\": \"get_weather_dkf0akqdazjb\",20 \"type\": \"function\",21 \"function\": {22 \"name\": \"get_weather\",23 \"arguments\": '{\"location\":\"Madrid\"}',24 },25 },26 {27 \"id\": \"get_weather_gh65bt2tcdy1\",28 \"type\": \"function\",29 \"function\": {30 \"name\": \"get_weather\",31 \"arguments\": '{\"location\":\"Brasilia\"}',32 },33 },34 ],35 },36 {37 \"role\": \"tool\",38 \"tool_call_id\": \"get_weather_dkf0akqdazjb\",39 \"content\": [40 {41 \"type\": \"document\",42 \"document\": {43 \"data\": '{\"temperature\": {\"madrid\": \"24°C\"}}',44 \"id\": \"1\",45 },46 }47 ],48 },49 {50 \"role\": \"tool\",51 \"tool_call_id\": \"get_weather_gh65bt2tcdy1\",52 \"content\": [53 {54 \"type\": \"document\",55 \"document\": {56 \"data\": '{\"temperature\": {\"brasilia\": \"28°C\"}}',57 \"id\": \"2\",58 },59 }60 ],61 },62]\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\", messages=messages, tools=tools3)45print(response.message.content[0].text)67for citation in response.message.citations:8 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1It's 24°C in Madrid and 28°C in Brasilia.23start=5 end=9 text='24°C' sources=[ToolSource(type='tool', id='1', tool_output={'temperature': '{\"madrid\":\"24°C\"}'})] type='TEXT_CONTENT' 45start=24 end=28 text='28°C' sources=[ToolSource(type='tool', id='2', tool_output={'temperature': '{\"brasilia\":\"28°C\"}'})] type='TEXT_CONTENT'\n```\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere3import json45co = cohere.ClientV2(6 \"COHERE_API_KEY\"7) # Get your free API key here: https://dashboard.cohere.com/api-keys89response = co.chat_stream(10 model=\"command-a-plus-05-2026\",11 messages=messages,12 tools=tools,13 citation_options={\"mode\": \"accurate\"},14)1516response_text = \"\"17citations = []18for chunk in response:19 if chunk:20 if chunk.type == \"content-delta\":21 response_text += chunk.delta.message.content.text22 print(chunk.delta.message.content.text, end=\"\")23 if chunk.type == \"citation-start\":24 citations.append(chunk.delta.message.citations)2526print(\"\\n\")27for citation in citations:28 print(citation, \"\\n\")\n```\n\nExample:\n```text\n1It is currently 24°C in Madrid and 28°C in Brasilia.23start=16 end=20 text='24°C' sources=[ToolSource(type='tool', id='1', tool_output={'temperature': '{\"madrid\":\"24°C\"}'})] type='TEXT_CONTENT' 45start=35 end=39 text='28°C' sources=[ToolSource(type='tool', id='2', tool_output={'temperature': '{\"brasilia\":\"28°C\"}'})] type='TEXT_CONTENT'\n```\n\nExample:\n```text\n1response = co.chat_stream(2 model=\"command-a-plus-05-2026\",3 messages=messages,4 tools=tools,5 citation_options={\"mode\": \"fast\"},6)78response_text = \"\"9for chunk in response:10 if chunk:11 if chunk.type == \"content-delta\":12 response_text += chunk.delta.message.content.text13 print(chunk.delta.message.content.text, end=\"\")14 if chunk.type == \"citation-start\":15 print(16 f\" [{chunk.delta.message.citations.sources[0].id}]\",17 end=\"\",18 )\n```\n\nExample:\n```text\n1It is currently 24°C [1] in Madrid and 28°C [2] in Brasilia.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.317Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":14,"totalLines":73,"estimatedTokens":1999}}123{"id":"doc-weaviate_and_cohere_integration_guide_cohere-4640a30e","source":"documentation","title":"Weaviate and Cohere (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/weaviate-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from google.colab import userdata23weaviate_url = userdata.get(\"WEAVIATE_ENDPOINT\")4weaviate_key = userdata.get(\"WEAVIATE_API_KEY\")5cohere_key = userdata.get(\"COHERE_API_KEY\")\n```\n\nExample:\n```text\n1!pip install -U weaviate-client -q\n```\n\nExample:\n```text\n1# Import the weaviate modules to interact with the Weaviate vector database2import weaviate3from weaviate.classes.init import Auth45# Define headers for the API requests, including the Cohere API key6headers = {7 \"X-Cohere-Api-Key\": cohere_key,8}910# Connect to the Weaviate cloud instance11client = weaviate.connect_to_weaviate_cloud(12 cluster_url=weaviate_url, # `weaviate_url`: your Weaviate URL13 auth_credentials=Auth.api_key(14 weaviate_key15 ), # `weaviate_key`: your Weaviate API key16 headers=headers,17)\n```\n\nExample:\n```text\n1from weaviate.classes.config import Configure23# This is where the \"Healthcare_Compliance\" collection is created in Weaviate.4client.collections.create(5 \"Healthcare_Compliance\",6 vectorizer_config=[7 # Configure a named vectorizer using Cohere's model8 Configure.NamedVectors.text2vec_cohere(9 name=\"title_vector\", # Name of the vectorizer10 source_properties=[11 \"title\"12 ], # Property to vectorize (in this case, the \"title\" field)13 model=\"embed-english-v3.0\", # Cohere model to use for vectorization14 )15 ],16)\n```\n\nExample:\n```text\n<weaviate.collections.collection.sync.Collection at 0x7f48a5604590>\n```\n\nExample:\n```text\n1# Define the list of healthcare compliance documents23hl_compliance_docs = [4 {5 \"title\": \"HIPAA Compliance Guide\",6 \"description\": \"Comprehensive overview of HIPAA regulations, including patient privacy rules, data security standards, and breach notification requirements.\",7 },8 {9 \"title\": \"FDA Drug Approval Process\",10 \"description\": \"Detailed explanation of the FDA's drug approval process, covering clinical trials, safety reviews, and post-market surveillance.\",11 },12 {13 \"title\": \"Telemedicine Regulations\",14 \"description\": \"Analysis of state and federal regulations governing telemedicine practices, including licensing, reimbursement, and patient consent.\",15 },16 {17 \"title\": \"Healthcare Data Security\",18 \"description\": \"Best practices for securing healthcare data, including encryption, access controls, and incident response planning.\",19 },20 {21 \"title\": \"Medicare and Medicaid Billing\",22 \"description\": \"Guide to billing and reimbursement processes for Medicare and Medicaid, including coding, claims submission, and audit compliance.\",23 },24 {25 \"title\": \"Patient Rights and Consent\",26 \"description\": \"Overview of patient rights under federal and state laws, including informed consent, access to medical records, and end-of-life decisions.\",27 },28 {29 \"title\": \"Healthcare Fraud and Abuse\",30 \"description\": \"Explanation of laws and regulations related to healthcare fraud, including the False Claims Act, Anti-Kickback Statute, and Stark Law.\",31 },32 {33 \"title\": \"Occupational Safety in Healthcare\",34 \"description\": \"Guidelines for ensuring workplace safety in healthcare settings, including infection control, hazard communication, and emergency preparedness.\",35 },36 {37 \"title\": \"Health Insurance Portability\",38 \"description\": \"Discussion of COBRA and other laws ensuring continuity of health insurance coverage during job transitions or life events.\",39 },40 {41 \"title\": \"Medical Device Regulations\",42 \"description\": \"Overview of FDA regulations for medical devices, including classification, premarket approval, and post-market surveillance.\",43 },44 {45 \"title\": \"Electronic Health Records (EHR) Standards\",46 \"description\": \"Explanation of standards and regulations for EHR systems, including interoperability, data exchange, and patient privacy.\",47 },48 {49 \"title\": \"Pharmacy Regulations\",50 \"description\": \"Overview of state and federal regulations governing pharmacy practices, including prescription drug monitoring, compounding, and controlled substances.\",51 },52 {53 \"title\": \"Mental Health Parity Act\",54 \"description\": \"Analysis of the Mental Health Parity and Addiction Equity Act, ensuring equal coverage for mental health and substance use disorder treatment.\",55 },56 {57 \"title\": \"Healthcare Quality Reporting\",58 \"description\": \"Guide to quality reporting requirements for healthcare providers, including measures, submission processes, and performance benchmarks.\",59 },60 {61 \"title\": \"Advance Directives and End-of-Life Care\",62 \"description\": \"Overview of laws and regulations governing advance directives, living wills, and end-of-life care decisions.\",63 },64]6566# Retrieve the \"Healthcare_Compliance\" collection from the Weaviate client67collection = client.collections.get(\"Healthcare_Compliance\")6869# Use a dynamic batch process to add multiple documents to the collection efficiently70with collection.batch.dynamic() as batch:71 for src_obj in hl_compliance_docs:72 # Add each document to the batch, specifying the \"title\" and \"description\" properties73 batch.add_object(74 properties={75 \"title\": src_obj[\"title\"],76 \"description\": src_obj[\"description\"],77 },78 )\n```\n\nExample:\n```text\n1# Import the MetadataQuery class from weaviate.classes.query to handle metadata in queries2from weaviate.classes.query import MetadataQuery34# Retrieve the \"Healthcare_Compliance\" collection from the Weaviate client5collection = client.collections.get(\"Healthcare_Compliance\")67# Perform a near_text search for documents related to \"policies related to drug compounding\"8response = collection.query.near_text(9 query=\"policies related to drug compounding\", # Search query10 limit=2, # Limit the number of results to 211 return_metadata=MetadataQuery(12 distance=True13 ), # Include distance metadata in the results14)1516# Iterate over the retrieved objects and print their details17for obj in response.objects:18 title = obj.properties.get(\"title\")19 description = obj.properties.get(\"description\")20 distance = (21 obj.metadata.distance22 ) # Get the distance metadata (A lower value for a distance means that two vectors are closer to one another than a higher value)23 print(f\"Title: {title}\")24 print(f\"Description: {description}\")25 print(f\"Distance: {distance}\")26 print(\"-\" * 50)\n```\n\nExample:\n```text\nTitle: Pharmacy RegulationsDescription: Overview of state and federal regulations governing pharmacy practices, including prescription drug monitoring, compounding, and controlled substances.Distance: 0.5904817581176758--------------------------------------------------Title: FDA Drug Approval ProcessDescription: Detailed explanation of the FDA's drug approval process, covering clinical trials, safety reviews, and post-market surveillance.Distance: 0.6262975931167603--------------------------------------------------\n```\n\nExample:\n```text\n1# Import the weaviate module to interact with the Weaviate vector database2import weaviate3from weaviate.classes.init import Auth45# Define headers for the API requests, including the Cohere API key6headers = {7 \"X-Cohere-Api-Key\": cohere_key,8}910# Connect to the Weaviate cloud instance11client = weaviate.connect_to_weaviate_cloud(12 cluster_url=weaviate_url, # `weaviate_url`: your Weaviate URL13 auth_credentials=Auth.api_key(14 weaviate_key15 ), # `weaviate_key`: your Weaviate API key16 headers=headers, # Include the Cohere API key in the headers17)\n```\n\nExample:\n```text\n1from weaviate.classes.config import Configure, Property, DataType23# Create a new collection named \"Legal_Docs\" in the Weaviate database4client.collections.create(5 name=\"Legal_Docs\",6 properties=[7 # Define a property named \"title\" with data type TEXT8 Property(name=\"title\", data_type=DataType.TEXT),9 ],10 # Configure the vectorizer to use Cohere's text2vec model11 vectorizer_config=Configure.Vectorizer.text2vec_cohere(12 model=\"embed-english-v3.0\" # Specify the Cohere model to use for vectorization13 ),14 # Configure the reranker to use Cohere's rerank model15 reranker_config=Configure.Reranker.cohere(16 model=\"rerank-english-v3.0\" # Specify the Cohere model to use for reranking17 ),18)\n```\n\nExample:\n```text\n1legal_documents = [2 {3 \"title\": \"Contract Law Basics\",4 \"description\": \"An in-depth introduction to contract law, covering essential elements such as offer, acceptance, consideration, and mutual assent. Explores types of contracts, including express, implied, and unilateral contracts, as well as remedies for breach of contract, such as damages, specific performance, and rescission.\",5 },6 {7 \"title\": \"Intellectual Property Rights\",8 \"description\": \"Comprehensive overview of intellectual property laws, including patents, trademarks, copyrights, and trade secrets. Discusses the process of obtaining patents, trademark registration, and copyright protection, as well as strategies for enforcing intellectual property rights and defending against infringement claims.\",9 },10 {11 \"title\": \"Employment Law Guide\",12 \"description\": \"Detailed guide to employment laws, covering hiring practices, termination procedures, anti-discrimination laws, and workplace safety regulations. Includes information on employee rights, such as minimum wage, overtime pay, and family and medical leave, as well as employer obligations under federal and state laws.\",13 },14 {15 \"title\": \"Criminal Law Procedures\",16 \"description\": \"Step-by-step explanation of criminal law procedures, from arrest and booking to trial and sentencing. Covers the rights of the accused, including the right to counsel, the right to remain silent, and the right to a fair trial, as well as rules of evidence and burden of proof in criminal cases.\",17 },18 {19 \"title\": \"Real Estate Transactions\",20 \"description\": \"Comprehensive guide to real estate transactions, including purchase agreements, title searches, property inspections, and closing processes. Discusses common issues such as title defects, financing contingencies, and property disclosures, as well as the role of real estate agents and attorneys in the transaction process.\",21 },22 {23 \"title\": \"Corporate Governance\",24 \"description\": \"In-depth overview of corporate governance principles, including the roles and responsibilities of boards of directors, shareholder rights, and compliance with securities laws. Explores best practices for board composition, executive compensation, and risk management, as well as strategies for maintaining transparency and accountability in corporate decision-making.\",25 },26 {27 \"title\": \"Family Law Overview\",28 \"description\": \"Comprehensive introduction to family law, covering marriage, divorce, child custody, child support, and adoption processes. Discusses the legal requirements for marriage and divorce, factors considered in child custody determinations, and the rights and obligations of adoptive parents under state and federal laws.\",29 },30 {31 \"title\": \"Tax Law for Businesses\",32 \"description\": \"Detailed guide to tax laws affecting businesses, including corporate income tax, payroll taxes, sales and use taxes, and tax deductions. Explores tax planning strategies, such as deferring income and accelerating expenses, as well as compliance requirements and penalties for non-compliance with tax laws.\",33 },34 {35 \"title\": \"Immigration Law Basics\",36 \"description\": \"Comprehensive overview of immigration laws, including visa categories, citizenship requirements, and deportation processes. Discusses the rights and obligations of immigrants, including access to public benefits and protection from discrimination, as well as the role of immigration attorneys in navigating the immigration system.\",37 },38 {39 \"title\": \"Environmental Regulations\",40 \"description\": \"In-depth overview of environmental laws and regulations, including air and water quality standards, hazardous waste management, and endangered species protection. Explores the role of federal and state agencies in enforcing environmental laws, as well as strategies for businesses to achieve compliance and minimize environmental impact.\",41 },42 {43 \"title\": \"Consumer Protection Laws\",44 \"description\": \"Comprehensive guide to consumer protection laws, including truth in advertising, product safety, and debt collection practices. Discusses the rights of consumers under federal and state laws, such as the right to sue for damages and the right to cancel certain contracts, as well as the role of government agencies in enforcing consumer protection laws.\",45 },46 {47 \"title\": \"Estate Planning Essentials\",48 \"description\": \"Detailed overview of estate planning, including wills, trusts, powers of attorney, and advance healthcare directives. Explores strategies for minimizing estate taxes, protecting assets from creditors, and ensuring that assets are distributed according to the individual's wishes after death.\",49 },50 {51 \"title\": \"Bankruptcy Law Overview\",52 \"description\": \"Comprehensive introduction to bankruptcy law, including Chapter 7 and Chapter 13 bankruptcy proceedings. Discusses the eligibility requirements for filing bankruptcy, the process of liquidating assets and discharging debts, and the impact of bankruptcy on credit scores and future financial opportunities.\",53 },54 {55 \"title\": \"International Trade Law\",56 \"description\": \"In-depth overview of international trade laws, including tariffs, quotas, and trade agreements. Explores the role of international organizations such as the World Trade Organization (WTO) in regulating global trade, as well as strategies for businesses to navigate trade barriers and comply with international trade regulations.\",57 },58 {59 \"title\": \"Healthcare Law and Regulations\",60 \"description\": \"Comprehensive guide to healthcare laws and regulations, including patient privacy rights, healthcare provider licensing, and medical malpractice liability. Discusses the impact of laws such as the Affordable Care Act (ACA) and the Health Insurance Portability and Accountability Act (HIPAA) on healthcare providers and patients, as well as strategies for ensuring compliance with healthcare regulations.\",61 },62]\n```\n\nExample:\n```text\n1# Retrieve the \"Legal_Docs\" collection from the Weaviate client2collection = client.collections.get(\"Legal_Docs\")34# Use a dynamic batch process to add multiple documents to the collection efficiently5with collection.batch.dynamic() as batch:6 for src_obj in legal_documents:7 # Add each document to the batch, specifying the \"title\" and \"description\" properties8 batch.add_object(9 properties={10 \"title\": src_obj[\"title\"],11 \"description\": src_obj[\"description\"],12 },13 )\n```\n\nExample:\n```text\n1search_query = \"eligibility requirements for filing bankruptcy\"\n```\n\nExample:\n```text\n1# Import the MetadataQuery class from weaviate.classes.query to handle metadata in queries2from weaviate.classes.query import MetadataQuery34# Retrieve the \"Legal_Docs\" collection from the Weaviate client5collection = client.collections.get(\"Legal_Docs\")67# Perform a near_text semantic search for documents8response = collection.query.near_text(9 query=search_query, # Search query10 limit=3, # Limit the number of results to 311 return_metadata=MetadataQuery(distance=True) # Include distance metadata in the results12)1314print(\"Semantic Search\")15print(\"*\" * 50)1617# Iterate over the retrieved objects and print their details18for obj in response.objects:19 title = obj.properties.get(\"title\")20 description = obj.properties.get(\"description\")21 metadata_distance = obj.metadata.distance22 print(f\"Title: {title}\")23 print(f\"Description: {description}\")24 print(f\"Metadata Distance: {metadata_distance}\")25 print(\"-\" * 50)\n```\n\nExample:\n```text\nSemantic Search**************************************************Title: Bankruptcy Law OverviewDescription: Comprehensive introduction to bankruptcy law, including Chapter 7 and Chapter 13 bankruptcy proceedings. Discusses the eligibility requirements for filing bankruptcy, the process of liquidating assets and discharging debts, and the impact of bankruptcy on credit scores and future financial opportunities.Metadata Distance: 0.41729819774627686--------------------------------------------------Title: Tax Law for BusinessesDescription: Detailed guide to tax laws affecting businesses, including corporate income tax, payroll taxes, sales and use taxes, and tax deductions. Explores tax planning strategies, such as deferring income and accelerating expenses, as well as compliance requirements and penalties for non-compliance with tax laws.Metadata Distance: 0.6903179883956909--------------------------------------------------Title: Consumer Protection LawsDescription: Comprehensive guide to consumer protection laws, including truth in advertising, product safety, and debt collection practices. Discusses the rights of consumers under federal and state laws, such as the right to sue for damages and the right to cancel certain contracts, as well as the role of government agencies in enforcing consumer protection laws.Metadata Distance: 0.7075160145759583--------------------------------------------------\n```\n\nExample:\n```text\n1# Import the Rerank class from weaviate.classes.query to enable reranking in queries2from weaviate.classes.query import Rerank34# Perform a near_text search with reranking for documents related to \"property contracts and zoning regulations\"5rerank_response = collection.query.near_text(6 query=search_query,7 limit=3,8 rerank=Rerank(9 prop=\"description\", # Property to rerank based on (description in this case)10 query=search_query, # Query to use for reranking11 ),12)1314# Display the reranked search results15print(\"Reranked Search Results:\")16for obj in rerank_response.objects:17 title = obj.properties.get(\"title\")18 description = obj.properties.get(\"description\")19 rerank_score = getattr(20 obj.metadata, \"rerank_score\", None21 ) # Get the rerank score metadata22 print(f\"Title: {title}\")23 print(f\"Description: {description}\")24 print(f\"Rerank Score: {rerank_score}\")25 print(\"-\" * 50)\n```\n\nExample:\n```text\nReranked Search Results:Title: Bankruptcy Law OverviewDescription: Comprehensive introduction to bankruptcy law, including Chapter 7 and Chapter 13 bankruptcy proceedings. Discusses the eligibility requirements for filing bankruptcy, the process of liquidating assets and discharging debts, and the impact of bankruptcy on credit scores and future financial opportunities.Rerank Score: 0.8951567--------------------------------------------------Title: Tax Law for BusinessesDescription: Detailed guide to tax laws affecting businesses, including corporate income tax, payroll taxes, sales and use taxes, and tax deductions. Explores tax planning strategies, such as deferring income and accelerating expenses, as well as compliance requirements and penalties for non-compliance with tax laws.Rerank Score: 7.071895e-06--------------------------------------------------Title: Consumer Protection LawsDescription: Comprehensive guide to consumer protection laws, including truth in advertising, product safety, and debt collection practices. Discusses the rights of consumers under federal and state laws, such as the right to sue for damages and the right to cancel certain contracts, as well as the role of government agencies in enforcing consumer protection laws.Rerank Score: 6.4895394e-06--------------------------------------------------\n```\n\nExample:\n```text\n1from weaviate.classes.config import Configure2from weaviate.classes.generate import GenerativeConfig34# Create a new collection named \"Legal_Docs\" in the Weaviate database5client.collections.create(6 name=\"Legal_Docs_RAG\",7 properties=[8 # Define a property named \"title\" with data type TEXT9 Property(name=\"title\", data_type=DataType.TEXT),10 ],11 # Configure the vectorizer to use Cohere's text2vec model12 vectorizer_config=Configure.Vectorizer.text2vec_cohere(13 model=\"embed-english-v3.0\" # Specify the Cohere model to use for vectorization14 ),15 # Configure the reranker to use Cohere's rerank model16 reranker_config=Configure.Reranker.cohere(17 model=\"rerank-english-v3.0\" # Specify the Cohere model to use for reranking18 ),19 # Configure the generative model to use Cohere's command r plus model20 generative_config=Configure.Generative.cohere(21 model=\"command-r-plus\"22 ),23)\n```\n\nExample:\n```text\n<weaviate.collections.collection.sync.Collection at 0x7f48afc06410>\n```\n\nExample:\n```text\n1# Retrieve the \"Legal_Docs_RAG\" collection from the Weaviate client2collection = client.collections.get(\"Legal_Docs_RAG\")34# Use a dynamic batch process to add multiple documents to the collection efficiently5with collection.batch.dynamic() as batch:6 for src_obj in legal_documents:7 # Add each document to the batch, specifying the \"title\" and \"description\" properties8 batch.add_object(9 properties={10 \"title\": src_obj[\"title\"],11 \"description\": src_obj[\"description\"],12 },13 )\n```\n\nExample:\n```text\n1from weaviate.classes.config import Configure2from weaviate.classes.generate import GenerativeConfig34# To generate text for each object in the search results, use the single prompt method.5# The example below generates outputs for each of the n search results, where n is specified by the limit parameter.67collection = client.collections.get(\"Legal_Docs_RAG\")8response = collection.generate.near_text(9 query=search_query,10 limit=1,11 single_prompt=\"Translate this into French - {title}: {description}\",12)1314for obj in response.objects:15 print(\"Retrieved results\")16 print(\"-----------------\")17 print(obj.properties[\"title\"])18 print(obj.properties[\"description\"])19 print(\"Generated output\")20 print(\"-----------------\")21 print(obj.generated)\n```\n\nExample:\n```text\nRetrieved results-----------------Bankruptcy Law OverviewComprehensive introduction to bankruptcy law, including Chapter 7 and Chapter 13 bankruptcy proceedings. Discusses the eligibility requirements for filing bankruptcy, the process of liquidating assets and discharging debts, and the impact of bankruptcy on credit scores and future financial opportunities.Generated output-----------------Voici une traduction possible :Aperçu du droit des faillites : Introduction complète au droit des faillites, y compris les procédures de faillite en vertu des chapitres 7 et 13. Discute des conditions d'admissibilité pour déposer une demande de faillite, du processus de liquidation des actifs et de libération des dettes, ainsi que de l'impact de la faillite sur les cotes de crédit et les opportunités financières futures.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.318Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":22,"totalLines":113,"estimatedTokens":5965}}124{"id":"doc-an_amazon_sagemaker_setup_guide_cohere-23c99c9b","source":"documentation","title":"An Amazon SageMaker Setup Guide | Cohere","url":"https://docs.cohere.com/docs/amazon-sagemaker-setup-guide","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.SagemakerClient(4 aws_region=\"us-east-1\",5 aws_access_key=\"...\",6 aws_secret_key=\"...\",7 aws_session_token=\"...\",8)910# Input parameters for embed. In this example we are embedding hacker news post titles.11texts = [12 \"Interesting (Non software) books?\",13 \"Non-tech books that have helped you grow professionally?\",14 \"I sold my company last month for $5m. What do I do with the money?\",15 \"How are you getting through (and back from) burning out?\",16 \"I made $24k over the last month. Now what?\",17 \"What kind of personal financial investment do you do?\",18 \"Should I quit the field of software development?\",19]20input_type = \"clustering\"21truncate = \"NONE\" # optional22model_id = \"<YOUR ENDPOINT NAME>\" # On SageMaker, you create a model name that you'll pass here.232425# Invoke the model and print the response26result = co.embed(27 model=model_id,28 input_type=input_type,29 texts=texts,30 truncate=truncate,31)3233print(result)\n```\n\nExample:\n```text\n1import cohere23co = cohere.SagemakerClient(4 aws_region=\"us-east-1\",5 aws_access_key=\"...\",6 aws_secret_key=\"...\",7 aws_session_token=\"...\",8)910# Invoke the model and print the response11result = co.chat(message=\"Write a LinkedIn post about starting a career in tech:\",12 model=\"<YOUR ENDPOINT NAME>\") # On SageMaker, you create a model name that you'll pass here. 1314print(result)\n```\n\nExample:\n```text\n1import cohere23co = cohere.SagemakerClient(4 aws_region=\"us-east-1\",5 aws_access_key=\"...\",6 aws_secret_key=\"...\",7 aws_session_token=\"...\",8)910# Set up your documents and query11query = \"YOUR QUERY\"12docs = [13 \"String 1\",14 \"String 2\"15]1617# Invoke the model and print the response18results = co.rerank(19 model=\"<YOUR RERANK-V4.0 ENDPOINT NAME>\", # On SageMaker, you create a model name that you'll pass here.20 query=query,21 documents=docs,22 top_n=2,23)2425print(result)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.319Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":3,"totalLines":18,"estimatedTokens":546}}125{"id":"doc-aws_private_deployment_guide_ec2_and_eks_cohere-bfdb6e23","source":"documentation","title":"AWS Private Deployment Guide (EC2 and EKS) | Cohere","url":"https://docs.cohere.com/docs/aws-private-deployment","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$sudo apt install -y ubuntu-drivers-common$sudo ubuntu-drivers install$sudo apt install nvidia-cuda-toolkit\n```\n\nExample:\n```text\n$curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list$sed -i -e '/experimental/ s/^#//g' /etc/apt/sources.list.d/nvidia-container-toolkit.list$sudo apt-get update$sudo apt-get install -y nvidia-container-toolkit\n```\n\nExample:\n```text\n$sudo reboot\n```\n\nExample:\n```text\n$nvidia-smi\n```\n\nExample:\n```text\n$sudo apt-get update$sudo apt-get install docker.io -ysudo systemctl start docker$sudo docker run hello-world$sudo systemctl enable docker$docker --version\n```\n\nExample:\n```text\n$export CUSTOMER_TAG=proxy.replicated.com/proxy/cohere/us-docker.pkg.dev/cohere-artifacts/replicated/single-serving-embed-multilingual-03:<YOUR_MODEL_TAG>$export LICENSE_ID=\"<YOUR_LICENSE_ID>\"$export DOCKER_CONFIG=$(mktemp -d)$cat <<EOF > \"${DOCKER_CONFIG}/config.json\" { \"auths\": { \"proxy.replicated.com\": {\"auth\": \"$(echo -n \"${LICENSE_ID}:${LICENSE_ID}\" | base64 | tr -d '\\n')\"}}EOF\n```\n\nExample:\n```text\n$sudo docker pull $CUSTOMER_TAG\n```\n\nExample:\n```text\n$sudo chmod 666 /var/run/docker.sock\n```\n\nExample:\n```text\n$sudo docker images\n```\n\nExample:\n```text\n$sudo docker run -d --rm --name embed-english --gpus=1 --net=host proxy.replicated.com/proxy/cohere/us-docker.pkg.dev/cohere-artifacts/replicated/single-serving-embed-multilingual-03:<YOUR_MODEL_TAG>$$sudo docker ps\n```\n\nExample:\n```text\n$curl --header \"Content-Type: application/json\" --request POST http://localhost:8080/embed --data-raw '{\"texts\": [\"testing multilingual embeddings\"], \"input_type\": \"classification\"}'\n```\n\nExample:\n```text\n$export CUSTOMER_TAG=proxy.replicated.com/proxy/cohere/us-docker.pkg.dev/cohere-artifacts/replicated/single-serving-embed-multilingual-03:<YOUR_MODEL_TAG>$export LICENSE_ID=\"<YOUR_LICENSE_ID>\"$export DOCKER_CONFIG=$(mktemp -d)$cat <<EOF > \"${DOCKER_CONFIG}/config.json\" { \"auths\": { \"proxy.replicated.com\": {\"auth\": \"$(echo -n \"${LICENSE_ID}:${LICENSE_ID}\" | base64 | tr -d '\\n')\"}}EOF>kubectl create secret generic cohere-pull-secret --from-file=.dockerconfigjson=\"{$DOCKER_CONFIG}/config.json\" --type=kubernetes.io/dockerconfigjson\n```\n\nExample:\n```text\n$kubectl apply -f cohere.yaml$kubectl get pods$kubectl logs -f <pod-name-from-above-command>\n```\n\nExample:\n```text\n$kubectl port-forward svc/cohere 8080:8080\n```\n\nExample:\n```text\n$curl --header \"Content-Type: application/json\" --request POST http://localhost:8080/embed --data-raw '{\"texts\": [\"testing embeddings in english\"], \"input_type\": \"classification\"}'\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.319Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":776}}126{"id":"doc-deploying_models_in_private_environments_cohere-7324345d","source":"documentation","title":"Deploying Models in Private Environments | Cohere","url":"https://docs.cohere.com/docs/single-container-on-private-clouds","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\nLICENSE_ID=\"<YOUR LICENSE ID>\"cat <<EOF > ~/.docker/config.json { \"auths\": { \"proxy.replicated.com\": { \"auth\": \"$(echo -n \"${LICENSE_ID}:${LICENSE_ID}\" | base64 | tr -d '\\n')\" } }}EOF\n```\n\nExample:\n```text\nLICENSE_ID=\"<YOUR LICENSE ID>\"export DOCKER_CONFIG=$(mktemp -d)cat <<EOF > \"${DOCKER_CONFIG}/config.json\"{ \"auths\": { \"proxy.replicated.com\": { \"auth\": \"$(echo -n \"${LICENSE_ID}:${LICENSE_ID}\" | base64 | tr -d '\\n')\" } }}EOF\n```\n\nExample:\n```text\nCUSTOMER_TAG=image_tag_from_cohere # provided by Coheredocker pull $CUSTOMER_TAG\n```\n\nExample:\n```text\ndocker run -d --rm --name embed-v4 --gpus=1 --net=host $IMAGE_TAG# wait 5-10 seconds for the container to start# you can use `curl http://localhost:8080/ping` to check for readinesscurl --header \"Content-Type: application/json\" --request POST http://localhost:8080/embed --data-raw '{\"input_type\": \"search_query\", \"texts\":[\"Why are embeddings good\"], \"embedding_types\": [\"float\"]}'{\"id\":\"6d54d453-f2c8-44da-aab8-39e3c11d29d5\",\"texts\":[\"Why are embeddings good\"],\"embeddings\":{\"float\":[[0.033935547,0.06347656,0.020263672,-0.020507812,0.014160156,0.0038757324,-0.07421875,-0.05859375,...docker stop embed-v4\n```\n\nExample:\n```text\n1kubectl create secret generic cohere-pull-secret \\2 --from-file=.dockerconfigjson=\"~/.docker/config.json\" \\3 --type=kubernetes.io/dockerconfigjson\n```\n\nExample:\n```text\nAPP=cohere # or any other name you want to useIMAGE= <IMAGE_TAG_FROM_COHERE> # replace with the image cohere providedGPUS= <Number of GPUs for the target model> cat <<EOF > cohere.yaml---apiVersion: apps/v1kind: Deploymentmetadata: labels: app: ${APP} name: ${APP}spec: replicas: 1 selector: matchLabels: app: ${APP} strategy: {} template: metadata: labels: app: ${APP} spec: imagePullSecrets: - name: cohere-pull-secret containers: - image: ${IMAGE} name: ${APP} resources: limits: nvidia.com/gpu: ${GPUS}---apiVersion: v1kind: Servicemetadata: labels: app: ${APP} name: ${APP}spec: ports: - name: http port: 8080 protocol: TCP targetPort: 8080 selector: app: ${APP} type: ClusterIP---EOF\n```\n\nExample:\n```text\nkubectl apply -f cohere.yaml\n```\n\nExample:\n```text\n# once the pod is runningkubectl port-forward svc/${APP} 8080:8080# Forwarding from 127.0.0.1:8080 -> 8080# Forwarding from [::1]:8080 -> 8080# Handling connection for 8080\n```\n\nExample:\n```text\ncurl --header \"Content-Type: application/json\" --request POST http://localhost:8080/embed --data-raw '{\"texts\": [\"testing embeddings in english\"], \"input_type\": \"classification\"}'# {\"id\":\"2ffe4bca-8664-4456-b858-1b3b15411f2c\",\"embeddings\":[[-0.5019531,-2.0917969,-1.6220703,-1.2919922,-0.80029297,1.3173828,1.4677734,-1.7763672,0.03869629,1.9033203...}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.320Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":9,"totalLines":48,"estimatedTokens":763}}127{"id":"doc-deploy_finetuned_command_models_from_aws_marketp-973630c8","source":"documentation","title":"Deploy Finetuned Command Models from AWS Marketplace | Cohere","url":"https://docs.cohere.com/docs/bring-your-finetuned-models-to-sagemaker","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\npip install \"cohere>=5.11.0\"\n```\n\nExample:\n```text\n1import cohere2import os3import sagemaker as sage45from sagemaker.s3 import S3Uploader\n```\n\nExample:\n```text\n1# Change \"<aws_profile>\" to your own AWS profile name2os.environ[\"AWS_PROFILE\"] = \"<aws_profile>\"\n```\n\nExample:\n```text\n1# The AWS region2region = \"<region>\"34# Get the arn of the bring your own finetuning algorithm by region5cohere_package = \"cohere-command-r-v2-byoft-8370167e649c32a1a5f00267cd334c2c\"6algorithm_map = {7 \"us-east-1\": f\"arn:aws:sagemaker:us-east-1:865070037744:algorithm/{cohere_package}\",8 \"us-east-2\": f\"arn:aws:sagemaker:us-east-2:057799348421:algorithm/{cohere_package}\",9 \"us-west-2\": f\"arn:aws:sagemaker:us-west-2:594846645681:algorithm/{cohere_package}\",10 \"eu-central-1\": f\"arn:aws:sagemaker:eu-central-1:446921602837:algorithm/{cohere_package}\",11 \"ap-southeast-1\": f\"arn:aws:sagemaker:ap-southeast-1:192199979996:algorithm/{cohere_package}\",12 \"ap-southeast-2\": f\"arn:aws:sagemaker:ap-southeast-2:666831318237:algorithm/{cohere_package}\",13 \"ap-northeast-1\": f\"arn:aws:sagemaker:ap-northeast-1:977537786026:algorithm/{cohere_package}\",14 \"ap-south-1\": f\"arn:aws:sagemaker:ap-south-1:077584701553:algorithm/{cohere_package}\",15}16if region not in algorithm_map:17 raise Exception(f\"Current region {region} is not supported.\")18arn = algorithm_map[region]1920# The local directory of your adapter weights. No need to specify this, if you bring your own merged weights21adapter_weights_dir = \"<adapter_weights_dir>\"2223# The local directory you want to save the merged weights. Or the local directory of your own merged weights, if you bring your own merged weights24merged_weights_dir = \"<merged_weights_dir>\"2526# The S3 directory you want to save the merged weights27s3_checkpoint_dir = \"<s3_checkpoint_dir>\"2829# The S3 directory you want to save the exported TensorRT-LLM engine. Make sure you do not reuse the same S3 directory across multiple runs30s3_output_dir = \"<s3_output_dir>\"3132# The name of the export33export_name = \"<export_name>\"3435# The name of the SageMaker endpoint36endpoint_name = \"<endpoint_name>\"3738# The instance type for export and inference. Now \"ml.p4de.24xlarge\" and \"ml.p5.48xlarge\" are supported39instance_type = \"<instance_type>\"\n```\n\nExample:\n```text\n1import torch23from peft import PeftModel4from transformers import CohereForCausalLM567def load_and_merge_model(base_model_name_or_path: str, adapter_weights_dir: str):8 \"\"\"9 Load the base model and the model finetuned by PEFT, and merge the adapter weights to the base weights to get a model with merged weights10 \"\"\"11 base_model = CohereForCausalLM.from_pretrained(base_model_name_or_path)12 peft_model = PeftModel.from_pretrained(base_model, adapter_weights_dir)13 merged_model = peft_model.merge_and_unload()14 return merged_model151617def save_hf_model(output_dir: str, model, tokenizer=None, args=None):18 \"\"\"19 Save a HuggingFace model (and optionally tokenizer as well as additional args) to a local directory20 \"\"\"21 os.makedirs(output_dir, exist_ok=True)22 model.save_pretrained(output_dir, state_dict=None, safe_serialization=True)23 if tokenizer is not None:24 tokenizer.save_pretrained(output_dir)25 if args is not None:26 torch.save(args, os.path.join(output_dir, \"training_args.bin\"))2728# Get the merged model from adapter weights29merged_model = load_and_merge_model(\"CohereForAI/c4ai-command-r-08-2024\", adapter_weights_dir)3031# Save the merged weights to your local directory32save_hf_model(merged_weights_dir, merged_model)\n```\n\nExample:\n```text\n1sess = sage.Session()2merged_weights = S3Uploader.upload(merged_weights_dir, s3_checkpoint_dir, sagemaker_session=sess)3print(\"merged_weights\", merged_weights)\n```\n\nExample:\n```text\n1co = cohere.SagemakerClient(aws_region=region)2co.sagemaker_finetuning.export_finetune(3 arn=arn,4 name=export_name,5 s3_checkpoint_dir=s3_checkpoint_dir,6 s3_output_dir=s3_output_dir,7 instance_type=instance_type,8 role=\"ServiceRoleSagemaker\",9)\n```\n\nExample:\n```text\n1co.sagemaker_finetuning.create_endpoint(2 arn=arn,3 endpoint_name=endpoint_name,4 s3_models_dir=s3_output_dir,5 recreate=True,6 instance_type=instance_type,7 role=\"ServiceRoleSagemaker\",8)\n```\n\nExample:\n```text\n1# If the endpoint is already deployed, you can directly connect to it2co.sagemaker_finetuning.connect_to_endpoint(endpoint_name=endpoint_name)34message = \"Classify the following text as either very negative, negative, neutral, positive or very positive: mr. deeds is , as comedy goes , very silly -- and in the best way.\"5result = co.sagemaker_finetuning.chat(message=message)6print(result)\n```\n\nExample:\n```text\n1import json2from tqdm import tqdm34eval_data_path = \"<path_to_scienceQA_eval.jsonl>\"56total = 07correct = 08for line in tqdm(open(eval_data_path).readlines()):9 total += 110 question_answer_json = json.loads(line)11 question = question_answer_json[\"messages\"][0][\"content\"]12 answer = question_answer_json[\"messages\"][1][\"content\"]13 model_ans = co.sagemaker_finetuning.chat(message=question, temperature=0).text14 if model_ans == answer:15 correct += 11617print(f\"Accuracy of finetuned model is %.3f\" % (correct / total))\n```\n\nExample:\n```text\n1co.sagemaker_finetuning.delete_endpoint()2co.sagemaker_finetuning.close()\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.321Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":11,"totalLines":58,"estimatedTokens":1398}}128{"id":"doc-cohere_text_generation_tutorial_cohere-9de54836","source":"documentation","title":"Cohere Text Generation Tutorial | Cohere","url":"https://docs.cohere.com/docs/text-generation-tutorial","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# pip install cohere23import cohere4import json56# Get your free API key: https://dashboard.cohere.com/api-keys7co = cohere.ClientV2(api_key=\"COHERE_API_KEY\")\n```\n\nExample:\n```text\n1# Add the user message2message = \"I'm joining a new startup called Co1t today. Could you help me write a short introduction message to my teammates.\"34# Generate the response5response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[{\"role\": \"user\", \"content\": message}],8)9# messages=[cohere.UserMessage(content=message)])1011print(response.message.content[0].text)\n```\n\nExample:\n```text\nSure! Here is a draft of an introduction message: \"Hi everyone! My name is [Your Name], and I am thrilled to be joining the Co1t team today. I am excited to get to know you all and contribute to the amazing work being done at this startup. A little about me: [Brief description of your role, experience, and interests]. Outside of work, I enjoy [Hobbies and interests]. I look forward to collaborating with you all and being a part of Co1t's journey. Let's connect and make something great together!\" Feel free to edit and personalize the message to your liking. Good luck with your new role at Co1t!\n```\n\nExample:\n```text\n1# Add the user message2message = \"I'm joining a new startup called Co1t today. Could you help me write a one-sentence introduction message to my teammates.\"34# Generate the response5response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[{\"role\": \"user\", \"content\": message}],8)9# messages=[cohere.UserMessage(content=message)])1011print(response.message.content[0].text)\n```\n\nExample:\n```text\n\"Hi everyone, my name is [Your Name], and I am thrilled to join the Co1t team today as a [Your Role], eager to contribute my skills and ideas to the company's growth and success!\"\n```\n\nExample:\n```text\n1# Add the user message2user_input = (3 \"Why can't I access the server? Is it a permissions issue?\"4)56# Create a prompt containing example outputs7message = f\"\"\"Write a ticket title for the following user request:89User request: Where are the usual storage places for project files?10Ticket title: Project File Storage Location1112User request: Emails won't send. What could be the issue?13Ticket title: Email Sending Issues1415User request: How can I set up a connection to the office printer?16Ticket title: Printer Connection Setup1718User request: {user_input}19Ticket title:\"\"\"2021# Generate the response22response = co.chat(23 model=\"command-a-plus-05-2026\",24 messages=[{\"role\": \"user\", \"content\": message}],25)2627print(response.message.content[0].text)\n```\n\nExample:\n```text\nTicket title: \"Server Access Permissions Issue\"\n```\n\nExample:\n```text\n1# Add the user message2message = \"I'm joining a new startup called Co1t today. Could you help me write a one-sentence introduction message to my teammates.\"34# Generate the response5response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[{\"role\": \"user\", \"content\": message}],8)910print(response.message.content[0].text)\n```\n\nExample:\n```text\n\"Hi, I'm [Your Name] and I'm thrilled to join the Co1t team today as a [Your Role], eager to contribute my skills and ideas to help drive innovation and success for our startup!\"\n```\n\nExample:\n```text\n1# Add the user message2message = \"I like learning about the industrial revolution and how it shapes the modern world. How I can introduce myself in five words or less.\"34# Generate the response multiple times by specifying a low temperature value5for idx in range(3):6 response = co.chat(7 model=\"command-a-plus-05-2026\",8 messages=[{\"role\": \"user\", \"content\": message}],9 temperature=0,10 )1112 print(f\"{idx+1}: {response.message.content[0].text}\\n\")\n```\n\nExample:\n```text\n1: \"Revolution Enthusiast\"2: \"Revolution Enthusiast\"3: \"Revolution Enthusiast\"\n```\n\nExample:\n```text\n1# Add the user message2message = \"I like learning about the industrial revolution and how it shapes the modern world. How I can introduce myself in five words or less.\"34# Generate the response multiple times by specifying a low temperature value5for idx in range(3):6 response = co.chat(7 model=\"command-a-plus-05-2026\",8 messages=[{\"role\": \"user\", \"content\": message}],9 temperature=1,10 )1112 print(f\"{idx+1}: {response.message.content[0].text}\\n\")\n```\n\nExample:\n```text\n1: Here is a suggestion: \"Revolution Enthusiast. History Fan.\" This introduction highlights your passion for the industrial revolution and its impact on history while keeping within the word limit.2: \"Revolution fan.\"3: \"IR enthusiast.\"\n```\n\nExample:\n```text\n1# Add the user message2user_input = (3 \"Why can't I access the server? Is it a permissions issue?\"4)5message = f\"\"\"Create an IT ticket for the following user request. Generate a JSON object.6{user_input}\"\"\"78# Generate the response multiple times by adding the JSON schema9response = co.chat(10 model=\"command-a-plus-05-2026\",11 messages=[{\"role\": \"user\", \"content\": message}],12 response_format={13 \"type\": \"json_object\",14 \"schema\": {15 \"type\": \"object\",16 \"required\": [\"title\", \"category\", \"status\"],17 \"properties\": {18 \"title\": {\"type\": \"string\"},19 \"category\": {20 \"type\": \"string\",21 \"enum\": [\"access\", \"software\"],22 },23 \"status\": {24 \"type\": \"string\",25 \"enum\": [\"open\", \"closed\"],26 },27 },28 },29 },30)3132json_object = json.loads(response.message.content[0].text)3334print(json_object)\n```\n\nExample:\n```text\n{'title': 'Unable to Access Server', 'category': 'access', 'status': 'open'}\n```\n\nExample:\n```text\n1# Add the user message2message = \"I'm joining a new startup called Co1t today. Could you help me write a one-sentence introduction message to my teammates.\"34# Generate the response by streaming it5response = co.chat_stream(6 model=\"command-a-plus-05-2026\",7 messages=[{\"role\": \"user\", \"content\": message}],8)910for event in response:11 if event:12 if event.type == \"content-delta\":13 print(event.delta.message.content.text, end=\"\")\n```\n\nExample:\n```text\n\"Hi, I'm [Your Name] and I'm thrilled to join the Co1t team today as a [Your Role], passionate about [Your Expertise], and excited to contribute to our shared mission of [Startup's Mission]!\"\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.322Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":17,"totalLines":88,"estimatedTokens":1662}}129{"id":"doc-cohere_on_the_microsoft_azure_platform_cohere-6af184e6","source":"documentation","title":"Cohere on the Microsoft Azure Platform | Cohere","url":"https://docs.cohere.com/docs/cohere-on-microsoft-azure","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import urllib.request2import json34# Configure payload data sending to API endpoint5data = {6 \"messages\": [7 {\"role\": \"system\", \"content\": \"You are a helpful assistant.\"},8 {\"role\": \"user\", \"content\": \"What is good about Wuhan?\"},9 ],10 \"max_tokens\": 500,11 \"temperature\": 0.3,12 \"stream\": \"True\",13}1415body = str.encode(json.dumps(data))1617# Replace the url with your API endpoint18url = (19 \"https://your-endpoint.inference.ai.azure.com/v1/chat/completions\"20)2122# Replace this with the key for the endpoint23api_key = \"your-auth-key\"24if not api_key:25 raise Exception(\"API Key is missing\")2627headers = {28 \"Content-Type\": \"application/json\",29 \"Authorization\": (api_key),30}3132req = urllib.request.Request(url, body, headers)3334try:35 response = urllib.request.urlopen(req)36 result = response.read()37 print(result)38except urllib.error.HTTPError as error:39 print(\"The request failed with status code: \" + str(error.code))40 # Print the headers - they include the requert ID and the timestamp, which are useful for debugging the failure41 print(error.info())42 print(error.read().decode(\"utf8\", \"ignore\"))\n```\n\nExample:\n```text\n1import urllib.request2import json34# Configure payload data sending to API endpoint5data = {\"input\": [\"hi\"]}67body = str.encode(json.dumps(data))89# Replace the url with your API endpoint10url = \"https://your-endpoint.inference.ai.azure.com/v1/embedding\"1112# Replace this with the key for the endpoint13api_key = \"your-auth-key\"14if not api_key:15 raise Exception(\"API Key is missing\")1617headers = {18 \"Content-Type\": \"application/json\",19 \"Authorization\": (api_key),20}2122req = urllib.request.Request(url, body, headers)2324try:25 response = urllib.request.urlopen(req)26 result = response.read()27 print(result)28except urllib.error.HTTPError as error:29 print(\"The request failed with status code: \" + str(error.code))30 # Print the headers - they include the requert ID and the timestamp, which are useful for debugging the failure31 print(error.info())32 print(error.read().decode(\"utf8\", \"ignore\"))\n```\n\nExample:\n```text\n1import cohere23co = cohere.Client(4 base_url=\"https://<endpoint>.<region>.inference.ai.azure.com/v1/rerank\",5 api_key=\"<key>\",6)78documents = [9 {10 \"Title\": \"Incorrect Password\",11 \"Content\": \"Hello, I have been trying to access my account for the past hour and it keeps saying my password is incorrect. Can you please help me?\",12 },13 {14 \"Title\": \"Confirmation Email Missed\",15 \"Content\": \"Hi, I recently purchased a product from your website but I never received a confirmation email. Can you please look into this for me?\",16 },17 {18 \"Title\": \"Questions about Return Policy\",19 \"Content\": \"Hello, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective.\",20 },21 {22 \"Title\": \"Customer Support is Busy\",23 \"Content\": \"Good morning, I have been trying to reach your customer support team for the past week but I keep getting a busy signal. Can you please help me?\",24 },25 {26 \"Title\": \"Received Wrong Item\",27 \"Content\": \"Hi, I have a question about my recent order. I received the wrong item and I need to return it.\",28 },29 {30 \"Title\": \"Customer Service is Unavailable\",31 \"Content\": \"Hello, I have been trying to reach your customer support team for the past hour but I keep getting a busy signal. Can you please help me?\",32 },33 {34 \"Title\": \"Return Policy for Defective Product\",35 \"Content\": \"Hi, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective.\",36 },37 {38 \"Title\": \"Wrong Item Received\",39 \"Content\": \"Good morning, I have a question about my recent order. I received the wrong item and I need to return it.\",40 },41 {42 \"Title\": \"Return Defective Product\",43 \"Content\": \"Hello, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective.\",44 },45]4647response = co.rerank(48 documents=documents,49 query=\"What emails have been about returning items?\",50 model=\"rerank-v4.0-pro\",51 rank_fields=[\"Title\", \"Content\"],52 top_n=5,53)\n```\n\nExample:\n```text\n1# pip install cohere23import cohere45# For Command models6co_chat = cohere.Client(7 api_key=\"AZURE_INFERENCE_CREDENTIAL\",8 base_url=\"AZURE_MODEL_ENDPOINT\", # Example - https://Cohere-command-r-plus-08-2024-xyz.eastus.models.ai.azure.com/9)1011# For Embed models12co_embed = cohere.Client(13 api_key=\"AZURE_INFERENCE_CREDENTIAL\",14 base_url=\"AZURE_MODEL_ENDPOINT\", # Example - https://cohere-embed-v4-xyz.eastus.models.ai.azure.com/15)1617# For Rerank models18co_rerank = cohere.Client(19 api_key=\"AZURE_INFERENCE_CREDENTIAL\",20 base_url=\"AZURE_MODEL_ENDPOINT\", # Example - https://cohere-rerank-v4-pro-xyz.eastus.models.ai.azure.com/21)\n```\n\nExample:\n```text\n1message = \"I'm joining a new startup called Co1t today. Could you help me write a short introduction message to my teammates.\"23response = co_chat.chat(message=message)45print(response)\n```\n\nExample:\n```text\n1faqs_short = [2 {3 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"4 },5 {6 \"text\": \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\"7 },8]910query = \"Are there fitness-related perks?\"1112response = co_chat.chat(message=query, documents=faqs_short)1314print(response)\n```\n\nExample:\n```text\n1docs = [2 \"Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.\",3 \"Finding Coffee Spots: For your caffeine fix, head to the break room's coffee machine or cross the street to the café for artisan coffee.\",4]56doc_emb = co_embed.embed(7 input_type=\"search_document\",8 texts=docs,9).embeddings\n```\n\nExample:\n```text\n1faqs_short = [2 {3 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"4 },5 {6 \"text\": \"Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.\"7 },8 {9 \"text\": \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\"10 },11]1213query = \"Are there fitness-related perks?\"1415results = co_rerank.rerank(16 query=query,17 documents=faqs_short,18 top_n=2,19 model=\"rerank-v4.0-pro\",20)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.322Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":8,"totalLines":43,"estimatedTokens":1791}}130{"id":"doc-building_a_chatbot_with_cohere_cohere-ee5f4d87","source":"documentation","title":"Building a Chatbot with Cohere | Cohere","url":"https://docs.cohere.com/docs/building-a-chatbot-with-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# pip install cohere23import cohere45# Get your free API key: https://dashboard.cohere.com/api-keys6co = cohere.ClientV2(api_key=\"COHERE_API_KEY\")\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\",3 messages=[4 {5 \"role\": \"user\",6 \"content\": \"I'm joining a new startup called Co1t today. Could you help me write a short introduction message to my teammates.\",7 },8 ],9)1011print(response.message)\n```\n\nExample:\n```text\n1{2 role='assistant',3 content=[4 {5 type='text', 6 text='Absolutely! Here’s a warm and professional introduction message you can use to connect with your new teammates at Co1t:\\n\\n---\\n\\n**Subject:** Excited to Join the Co1t Team! \\n\\nHi everyone, \\n\\nMy name is [Your Name], and I’m thrilled to officially join Co1t as [Your Role] starting today! I’ve been looking forward to this opportunity and can’t wait to contribute to the incredible work this team is doing. \\n\\nA little about me: [Share a brief personal or professional detail, e.g., \"I’ve spent the last few years working in [industry/field], and I’m passionate about [specific skill or interest].\" or \"Outside of work, I love [hobby or interest] and am always up for a good [book/podcast/movie] recommendation!\"] \\n\\nI’m excited to get to know each of you, learn from your experiences, and collaborate on driving Co1t’s mission forward. Please feel free to reach out—I’d love to chat and hear more about your roles and what you’re working on. \\n\\nLooking forward to an amazing journey together! \\n\\nBest regards, \\n[Your Name] \\n[Your Role] \\nCo1t \\n\\n---\\n\\nFeel free to customize it further to match your style and the culture of Co1t. Good luck on your first day! 🚀'7 }8 ],9}\n```\n\nExample:\n```text\n1# Create a custom system instruction that guides all of the Assistant's responses2system_instruction = \"\"\"## Task and Context3You assist new employees of Co1t with their first week of onboarding at Co1t, a startup founded by Mr. Colt.4If the user asks any questions unrelated to onboarding, politely refuse to answer them.56## Style Guide7Try to speak in rhymes as much as possible. Be professional.\"\"\"89# Send messages to the model10response = co.chat(11 model=\"command-a-plus-05-2026\",12 messages=[13 {\"role\": \"system\", \"content\": system_instruction},14 {15 \"role\": \"user\",16 \"content\": \"I'm joining a new startup called Co1t today. Could you help me write a short introduction message to my teammates.\",17 },18 ],19)2021print(response.message.content[0].text)\n```\n\nExample:\n```text\nSure, here's a rhyme to break the ice,A warm welcome to the team, so nice,Hi, I'm [Your Name], a new face,Ready to join the Co1t space,A journey begins, a path unknown,But together we'll make our mark, a foundation stone,Excited to learn and contribute my part,Let's create, innovate, and leave a lasting art,Looking forward to our adventures yet untold,With teamwork and passion, let's achieve our goals!Cheers to a great start!Your enthusiastic new mate.\n```\n\nExample:\n```text\n1messages = [2 {\"role\": \"system\", \"content\": system_instruction},3]45# user turn 16messages.append(7 {8 \"role\": \"user\",9 \"content\": \"I'm joining a new startup called Co1t today. Could you help me write a short introduction message to my teammates.\",10 },11)12response = co.chat(13 model=\"command-a-plus-05-2026\",14 messages=messages,15)1617# assistant turn 118messages.append(19 response.message20) # add the Assistant message to the messages array to include it in the chat history for the next turn2122# user turn 223messages.append({\"role\": \"user\", \"content\": \"Who founded co1t?\"})2425response = co.chat(26 model=\"command-a-plus-05-2026\",27 messages=messages,28)2930# assistant turn 231messages.append(response.message)3233print(response.message.content[0].text)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.323Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":6,"totalLines":33,"estimatedTokens":1034}}131{"id":"doc-semantic_search_cohere_on_azure_ai_foundry_coher-6eb8e7cd","source":"documentation","title":"Semantic search - Cohere on Azure AI Foundry | Cohere","url":"https://docs.cohere.com/docs/cohere-on-azure/azure-ai-sem-search","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# %pip install cohere hnswlib23import pandas as pd4import hnswlib5import re6import cohere78co = cohere.ClientV2(9 api_key=\"AZURE_API_KEY_EMBED\",10 base_url=\"AZURE_ENDPOINT_EMBED\", # example: \"https://embed-v-4-0-xyz.eastus.models.ai.azure.com/\"11)\n```\n\nExample:\n```text\n1url = \"https://raw.githubusercontent.com/cohere-ai/cohere-aws/main/notebooks/bedrock/multiFIN_train.csv\"2df = pd.read_csv(url)34# Inspect dataset5df.head(5)\n```\n\nExample:\n```text\n1# Ensure there is no duplicated text in the headers2def remove_duplicates(text):3 return re.sub(4 r\"((\\b\\w+\\b.{1,2}\\w+\\b)+).+\\1\", r\"\\1\", text, flags=re.I5 )678df[\"text\"] = df[\"text\"].apply(remove_duplicates)910# Keep only selected languages11languages = [\"English\", \"Spanish\", \"Danish\"]12df = df.loc[df[\"lang\"].isin(languages)]1314# Pick the top 80 longest articles15df[\"text_length\"] = df[\"text\"].str.len()16df.sort_values(by=[\"text_length\"], ascending=False, inplace=True)17top_80_df = df[:80]1819# Language distribution20top_80_df[\"lang\"].value_counts()\n```\n\nExample:\n```text\n1lang2Spanish 333English 294Danish 185Name: count, dtype: int64\n```\n\nExample:\n```text\n1# Embed documents2# Embed documents3docs = top_80_df[\"text\"].to_list()4docs_lang = top_80_df[\"lang\"].to_list()5translated_docs = top_80_df[6 \"translation\"7].to_list() # for reference when returning non-English results8doc_embs = co.embed(9 model=\"embed-v4.0\",10 texts=docs,11 input_type=\"search_document\",12 embedding_types=[\"float\"],13).embeddings.float1415# Create a search index16index = hnswlib.Index(space=\"ip\", dim=1536)17index.init_index(18 max_elements=len(doc_embs), ef_construction=512, M=6419)20index.add_items(doc_embs, list(range(len(doc_embs))))\n```\n\nExample:\n```text\n1# Retrieval of 4 closest docs to query2def retrieval(query):3 # Embed query and retrieve results4 query_emb = co.embed(5 model=\"embed-v4.0\", # Pass a dummy string6 texts=[query],7 input_type=\"search_query\",8 embedding_types=[\"float\"],9 ).embeddings.float1011 doc_ids = index.knn_query(query_emb, k=3)[0][12 013 ] # we will retrieve 3 closest neighbors1415 # Print and append results16 print(f\"QUERY: {query.upper()} \\n\")17 retrieved_docs, translated_retrieved_docs = [], []1819 for doc_id in doc_ids:20 # Append results21 retrieved_docs.append(docs[doc_id])22 translated_retrieved_docs.append(translated_docs[doc_id])2324 # Print results25 print(f\"ORIGINAL ({docs_lang[doc_id]}): {docs[doc_id]}\")26 if docs_lang[doc_id] != \"English\":27 print(f\"TRANSLATION: {translated_docs[doc_id]} \\n----\")28 else:29 print(\"----\")30 print(\"END OF RESULTS \\n\\n\")31 return retrieved_docs, translated_retrieved_docs\n```\n\nExample:\n```text\n1queries = [2 \"Can data science help meet sustainability goals?\", # English example3 \"Hvor kan jeg finde den seneste danske boligplan?\", # Danish example - \"Where can I find the latest Danish property plan?\"4]56for query in queries:7 retrieval(query)\n```\n\nExample:\n```text\n1QUERY: CAN DATA SCIENCE HELP MEET SUSTAINABILITY GOALS? 23ORIGINAL (English): Using AI to better manage the environment could reduce greenhouse gas emissions, boost global GDP by up to 38m jobs by 20304----5ORIGINAL (English): Quality of business reporting on the Sustainable Development Goals improves, but has a long way to go to meet and drive targets.6----7ORIGINAL (English): Only 10 years to achieve Sustainable Development Goals but businesses remain on starting blocks for integration and progress8----9END OF RESULTS 101112QUERY: HVOR KAN JEG FINDE DEN SENESTE DANSKE BOLIGPLAN? 1314ORIGINAL (Danish): Nyt fra CFOdirect: Ny PP&E-guide, FAQs om den nye leasingstandard, podcast om udfordringerne ved implementering af leasingstandarden og meget mere15TRANSLATION: New from CFOdirect: New PP&E guide, FAQs on the new leasing standard, podcast on the challenges of implementing the leasing standard and much more 16----17ORIGINAL (Danish): Lovforslag fremlagt om rentefri lån, udskudt frist for lønsumsafgift, førtidig udbetaling af skattekredit og loft på indestående på skattekontoen18TRANSLATION: Bills presented on interest -free loans, deferred deadline for payroll tax, early payment of tax credit and ceiling on the balance in the tax account 19----20ORIGINAL (Danish): Nyt fra CFOdirect: Shareholder-spørgsmål til ledelsen, SEC cybersikkerhedsguide, den amerikanske skattereform og meget mere21TRANSLATION: New from CFOdirect: Shareholder questions for management, the SEC cybersecurity guide, US tax reform and more 22----23END OF RESULTS\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.323Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":8,"totalLines":43,"estimatedTokens":1214}}132{"id":"doc-cohere_chat_on_langchain_integration_guide_coher-71a998d5","source":"documentation","title":"Cohere Chat on LangChain (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/chat-on-langchain","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from langchain_cohere import ChatCohere2from langchain_core.messages import AIMessage, HumanMessage34# Define the Cohere LLM5llm = ChatCohere(6 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"7)89# Send a chat message without chat history10current_message = [HumanMessage(content=\"knock knock\")]11print(llm.invoke(current_message))1213# Send a chat message with chat history, note the last message is the current user message14current_message_and_history = [15 HumanMessage(content=\"knock knock\"),16 AIMessage(content=\"Who's there?\"),17 HumanMessage(content=\"Tank\"),18]19print(llm.invoke(current_message_and_history))\n```\n\nExample:\n```text\n1from langchain_cohere import ChatCohere2from langchain_core.messages import HumanMessage34# Define the Cohere LLM5llm = ChatCohere(6 cohere_api_key=\"COHERE_API_KEY\",7 model=\"command-a-reasoning-08-2025\",8)910response = llm.invoke(11 [12 HumanMessage(13 content=\"Alice has 3 brothers and 2 sisters. How many sisters does Alice's brother have?\"14 )15 ]16)1718# The reasoning model returns its content as a list of blocks: a \"reasoning\"19# block (the model's reasoning) followed by a \"text\" block (the final answer).20for block in response.content:21 if block[\"type\"] == \"reasoning\":22 for step in block[\"summary\"]:23 print(\"Reasoning:\", step[\"text\"])24 elif block[\"type\"] == \"text\":25 print(\"Answer:\", block[\"text\"])\n```\n\nExample:\n```text\n1from langchain_cohere import ChatCohere2from langchain_core.messages import HumanMessage34# Define the Cohere LLM5llm = ChatCohere(6 cohere_api_key=\"COHERE_API_KEY\",7 model=\"command-a-vision-07-2025\",8)910# Use a publicly accessible image URL (a base64 data URI also works)11image_url = \"https://raw.githubusercontent.com/cohere-ai/cohere-developer-experience/main/fern/assets/images/waste-management-request.png\"1213# Pass the image alongside a text prompt14message = HumanMessage(15 content=[16 {\"type\": \"text\", \"text\": \"What is shown in this image?\"},17 {\"type\": \"image_url\", \"image_url\": {\"url\": image_url}},18 ]19)2021print(llm.invoke([message]).content)\n```\n\nExample:\n```text\n1import os23from langchain.agents import create_agent4from langchain_cohere import ChatCohere5from langchain_tavily import TavilySearch67# Internet search tool. Replace the placeholder with your Tavily API key.8os.environ[\"TAVILY_API_KEY\"] = \"TAVILY_API_KEY\"9internet_search = TavilySearch()1011# Define the Cohere LLM12llm = ChatCohere(13 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"14)1516# Create an agent with the internet search tool17agent = create_agent(llm, tools=[internet_search])1819# Run the agent20result = agent.invoke(21 {\"messages\": [(\"user\", \"I want to write an essay. Any tips?\")]}22)2324# See Cohere's response25print(result[\"messages\"][-1].content)\n```\n\nExample:\n```text\n1import wikipedia23from langchain_cohere import ChatCohere4from langchain_community.retrievers import WikipediaRetriever56# Wikipedia requires a descriptive User-Agent; set one before querying.7wikipedia.set_user_agent(\"my-app/1.0 (you@example.com)\")89# User query we will use for the generation10user_query = \"What is Cohere?\"1112# Define the Cohere LLM13llm = ChatCohere(14 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"15)1617# Retrieve documents with any LangChain retriever18wiki_retriever = WikipediaRetriever()19wiki_docs = wiki_retriever.invoke(user_query)2021# Ground the answer in the retrieved documents22response = llm.invoke(user_query, documents=wiki_docs)2324# Print the answer25print(\"Answer:\")26print(response.content)27# Print the citations that ground the answer in the documents28print(\"Citations:\")29print(response.additional_kwargs.get(\"citations\"))\n```\n\nExample:\n```text\n1from langchain_cohere import ChatCohere2from langchain_core.documents import Document34# Define the Cohere LLM5llm = ChatCohere(6 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"7)89# Supply your own documents (these might come from elsewhere in your application)10documents = [11 Document(12 page_content=\"LangChain supports Cohere RAG!\",13 metadata={\"id\": \"id-1\"},14 ),15 Document(16 page_content=\"The sky is blue!\", metadata={\"id\": \"id-2\"}17 ),18]1920# Ground the answer in the documents21response = llm.invoke(22 \"Does LangChain support Cohere RAG?\", documents=documents23)2425# Print the answer26print(\"Answer:\")27print(response.content)28# Print the citations that ground the answer in the documents29print(\"Citations:\")30print(response.additional_kwargs.get(\"citations\"))\n```\n\nExample:\n```text\n1import os23from langchain_cohere import ChatCohere4from langchain_core.messages import HumanMessage, ToolMessage5from langchain_tavily import TavilySearch67# Web search tool. Replace the placeholder with your Tavily API key.8os.environ[\"TAVILY_API_KEY\"] = \"TAVILY_API_KEY\"9web_search = TavilySearch()1011# Define the Cohere LLM and bind the search tool12llm = ChatCohere(13 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"14)15llm_with_tools = llm.bind_tools([web_search])1617# 1. Force a search on the first turn so the answer is grounded18messages = [HumanMessage(\"Who founded Cohere?\")]19ai_message = llm_with_tools.invoke(messages, tool_choice=\"REQUIRED\")20messages.append(ai_message)2122# 2. Run the search and pass the results back to the model23for tool_call in ai_message.tool_calls:24 results = web_search.invoke(tool_call[\"args\"])25 messages.append(26 ToolMessage(27 content=str(results), tool_call_id=tool_call[\"id\"]28 )29 )3031# 3. The model answers, grounded in the search results32final = llm_with_tools.invoke(messages)33print(\"Answer:\", final.content)34# Cohere returns citations that ground the answer in the search results35# (`.get` avoids a KeyError if the answer came back without any citations)36print(\"Citations:\", final.additional_kwargs.get(\"citations\"))\n```\n\nExample:\n```text\n1from langchain_cohere import ChatCohere2from langchain_core.documents import Document3from langchain_core.prompts import ChatPromptTemplate4from langchain_classic.chains.combine_documents import (5 create_stuff_documents_chain,6)78prompt = ChatPromptTemplate.from_messages(9 [(\"human\", \"What are everyone's favorite colors:\\n\\n{context}\")]10)1112# Define the Cohere LLM13llm = ChatCohere(14 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"15)1617chain = create_stuff_documents_chain(llm, prompt)1819docs = [20 Document(page_content=\"Jesse loves red but not yellow\"),21 Document(22 page_content=\"Jamal loves green but not as much as he loves orange\"23 ),24]2526res = chain.invoke({\"context\": docs})27print(res)\n```\n\nExample:\n```text\n1from langchain_cohere import ChatCohere23# Define the Cohere LLM4llm = ChatCohere(5 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"6)78res = llm.invoke(9 \"John is five years old\",10 response_format={11 \"type\": \"json_object\",12 \"schema\": {13 \"title\": \"Person\",14 \"description\": \"Identifies the age and name of a person\",15 \"type\": \"object\",16 \"properties\": {17 \"name\": {18 \"type\": \"string\",19 \"description\": \"Name of the person\",20 },21 \"age\": {22 \"type\": \"number\",23 \"description\": \"Age of the person\",24 },25 },26 \"required\": [27 \"name\",28 \"age\",29 ],30 },31 },32)3334print(res)\n```\n\nExample:\n```text\n1from langchain_cohere import ChatCohere2from langchain_classic.chains.summarize import load_summarize_chain3from langchain_community.document_loaders import WebBaseLoader45loader = WebBaseLoader(\"https://docs.cohere.com/docs/cohere-toolkit\")6docs = loader.load()78# Define the Cohere LLM9llm = ChatCohere(10 cohere_api_key=\"COHERE_API_KEY\",11 model=\"command-a-03-2025\",12 temperature=0,13)1415chain = load_summarize_chain(llm, chain_type=\"stuff\")1617result = chain.invoke({\"input_documents\": docs})18print(result[\"output_text\"])\n```\n\nExample:\n```text\n1llm = ChatCohere(2 base_url=\"<YOUR_DEPLOYMENT_URL>\",3 cohere_api_key=\"COHERE_API_KEY\",4 model=\"MODEL_NAME\",5)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.324Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":11,"totalLines":58,"estimatedTokens":2136}}133{"id":"doc-cohere_tools_on_langchain_integration_guide_cohe-2a3ab698","source":"documentation","title":"Cohere Tools on LangChain (Integration Guide) | Cohere","url":"https://docs.cohere.com/docs/tools-on-langchain","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import os23from langchain.agents import create_agent4from langchain_cohere import ChatCohere5from langchain_tavily import TavilySearch67# Internet search tool. Replace the placeholder with your Tavily API key.8os.environ[\"TAVILY_API_KEY\"] = \"TAVILY_API_KEY\"910internet_search = TavilySearch()1112# Define the Cohere LLM13llm = ChatCohere(14 cohere_api_key=\"COHERE_API_KEY\",15 model=\"command-a-03-2025\",16 temperature=0,17)1819# System instruction for the agent20system_prompt = \"\"\"21You are an expert who answers the user's question by searching the internet for the most relevant, up-to-date information.22\"\"\"2324# Create a multi-step agent, passing the instruction via `system_prompt`25agent = create_agent(26 llm, tools=[internet_search], system_prompt=system_prompt27)2829# The agent can search multiple times to answer the question30result = agent.invoke(31 {32 \"messages\": [33 (\"user\", \"Who is the mayor of the capital of Ontario?\")34 ]35 }36)3738print(result[\"messages\"][-1].content)\n```\n\nExample:\n```text\n1from langchain_cohere import ChatCohere2from langchain_core.messages import HumanMessage, SystemMessage3from pydantic import BaseModel, Field456# Data model7class web_search(BaseModel):8 \"\"\"9 The internet. Use web_search for questions that are related to anything else than agents, prompt engineering, and adversarial attacks.10 \"\"\"1112 query: str = Field(13 description=\"The query to use when searching the internet.\"14 )151617class vectorstore(BaseModel):18 \"\"\"19 A vectorstore containing documents related to agents, prompt engineering, and adversarial attacks. Use the vectorstore for questions on these topics.20 \"\"\"2122 query: str = Field(23 description=\"The query to use when searching the vectorstore.\"24 )252627# System instruction that tells the model how to route28system_message = SystemMessage(29 content=\"\"\"You are an expert at routing a user question to a vectorstore or web search.30The vectorstore contains documents related to agents, prompt engineering, and adversarial attacks.31Use the vectorstore for questions on these topics. Otherwise, use web-search.\"\"\"32)3334# Define the Cohere LLM35llm = ChatCohere(36 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"37)3839# Bind the tools to the model40llm_with_tools = llm.bind_tools(tools=[web_search, vectorstore])4142# The model routes this question to web search43messages = [44 system_message,45 HumanMessage(\"Who will the Bears draft first in the NFL draft?\"),46]47response = llm_with_tools.invoke(messages)48print(response.tool_calls)4950# The model routes this question to the vectorstore51messages = [52 system_message,53 HumanMessage(\"What are the types of agent memory?\"),54]55response = llm_with_tools.invoke(messages)56print(response.tool_calls)5758# When no tool is needed, `.tool_calls` is an empty list59messages = [system_message, HumanMessage(\"Hi, how are you?\")]60response = llm_with_tools.invoke(messages)61print(response.tool_calls)\n```\n\nExample:\n```text\n1from langchain.agents import create_agent2from langchain_cohere import ChatCohere3from langchain_community.agent_toolkits import SQLDatabaseToolkit4from langchain_community.utilities import SQLDatabase5import urllib.request67# Download the Chinook SQLite database8url = \"https://github.com/lerocha/chinook-database/raw/master/ChinookDatabase/DataSources/Chinook_Sqlite.sqlite\"9urllib.request.urlretrieve(url, \"Chinook.db\")10print(\"Chinook database downloaded successfully.\")1112db = SQLDatabase.from_uri(\"sqlite:///Chinook.db\")13print(db.dialect)14print(db.get_usable_table_names())15db.run(\"SELECT * FROM Artist LIMIT 10;\")1617# Define the Cohere LLM18llm = ChatCohere(19 cohere_api_key=\"COHERE_API_KEY\",20 model=\"command-a-03-2025\",21 temperature=0,22)2324# Build a SQL agent from the database toolkit's tools25toolkit = SQLDatabaseToolkit(db=db, llm=llm)26agent_executor = create_agent(llm, tools=toolkit.get_tools())2728result = agent_executor.invoke(29 {30 \"messages\": [31 (\"user\", \"Show me the first 5 rows of the Album table.\")32 ]33 }34)35print(result[\"messages\"][-1].content)\n```\n\nExample:\n```text\n1from langchain.agents import create_agent2from langchain_cohere import ChatCohere3from langchain_experimental.tools import PythonAstREPLTool4import pandas as pd5import urllib.request67# Download the Titanic CSV and load it into a dataframe8url = \"https://raw.githubusercontent.com/pandas-dev/pandas/main/doc/data/titanic.csv\"9urllib.request.urlretrieve(url, \"titanic.csv\")10df = pd.read_csv(\"titanic.csv\")1112# Give the agent a Python REPL with the dataframe (`df`) in scope so it can13# answer arbitrary questions about the CSV by writing pandas code.14python_tool = PythonAstREPLTool(locals={\"df\": df})1516# Define the Cohere LLM17llm = ChatCohere(18 cohere_api_key=\"COHERE_API_KEY\",19 model=\"command-a-03-2025\",20 temperature=0,21)2223# Give the model the dataframe's columns and a preview so it knows the schema24# before it writes any pandas code.25system_prompt = (26 \"You are a data analyst working with a pandas dataframe named `df`.\\n\"27 f\"The dataframe columns are: {list(df.columns)}.\\n\"28 f\"Here is `df.head()`:\\n{df.head().to_string()}\\n\\n\"29 \"Answer the user's question by writing pandas code against `df` and running \"30 \"it with the Python tool, then report the result.\"31)3233agent_executor = create_agent(34 llm, tools=[python_tool], system_prompt=system_prompt35)3637result = agent_executor.invoke(38 {\"messages\": [(\"user\", \"How many people were on the titanic?\")]}39)40print(result[\"messages\"][-1].content)\n```\n\nExample:\n```text\n1from langchain_core.tools import tool2from langchain_cohere import ChatCohere345@tool6def add(a: int, b: int) -> int:7 \"\"\"Adds a and b.\"\"\"8 return a + b91011@tool12def multiply(a: int, b: int) -> int:13 \"\"\"Multiplies a and b.\"\"\"14 return a * b151617tools = [add, multiply]1819# Define the Cohere LLM20llm = ChatCohere(21 cohere_api_key=\"COHERE_API_KEY\",22 model=\"command-a-03-2025\",23 temperature=0,24)2526llm_with_tools = llm.bind_tools(tools)2728query = \"What is 3 * 12? Also, what is 11 + 49?\"2930for chunk in llm_with_tools.stream(query):31 if chunk.tool_call_chunks:32 print(chunk.tool_call_chunks)\n```\n\nExample:\n```text\n1from typing import Annotated2from typing_extensions import TypedDict3from langgraph.graph import StateGraph, START, END4from langgraph.graph.message import add_messages5from langchain_cohere import ChatCohere678# Create a state graph9class State(TypedDict):10 messages: Annotated[list, add_messages]111213graph_builder = StateGraph(State)1415# Define the Cohere LLM16llm = ChatCohere(17 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"18)192021# Add nodes22def chatbot(state: State):23 return {\"messages\": [llm.invoke(state[\"messages\"])]}242526graph_builder.add_node(\"chatbot\", chatbot)27graph_builder.add_edge(START, \"chatbot\")28graph_builder.add_edge(\"chatbot\", END)2930# Compile the graph31graph = graph_builder.compile()3233# Run the chatbot34while True:35 user_input = input(\"User: \")36 print(\"User: \" + user_input)37 if user_input.lower() in [\"quit\", \"exit\", \"q\"]:38 print(\"Goodbye!\")39 break40 for event in graph.stream({\"messages\": (\"user\", user_input)}):41 for value in event.values():42 print(\"Assistant:\", value[\"messages\"][-1].content)\n```\n\nExample:\n```text\n1from langchain_tavily import TavilySearch2from langchain_cohere import ChatCohere3from langgraph.graph import StateGraph, START4from langgraph.graph.message import add_messages5from langchain_core.messages import ToolMessage6from langchain_core.messages import BaseMessage7from typing import Annotated, Literal8from typing_extensions import TypedDict9import json1011# Create a tool12tool = TavilySearch(max_results=2)13tools = [tool]141516# Create a state graph17class State(TypedDict):18 messages: Annotated[list, add_messages]192021graph_builder = StateGraph(State)2223# Define the LLM24llm = ChatCohere(25 cohere_api_key=\"COHERE_API_KEY\", model=\"command-a-03-2025\"26)2728# Bind the tools to the LLM29llm_with_tools = llm.bind_tools(tools)303132# Add nodes33def chatbot(state: State):34 return {\"messages\": [llm_with_tools.invoke(state[\"messages\"])]}353637graph_builder.add_node(\"chatbot\", chatbot)383940class BasicToolNode:41 \"\"\"A node that runs the tools requested in the last AIMessage.\"\"\"4243 def __init__(self, tools: list) -> None:44 self.tools_by_name = {tool.name: tool for tool in tools}4546 def __call__(self, inputs: dict):47 if messages := inputs.get(\"messages\", []):48 message = messages[-1]49 else:50 raise ValueError(\"No message found in input\")51 outputs = []52 for tool_call in message.tool_calls:53 tool_result = self.tools_by_name[54 tool_call[\"name\"]55 ].invoke(tool_call[\"args\"])56 outputs.append(57 ToolMessage(58 content=json.dumps(tool_result),59 name=tool_call[\"name\"],60 tool_call_id=tool_call[\"id\"],61 )62 )63 return {\"messages\": outputs}646566tool_node = BasicToolNode(tools=[tool])67graph_builder.add_node(\"tools\", tool_node)686970def route_tools(71 state: State,72) -> Literal[\"tools\", \"__end__\"]:73 \"\"\"74 Use in the conditional_edge to route to the ToolNode if the last message75 has tool calls. Otherwise, route to the end.76 \"\"\"77 if isinstance(state, list):78 ai_message = state[-1]79 elif messages := state.get(\"messages\", []):80 ai_message = messages[-1]81 else:82 raise ValueError(83 f\"No messages found in input state to tool_edge: {state}\"84 )85 if (86 hasattr(ai_message, \"tool_calls\")87 and len(ai_message.tool_calls) > 088 ):89 return \"tools\"90 return \"__end__\"919293graph_builder.add_conditional_edges(94 \"chatbot\",95 route_tools,96 {\"tools\": \"tools\", \"__end__\": \"__end__\"},97)98graph_builder.add_edge(\"tools\", \"chatbot\")99graph_builder.add_edge(START, \"chatbot\")100101# Compile the graph102graph = graph_builder.compile()103104# Run the chatbot105while True:106 user_input = input(\"User: \")107 if user_input.lower() in [\"quit\", \"exit\", \"q\"]:108 print(\"Goodbye!\")109 break110 for event in graph.stream({\"messages\": [(\"user\", user_input)]}):111 for value in event.values():112 if isinstance(value[\"messages\"][-1], BaseMessage):113 print(\"Assistant:\", value[\"messages\"][-1].content)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.324Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":7,"totalLines":38,"estimatedTokens":2727}}134{"id":"doc-reranking_cohere_on_azure_ai_foundry_cohere-e1a81a6a","source":"documentation","title":"Reranking - Cohere on Azure AI Foundry | Cohere","url":"https://docs.cohere.com/docs/cohere-on-azure/azure-ai-reranking","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# %pip install cohere23import cohere45co = cohere.ClientV2(6 api_key=\"AZURE_API_KEY_RERANK\",7 base_url=\"AZURE_ENDPOINT_RERANK\", # example: \"https://cohere-rerank-v3-multilingual-xyz.eastus.models.ai.azure.com/\"8)\n```\n\nExample:\n```text\n1documents = [2 {3 \"Title\": \"Incorrect Password\",4 \"Content\": \"Hello, I have been trying to access my account for the past hour and it keeps saying my password is incorrect. Can you please help me?\",5 },6 {7 \"Title\": \"Confirmation Email Missed\",8 \"Content\": \"Hi, I recently purchased a product from your website but I never received a confirmation email. Can you please look into this for me?\",9 },10 {11 \"Title\": \"Questions about Return Policy\",12 \"Content\": \"Hello, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective.\",13 },14 {15 \"Title\": \"Customer Support is Busy\",16 \"Content\": \"Good morning, I have been trying to reach your customer support team for the past week but I keep getting a busy signal. Can you please help me?\",17 },18 {19 \"Title\": \"Received Wrong Item\",20 \"Content\": \"Hi, I have a question about my recent order. I received the wrong item and I need to return it.\",21 },22 {23 \"Title\": \"Customer Service is Unavailable\",24 \"Content\": \"Hello, I have been trying to reach your customer support team for the past hour but I keep getting a busy signal. Can you please help me?\",25 },26 {27 \"Title\": \"Return Policy for Defective Product\",28 \"Content\": \"Hi, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective.\",29 },30 {31 \"Title\": \"Wrong Item Received\",32 \"Content\": \"Good morning, I have a question about my recent order. I received the wrong item and I need to return it.\",33 },34 {35 \"Title\": \"Return Defective Product\",36 \"Content\": \"Hello, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective.\",37 },38]\n```\n\nExample:\n```text\n1import yaml23yaml_docs = [yaml.dump(doc, sort_keys=False) for doc in documents]45query = \"What emails have been about refunds?\"67results = co.rerank(8 model=\"model\", # Pass a dummy string9 documents=yaml_docs,10 query=query,11 top_n=3,12)\n```\n\nExample:\n```text\n1def return_results(results, documents):2 for idx, result in enumerate(results.results):3 print(f\"Rank: {idx+1}\")4 print(f\"Score: {result.relevance_score}\")5 print(f\"Document: {documents[result.index]}\\n\")678return_results(results, documents)\n```\n\nExample:\n```text\n1Rank: 12Score: 8.547617e-053Document: {'Title': 'Return Defective Product', 'Content': 'Hello, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective.'}45Rank: 26Score: 5.1442214e-057Document: {'Title': 'Questions about Return Policy', 'Content': 'Hello, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective.'}89Rank: 310Score: 3.591301e-0511Document: {'Title': 'Return Policy for Defective Product', 'Content': 'Hi, I have a question about the return policy for this product. I purchased it a few weeks ago and it is defective.'}\n```\n\nExample:\n```text\n1# Define the documents2emails = [3 {4 \"from\": \"hr@co1t.com\",5 \"to\": \"david@co1t.com\",6 \"date\": \"2024-06-24\",7 \"subject\": \"A Warm Welcome to Co1t!\",8 \"text\": \"We are delighted to welcome you to the team! As you embark on your journey with us, you'll find attached an agenda to guide you through your first week.\",9 },10 {11 \"from\": \"it@co1t.com\",12 \"to\": \"david@co1t.com\",13 \"date\": \"2024-06-24\",14 \"subject\": \"Setting Up Your IT Needs\",15 \"text\": \"Greetings! To ensure a seamless start, please refer to the attached comprehensive guide, which will assist you in setting up all your work accounts.\",16 },17 {18 \"from\": \"john@co1t.com\",19 \"to\": \"david@co1t.com\",20 \"date\": \"2024-06-24\",21 \"subject\": \"First Week Check-In\",22 \"text\": \"Hello! I hope you're settling in well. Let's connect briefly tomorrow to discuss how your first week has been going. Also, make sure to join us for a welcoming lunch this Thursday at noon—it's a great opportunity to get to know your colleagues!\",23 },24]2526yaml_emails = [yaml.dump(doc, sort_keys=False) for doc in emails]\n```\n\nExample:\n```text\n1# Add the user query2query = \"Any email about check ins?\"34# Rerank the documents5results = co.rerank(6 model=\"model\", # Pass a dummy string7 query=query,8 documents=yaml_emails,9 top_n=2,10)1112return_results(results, emails)\n```\n\nExample:\n```text\n1Rank: 12Score: 0.134775923Document: {'from': 'john@co1t.com', 'to': 'david@co1t.com', 'date': '2024-06-24', 'subject': 'First Week Check-In', 'text': \"Hello! I hope you're settling in well. Let's connect briefly tomorrow to discuss how your first week has been going. Also, make sure to join us for a welcoming lunch this Thursday at noon—it's a great opportunity to get to know your colleagues!\"}45Rank: 26Score: 0.00100834357Document: {'from': 'it@co1t.com', 'to': 'david@co1t.com', 'date': '2024-06-24', 'subject': 'Setting Up Your IT Needs', 'text': 'Greetings! To ensure a seamless start, please refer to the attached comprehensive guide, which will assist you in setting up all your work accounts.'}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.325Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":8,"totalLines":43,"estimatedTokens":1439}}135{"id":"doc-generating_multi_faceted_queries_cohere-df2f6a02","source":"documentation","title":"Generating Multi-Faceted Queries | Cohere","url":"https://docs.cohere.com/docs/generating-multi-faceted-queries","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1! pip install cohere -qq\n```\n\nExample:\n```text\n1import json2import os3import cohere45co = cohere.ClientV2(6 \"COHERE_API_KEY\"7) # Get your free API key: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1from tool_def import (2 search_code_examples_detailed,3 search_code_examples_detailed_tool,4)\n```\n\nExample:\n```text\n1functions_map = {2 \"search_code_examples_detailed\": search_code_examples_detailed,3}\n```\n\nExample:\n```text\n1tools = [search_code_examples_detailed_tool]\n```\n\nExample:\n```text\n1system_message = \"\"\"## Task and Context2You are an assistant who helps developers find code examples and tutorials on using Cohere.\"\"\"\n```\n\nExample:\n```text\n1model = \"command-a-plus-05-2026\"234def run_agent(query, messages=None):5 if messages is None:6 messages = []78 if \"system\" not in {m.get(\"role\") for m in messages}:9 messages.append({\"role\": \"system\", \"content\": system_message})1011 # Step 1: get user message12 print(f\"QUESTION:\\n{query}\")13 print(\"=\" * 50)1415 messages.append({\"role\": \"user\", \"content\": query})1617 # Step 2: Generate tool calls (if any)18 response = co.chat(19 model=model, messages=messages, tools=tools, temperature=0.320 )2122 while response.message.tool_calls:2324 print(\"TOOL PLAN:\")25 print(response.message.tool_plan, \"\\n\")26 print(\"TOOL CALLS:\")27 for tc in response.message.tool_calls:28 print(29 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"30 )31 print(\"=\" * 50)3233 messages.append(response.message)3435 # Step 3: Get tool results36 for tc in response.message.tool_calls:37 tool_result = functions_map[tc.function.name](38 **json.loads(tc.function.arguments)39 )40 tool_content = []41 for data in tool_result:42 tool_content.append(43 {44 \"type\": \"document\",45 \"document\": {\"data\": json.dumps(data)},46 }47 )48 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated49 messages.append(50 {51 \"role\": \"tool\",52 \"tool_call_id\": tc.id,53 \"content\": tool_content,54 }55 )5657 # Step 4: Generate response and citations58 response = co.chat(59 model=model,60 messages=messages,61 tools=tools,62 temperature=0.3,63 )6465 messages.append(66 {67 \"role\": \"assistant\",68 \"content\": response.message.content[0].text,69 }70 )7172 # Print final response73 print(\"RESPONSE:\")74 print(response.message.content[0].text)75 print(\"=\" * 50)7677 # Print citations (if any)78 verbose_source = (79 False # Change to True to display the contents of a source80 )81 if response.message.citations:82 print(\"CITATIONS:\\n\")83 for citation in response.message.citations:84 print(85 f\"Start: {citation.start}| End:{citation.end}| Text:'{citation.text}' \"86 )87 print(\"Sources:\")88 for idx, source in enumerate(citation.sources):89 print(f\"{idx+1}. {source.id}\")90 if verbose_source:91 print(f\"{source.tool_output}\")92 print(\"\\n\")9394 return messages\n```\n\nExample:\n```text\n1messages = run_agent(\"Do you have any RAG code examples\")2# Tool name: search_code_examples | Parameters: {\"query\":\"RAG code examples\"}\n```\n\nExample:\n```text\n1QUESTION:2Do you have any RAG code examples3==================================================4TOOL PLAN:5I will search for RAG code examples. 67TOOL CALLS:8Tool name: search_code_examples_detailed | Parameters: {\"query\":\"RAG\"}9==================================================10RESPONSE:11I found one code example for RAG with Chat, Embed and Rerank via Pinecone.12==================================================13CITATIONS:1415Start: 38| End:74| Text:'Chat, Embed and Rerank via Pinecone.' 16Sources:171. search_code_examples_detailed_kqa6j5x92e3k:2\n```\n\nExample:\n```text\n1messages = run_agent(\"Javascript tutorials on summarization\")2# Tool name: search_code_examples | Parameters: {\"programming_language\":\"js\",\"query\":\"...\"}\n```\n\nExample:\n```text\n1QUESTION:2Javascript tutorials on summarization3==================================================4TOOL PLAN:5I will search for Javascript tutorials on summarization. 67TOOL CALLS:8Tool name: search_code_examples_detailed | Parameters: {\"query\":\"summarization\",\"programming_language\":\"js\"}9==================================================10RESPONSE:11I found one JavaScript tutorial on summarization:12- Build a Chrome extension to summarize web pages13==================================================14CITATIONS:1516Start: 52| End:99| Text:'Build a Chrome extension to summarize web pages' 17Sources:181. search_code_examples_detailed_mz15bkavd7r1:0\n```\n\nExample:\n```text\n1messages = run_agent(2 \"Code examples of using embed and rerank endpoints.\"3)45# Tool name: search_code_examples | Parameters: {\"endpoints\":[\"embed\",\"rerank\"],\"query\":\"...\"}\n```\n\nExample:\n```text\n1QUESTION:2Code examples of using embed and rerank endpoints.3==================================================4TOOL PLAN:5I will search for code examples of using embed and rerank endpoints. 67TOOL CALLS:8Tool name: search_code_examples_detailed | Parameters: {\"query\":\"code examples\",\"endpoints\":[\"embed\",\"rerank\"]}9==================================================10RESPONSE:11Here are some code examples of using the embed and rerank endpoints:12- Wikipedia Semantic Search with Cohere Embedding Archives13- RAG With Chat Embed and Rerank via Pinecone14- Build Chatbots That Know Your Business with MongoDB and Cohere15==================================================16CITATIONS:1718Start: 71| End:127| Text:'Wikipedia Semantic Search with Cohere Embedding Archives' 19Sources:201. search_code_examples_detailed_qjtk4xbt5g4n:0212223Start: 130| End:173| Text:'RAG With Chat Embed and Rerank via Pinecone' 24Sources:251. search_code_examples_detailed_qjtk4xbt5g4n:1262728Start: 176| End:238| Text:'Build Chatbots That Know Your Business with MongoDB and Cohere' 29Sources:301. search_code_examples_detailed_qjtk4xbt5g4n:2\n```\n\nExample:\n```text\n1messages = run_agent(\"Python examples of using the chat endpoint.\")23# Tool name: search_code_examples | Parameters: {\"endpoints\":[\"chat\"],\"programming_language\":\"py\",\"query\":\"...\"}\n```\n\nExample:\n```text\n1QUESTION:2Python examples of using the chat endpoint.3==================================================4TOOL PLAN:5I will search for Python examples of using the chat endpoint. 67TOOL CALLS:8Tool name: search_code_examples_detailed | Parameters: {\"query\":\"chat endpoint\",\"programming_language\":\"py\",\"endpoints\":[\"chat\"]}9==================================================10RESPONSE:11Here are some Python examples of using the chat endpoint:12- Calendar Agent with Native Multi Step Tool13- RAG With Chat Embed and Rerank via Pinecone14- Build Chatbots That Know Your Business with MongoDB and Cohere15==================================================16CITATIONS:1718Start: 60| End:102| Text:'Calendar Agent with Native Multi Step Tool' 19Sources:201. search_code_examples_detailed_79er2w6sycvr:0212223Start: 105| End:148| Text:'RAG With Chat Embed and Rerank via Pinecone' 24Sources:251. search_code_examples_detailed_79er2w6sycvr:2262728Start: 151| End:213| Text:'Build Chatbots That Know Your Business with MongoDB and Cohere' 29Sources:301. search_code_examples_detailed_79er2w6sycvr:3\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.326Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":2006}}136{"id":"doc-generate_parallel_queries_for_better_rag_retriev-521169e1","source":"documentation","title":"Generate Parallel Queries for Better RAG Retrieval | Cohere","url":"https://docs.cohere.com/docs/generating-parallel-queries","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1! pip install cohere langchain langchain-community pydantic -qq\n```\n\nExample:\n```text\n1import json2import os3import cohere45from tool_def import (6 search_developer_docs,7 search_developer_docs_tool,8 search_internet,9 search_internet_tool,10 search_code_examples,11 search_code_examples_tool,12)1314co = cohere.ClientV2(15 \"COHERE_API_KEY\"16) # Get your free API key: https://dashboard.cohere.com/api-keys1718os.environ[\"TAVILY_API_KEY\"] = (19 \"TAVILY_API_KEY\" # We'll need the Tavily API key to perform internet search. Get your API key: https://app.tavily.com/home20)\n```\n\nExample:\n```text\n1functions_map = {2 \"search_developer_docs\": search_developer_docs,3 \"search_internet\": search_internet,4 \"search_code_examples\": search_code_examples,5}\n```\n\nExample:\n```text\n1tools = [2 search_developer_docs_tool,3 search_internet_tool,4 search_code_examples_tool,5]\n```\n\nExample:\n```text\n1system_message = \"\"\"## Task and Context2You are an assistant who helps developers use Cohere. You are equipped with a number of tools that can provide different types of information. If you can't find the information you need from one tool, you should try other tools if there is a possibility that they could provide the information you need.\"\"\"\n```\n\nExample:\n```text\n1model = \"command-a-plus-05-2026\"234def run_agent(query, messages=None):5 if messages is None:6 messages = []78 if \"system\" not in {m.get(\"role\") for m in messages}:9 messages.append({\"role\": \"system\", \"content\": system_message})1011 # Step 1: get user message12 print(f\"QUESTION:\\n{query}\")13 print(\"=\" * 50)1415 messages.append({\"role\": \"user\", \"content\": query})1617 # Step 2: Generate tool calls (if any)18 response = co.chat(19 model=model, messages=messages, tools=tools, temperature=0.320 )2122 while response.message.tool_calls:2324 print(\"TOOL PLAN:\")25 print(response.message.tool_plan, \"\\n\")26 print(\"TOOL CALLS:\")27 for tc in response.message.tool_calls:28 print(29 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"30 )31 print(\"=\" * 50)3233 messages.append(response.message)3435 # Step 3: Get tool results36 for tc in response.message.tool_calls:37 tool_result = functions_map[tc.function.name](38 **json.loads(tc.function.arguments)39 )40 tool_content = []41 for data in tool_result:42 tool_content.append(43 {44 \"type\": \"document\",45 \"document\": {\"data\": json.dumps(data)},46 }47 )48 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated49 messages.append(50 {51 \"role\": \"tool\",52 \"tool_call_id\": tc.id,53 \"content\": tool_content,54 }55 )5657 # Step 4: Generate response and citations58 response = co.chat(59 model=model,60 messages=messages,61 tools=tools,62 temperature=0.3,63 )6465 messages.append(66 {67 \"role\": \"assistant\",68 \"content\": response.message.content[0].text,69 }70 )7172 # Print final response73 print(\"RESPONSE:\")74 print(response.message.content[0].text)75 print(\"=\" * 50)7677 # Print citations (if any)78 verbose_source = (79 False # Change to True to display the contents of a source80 )81 if response.message.citations:82 print(\"CITATIONS:\\n\")83 for citation in response.message.citations:84 print(85 f\"Start: {citation.start}| End:{citation.end}| Text:'{citation.text}' \"86 )87 print(\"Sources:\")88 for idx, source in enumerate(citation.sources):89 print(f\"{idx+1}. {source.id}\")90 if verbose_source:91 print(f\"{source.tool_output}\")92 print(\"\\n\")9394 return messages\n```\n\nExample:\n```text\n1messages = run_agent(\"Explain the Chat endpoint and the RAG feature\")\n```\n\nExample:\n```text\n1QUESTION:2Explain the Chat endpoint and the RAG feature3==================================================4TOOL PLAN:5I will search the Cohere developer documentation for the Chat endpoint and the RAG feature. 67TOOL CALLS:8Tool name: search_developer_docs | Parameters: {\"query\":\"Chat endpoint\"}9Tool name: search_developer_docs | Parameters: {\"query\":\"RAG feature\"}10==================================================11RESPONSE:12The Chat endpoint facilitates a conversational interface, allowing users to send messages to the model and receive text responses.1314Retrieval Augmented Generation (RAG) is a method for generating text using additional information fetched from an external data source, which can greatly increase the accuracy of the response.15==================================================16CITATIONS:1718Start: 18| End:56| Text:'facilitates a conversational interface' 19Sources:201. search_developer_docs_c059cbhr042g:3212. search_developer_docs_beycjq0ejbvx:3222324Start: 58| End:130| Text:'allowing users to send messages to the model and receive text responses.' 25Sources:261. search_developer_docs_c059cbhr042g:3272. search_developer_docs_beycjq0ejbvx:3282930Start: 132| End:162| Text:'Retrieval Augmented Generation' 31Sources:321. search_developer_docs_c059cbhr042g:4332. search_developer_docs_beycjq0ejbvx:4343536Start: 174| End:266| Text:'method for generating text using additional information fetched from an external data source' 37Sources:381. search_developer_docs_c059cbhr042g:4392. search_developer_docs_beycjq0ejbvx:4404142Start: 278| End:324| Text:'greatly increase the accuracy of the response.' 43Sources:441. search_developer_docs_c059cbhr042g:4452. search_developer_docs_beycjq0ejbvx:4\n```\n\nExample:\n```text\n1messages = run_agent(2 \"What is the Embed endpoint? Give me some code tutorials\"3)\n```\n\nExample:\n```text\n1QUESTION:2What is the Embed endpoint? Give me some code tutorials3==================================================4TOOL PLAN:5I will search for 'what is the Embed endpoint' and 'Embed endpoint code tutorials' at the same time. 67TOOL CALLS:8Tool name: search_developer_docs | Parameters: {\"query\":\"what is the Embed endpoint\"}9Tool name: search_code_examples | Parameters: {\"query\":\"Embed endpoint code tutorials\"}10==================================================11RESPONSE:12The Embed endpoint returns text embeddings. An embedding is a list of floating point numbers that captures semantic information about the text that it represents.1314I'm afraid I couldn't find any code tutorials for the Embed endpoint.15==================================================16CITATIONS:1718Start: 19| End:43| Text:'returns text embeddings.' 19Sources:201. search_developer_docs_pgzdgqd3k0sd:1212223Start: 62| End:162| Text:'list of floating point numbers that captures semantic information about the text that it represents.' 24Sources:251. search_developer_docs_pgzdgqd3k0sd:1\n```\n\nExample:\n```text\n1messages = run_agent(\"What is the Chat endpoint?\")\n```\n\nExample:\n```text\n1QUESTION:2What is the Chat endpoint?3==================================================4TOOL PLAN:5I will search the Cohere developer documentation for 'Chat endpoint'. 67TOOL CALLS:8Tool name: search_developer_docs | Parameters: {\"query\":\"Chat endpoint\"}9==================================================10RESPONSE:11The Chat endpoint facilitates a conversational interface, allowing users to send messages to the model and receive text responses.12==================================================13CITATIONS:1415Start: 18| End:130| Text:'facilitates a conversational interface, allowing users to send messages to the model and receive text responses.' 16Sources:171. search_developer_docs_qx7dht277mg7:3\n```\n\nExample:\n```text\n1messages = run_agent(2 \"How is it different from RAG? Also any code tutorials?\", messages3)\n```\n\nExample:\n```text\n1QUESTION:2How is it different from RAG? Also any code tutorials?3==================================================4TOOL PLAN:5I will search the Cohere developer documentation for 'Chat endpoint vs RAG' and 'Chat endpoint code tutorials'. 67TOOL CALLS:8Tool name: search_developer_docs | Parameters: {\"query\":\"Chat endpoint vs RAG\"}9Tool name: search_code_examples | Parameters: {\"query\":\"Chat endpoint\"}10==================================================11RESPONSE:12The Chat endpoint facilitates a conversational interface, allowing users to send messages to the model and receive text responses.1314RAG (Retrieval Augmented Generation) is a method for generating text using additional information fetched from an external data source, which can greatly increase the accuracy of the response.1516I could not find any code tutorials for the Chat endpoint, but I did find a tutorial on RAG with Chat Embed and Rerank via Pinecone.17==================================================18CITATIONS:1920Start: 414| End:458| Text:'RAG with Chat Embed and Rerank via Pinecone.' 21Sources:221. search_code_examples_h8mn6mdqbrc3:2\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.326Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":14,"totalLines":73,"estimatedTokens":2376}}137{"id":"doc-cohere_on_oracle_cloud_infrastructure_oci_cohere-e3f10ac0","source":"documentation","title":"Cohere on Oracle Cloud Infrastructure (OCI) | Cohere","url":"https://docs.cohere.com/docs/oracle-cloud-infrastructure-oci","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$pip install cohere[oci]\n```\n\nExample:\n```text\n1import cohere23client = cohere.OciClientV2(4 oci_region=\"us-chicago-1\",5 oci_compartment_id=\"ocid1.compartment.oc1...\",6)78response = client.chat(9 model=\"command-a-03-2025\",10 messages=[11 {\"role\": \"system\", \"content\": \"You are a helpful assistant.\"},12 {13 \"role\": \"user\",14 \"content\": \"Explain RAG in three sentences.\",15 },16 ],17)1819print(response.message.content[0].text)\n```\n\nExample:\n```text\n1import cohere23client = cohere.OciClient(4 oci_region=\"us-chicago-1\",5 oci_compartment_id=\"ocid1.compartment.oc1...\",6)78response = client.chat(9 model=\"command-r-plus-08-2024\",10 message=\"Explain RAG in three sentences.\",11)1213print(response.text)\n```\n\nExample:\n```text\n1import cohere23client = cohere.OciClientV2(4 oci_region=\"us-chicago-1\",5 oci_compartment_id=\"ocid1.compartment.oc1...\",6)78response = client.embed(9 model=\"embed-english-v3.0\",10 texts=[\"Oracle Cloud Infrastructure\", \"Generative AI service\"],11 input_type=\"search_document\",12)1314for i, embedding in enumerate(response.embeddings.float_):15 print(f\"Text {i}: {len(embedding)} dimensions\")\n```\n\nExample:\n```text\n1import cohere23client = cohere.OciClientV2(4 oci_region=\"us-chicago-1\",5 oci_compartment_id=\"ocid1.compartment.oc1...\",6)78for event in client.chat_stream(9 model=\"command-a-03-2025\",10 messages=[11 {\"role\": \"user\", \"content\": \"Explain RAG in three sentences.\"}12 ],13):14 if event.type == \"content-delta\":15 print(event.delta.message.content.text, end=\"\")\n```\n\nExample:\n```text\n1import cohere23client = cohere.OciClient(4 oci_region=\"us-chicago-1\",5 oci_compartment_id=\"ocid1.compartment.oc1...\",6)78for event in client.chat_stream(9 model=\"command-r-plus-08-2024\",10 message=\"Explain RAG in three sentences.\",11):12 if hasattr(event, \"text\") and event.text:13 print(event.text, end=\"\")\n```\n\nExample:\n```text\n1client = cohere.OciClientV2(2 oci_region=\"us-chicago-1\",3 oci_compartment_id=\"ocid1.compartment.oc1...\",4)\n```\n\nExample:\n```text\n1client = cohere.OciClientV2(2 oci_profile=\"MY_PROFILE\",3 oci_region=\"us-chicago-1\",4 oci_compartment_id=\"ocid1.compartment.oc1...\",5)\n```\n\nExample:\n```text\n1client = cohere.OciClientV2(2 oci_profile=\"MY_SESSION_PROFILE\", # Profile with security_token_file3 oci_region=\"us-chicago-1\",4 oci_compartment_id=\"ocid1.compartment.oc1...\",5)\n```\n\nExample:\n```text\n1client = cohere.OciClientV2(2 oci_user_id=\"ocid1.user.oc1...\",3 oci_fingerprint=\"xx:xx:xx:...\",4 oci_tenancy_id=\"ocid1.tenancy.oc1...\",5 oci_private_key_path=\"~/.oci/key.pem\",6 oci_region=\"us-chicago-1\",7 oci_compartment_id=\"ocid1.compartment.oc1...\",8)\n```\n\nExample:\n```text\n1client = cohere.OciClientV2(2 auth_type=\"instance_principal\",3 oci_region=\"us-chicago-1\",4 oci_compartment_id=\"ocid1.compartment.oc1...\",5)\n```\n\nExample:\n```text\n1client = cohere.OciClientV2(2 auth_type=\"resource_principal\",3 oci_region=\"us-chicago-1\",4 oci_compartment_id=\"ocid1.compartment.oc1...\",5)\n```\n\nExample:\n```text\n1import cohere23client = cohere.OciClientV2(4 oci_region=\"us-chicago-1\",5 oci_compartment_id=\"ocid1.compartment.oc1...\",6)78response = client.chat(9 model=\"command-a-03-2025\",10 messages=[11 {\"role\": \"user\", \"content\": \"What's the weather in Toronto?\"}12 ],13 max_tokens=200,14 tools=[15 {16 \"type\": \"function\",17 \"function\": {18 \"name\": \"get_weather\",19 \"description\": \"Get current weather for a location\",20 \"parameters\": {21 \"type\": \"object\",22 \"properties\": {23 \"location\": {24 \"type\": \"string\",25 \"description\": \"City name\",26 }27 },28 \"required\": [\"location\"],29 },30 },31 }32 ],33)3435if response.message.tool_calls:36 for tc in response.message.tool_calls:37 print(f\"{tc.function.name}({tc.function.arguments})\")38# Output: get_weather({\"location\":\"Toronto\"})\n```\n\nExample:\n```text\n1import cohere2import base6434client = cohere.OciClientV2(5 oci_region=\"us-chicago-1\",6 oci_compartment_id=\"ocid1.compartment.oc1...\",7)89# Read and encode an image10with open(\"document.png\", \"rb\") as f:11 img_b64 = base64.b64encode(f.read()).decode()1213response = client.chat(14 model=\"command-a-vision\",15 messages=[16 {17 \"role\": \"user\",18 \"content\": [19 {20 \"type\": \"text\",21 \"text\": \"Describe what you see in this image.\",22 },23 {24 \"type\": \"image_url\",25 \"image_url\": {26 \"url\": f\"data:image/png;base64,{img_b64}\"27 },28 },29 ],30 }31 ],32)3334print(response.message.content[0].text)\n```\n\nExample:\n```text\n1import cohere23client = cohere.OciClientV2(4 oci_region=\"us-chicago-1\",5 oci_compartment_id=\"ocid1.compartment.oc1...\",6)78response = client.embed(9 model=\"embed-v4.0\",10 texts=[\"Oracle Cloud Infrastructure\", \"Generative AI service\"],11 input_type=\"search_document\",12)1314for i, embedding in enumerate(response.embeddings.float_):15 print(f\"Text {i}: {len(embedding)} dimensions\")16# Output: 1536 dimensions per text\n```\n\nExample:\n```text\n1import cohere2import base6434# Initialize V2 client for Command A models5client = cohere.OciClientV2(6 oci_region=\"us-chicago-1\",7 oci_compartment_id=\"ocid1.compartment.oc1...\",8)910# --- Step 1: Build a knowledge base with embeddings ---1112documents = [13 \"Oracle Cloud Infrastructure provides enterprise-grade AI services.\",14 \"Cohere Command A is a 111B parameter model with 256K context window.\",15 \"OCI Generative AI is FedRAMP High and DISA IL5 authorized.\",16]1718doc_embeddings = client.embed(19 model=\"embed-english-v3.0\",20 texts=documents,21 input_type=\"search_document\",22).embeddings.float_2324query_embedding = client.embed(25 model=\"embed-english-v3.0\",26 texts=[\"What security certifications does OCI have?\"],27 input_type=\"search_query\",28).embeddings.float_[0]2930# Find the most relevant document (cosine similarity)31best_idx = max(32 range(len(documents)),33 key=lambda i: sum(34 a * b for a, b in zip(query_embedding, doc_embeddings[i])35 ),36)37print(f\"Best match: {documents[best_idx]}\")3839# --- Step 2: Grounded chat with retrieved context ---4041response = client.chat(42 model=\"command-a-03-2025\",43 messages=[44 {45 \"role\": \"system\",46 \"content\": \"Answer based on the provided context only.\",47 },48 {49 \"role\": \"user\",50 \"content\": f\"Context: {documents[best_idx]}\\n\\nWhat certifications does OCI have?\",51 },52 ],53 temperature=0.3,54)55print(f\"Answer: {response.message.content[0].text}\")5657# --- Step 3: Tool use — call an external API ---5859response = client.chat(60 model=\"command-a-03-2025\",61 messages=[62 {63 \"role\": \"user\",64 \"content\": \"What's the current stock price of ORCL?\",65 }66 ],67 tools=[68 {69 \"type\": \"function\",70 \"function\": {71 \"name\": \"get_stock_price\",72 \"description\": \"Get the current stock price for a ticker symbol\",73 \"parameters\": {74 \"type\": \"object\",75 \"properties\": {76 \"ticker\": {77 \"type\": \"string\",78 \"description\": \"Stock ticker symbol\",79 }80 },81 \"required\": [\"ticker\"],82 },83 },84 }85 ],86)8788# Model returns a tool call89tool_call = response.message.tool_calls[0]90print(91 f\"Tool call: {tool_call.function.name}({tool_call.function.arguments})\"92)9394# Send the tool result back95final = client.chat(96 model=\"command-a-03-2025\",97 messages=[98 {99 \"role\": \"user\",100 \"content\": \"What's the current stock price of ORCL?\",101 },102 {103 \"role\": \"assistant\",104 \"tool_calls\": [105 {106 \"id\": tool_call.id,107 \"type\": \"function\",108 \"function\": {109 \"name\": tool_call.function.name,110 \"arguments\": tool_call.function.arguments,111 },112 }113 ],114 \"tool_plan\": response.message.tool_plan,115 },116 {117 \"role\": \"tool\",118 \"tool_call_id\": tool_call.id,119 \"content\": [120 {121 \"type\": \"text\",122 \"text\": '{\"ticker\": \"ORCL\", \"price\": 187.42, \"currency\": \"USD\"}',123 }124 ],125 },126 ],127 tools=[128 {129 \"type\": \"function\",130 \"function\": {131 \"name\": \"get_stock_price\",132 \"description\": \"Get the current stock price for a ticker symbol\",133 \"parameters\": {134 \"type\": \"object\",135 \"properties\": {\"ticker\": {\"type\": \"string\"}},136 \"required\": [\"ticker\"],137 },138 },139 }140 ],141)142print(f\"Final answer: {final.message.content[0].text}\")143144# --- Step 4: Vision — analyze an image ---145146with open(\"chart.png\", \"rb\") as f:147 img_b64 = base64.b64encode(f.read()).decode()148149response = client.chat(150 model=\"command-a-vision\",151 messages=[152 {153 \"role\": \"user\",154 \"content\": [155 {156 \"type\": \"text\",157 \"text\": \"Describe the trend shown in this chart.\",158 },159 {160 \"type\": \"image_url\",161 \"image_url\": {162 \"url\": f\"data:image/png;base64,{img_b64}\"163 },164 },165 ],166 }167 ],168)169print(f\"Vision: {response.message.content[0].text}\")170171# --- Step 5: Stream a response in real time ---172173print(\"Streaming: \", end=\"\")174for event in client.chat_stream(175 model=\"command-a-03-2025\",176 messages=[177 {178 \"role\": \"user\",179 \"content\": \"Summarize why enterprises choose OCI for AI.\",180 }181 ],182):183 if event.type == \"content-delta\":184 print(event.delta.message.content.text, end=\"\")185print()\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.327Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":16,"totalLines":83,"estimatedTokens":2772}}138{"id":"doc-semantic_search_with_cohere_models_cohere-52cd1c87","source":"documentation","title":"Semantic Search with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/semantic-search-with-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# pip install cohere23import cohere4import numpy as np56# Get your free API key: https://dashboard.cohere.com/api-keys7co = cohere.ClientV2(api_key=\"COHERE_API_KEY\")\n```\n\nExample:\n```text\n1# Define the documents2faqs_long = [3 {4 \"text\": \"Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.\"5 },6 {7 \"text\": \"Finding Coffee Spots: For your caffeine fix, head to the break room's coffee machine or cross the street to the café for artisan coffee.\"8 },9 {10 \"text\": \"Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!\"11 },12 {13 \"text\": \"Working Hours Flexibility: We prioritize work-life balance. While our core hours are 9 AM to 5 PM, we offer flexibility to adjust as needed.\"14 },15 {16 \"text\": \"Side Projects Policy: We encourage you to pursue your passions. Just be mindful of any potential conflicts of interest with our business.\"17 },18 {19 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"20 },21 {22 \"text\": \"Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.\"23 },24 {25 \"text\": \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\"26 },27 {28 \"text\": \"Performance Reviews Frequency: We conduct informal check-ins every quarter and formal performance reviews twice a year.\"29 },30 {31 \"text\": \"Proposing New Ideas: Innovation is welcomed! Share your brilliant ideas at our weekly team meetings or directly with your team lead.\"32 },33]3435# Embed the documents36doc_emb = co.embed(37 model=\"embed-v4.0\",38 input_type=\"search_document\",39 texts=[doc[\"text\"] for doc in documents],40).embeddings\n```\n\nExample:\n```text\n1# Add the user query2query = \"How do I stay connected to what's happening at the company?\"34# Embed the query5query_emb = co.embed(6 model=\"embed-v4.0\",7 input_type=\"search_query\",8 texts=[query],9).embeddings\n```\n\nExample:\n```text\n1# Compute dot product similarity and display results2def return_results(query_emb, doc_emb, documents):3 n = 24 scores = np.dot(query_emb, np.transpose(doc_emb))[0]5 scores_sorted = sorted(6 enumerate(scores), key=lambda x: x[1], reverse=True7 )[:n]89 for idx, item in enumerate(scores_sorted):10 print(f\"Rank: {idx+1}\")11 print(f\"Score: {item[1]}\")12 print(f\"Document: {documents[item[0]]}\\n\")131415return_results(query_emb, doc_emb, documents)\n```\n\nExample:\n```text\nRank: 1Score: 0.352135965228231Document: {'text': 'Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.'}Rank: 2Score: 0.31995661889273097Document: {'text': 'Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.'}\n```\n\nExample:\n```text\n1# Define the documents2faqs_short_fr = [3 {4 \"text\": \"Remboursement des frais de voyage : Gérez facilement vos frais de voyage en les soumettant via notre outil financier. Les approbations sont rapides et simples.\"5 },6 {7 \"text\": \"Travailler de l'étranger : Il est possible de travailler à distance depuis un autre pays. Il suffit de coordonner avec votre responsable et de vous assurer d'être disponible pendant les heures de travail.\"8 },9 {10 \"text\": \"Avantages pour la santé et le bien-être : Nous nous soucions de votre bien-être et proposons des adhésions à des salles de sport, des cours de yoga sur site et une assurance santé complète.\"11 },12 {13 \"text\": \"Fréquence des évaluations de performance : Nous organisons des bilans informels tous les trimestres et des évaluations formelles deux fois par an.\"14 },15]1617documents = faqs_short_fr1819# Embed the documents20doc_emb = co.embed(21 model=\"embed-v4.0\",22 input_type=\"search_document\",23 texts=[doc[\"text\"] for doc in documents],24).embeddings2526# Add the user query27query = \"What's your remote-working policy?\"2829# Embed the query30query_emb = co.embed(31 model=\"embed-v4.0\",32 input_type=\"search_query\",33 texts=[query],34).embeddings3536# Compute dot product similarity and display results37return_results(query_emb, doc_emb, documents)\n```\n\nExample:\n```text\nRank: 1Score: 0.442758615743984Document: {'text': \"Travailler de l'étranger : Il est possible de travailler à distance depuis un autre pays. Il suffit de coordonner avec votre responsable et de vous assurer d'être disponible pendant les heures de travail.\"}Rank: 2Score: 0.32783563708365726Document: {'text': 'Avantages pour la santé et le bien-être : Nous nous soucions de votre bien-être et proposons des adhésions à des salles de sport, des cours de yoga sur site et une assurance santé complète.'}\n```\n\nExample:\n```text\n1# Define the documents2documents = faqs_long34# Embed the documents with the given embedding types5doc_emb = co.embed(6 model=\"embed-v4.0\",7 embedding_types=[\"float\", \"int8\"],8 input_type=\"search_document\",9 texts=[doc[\"text\"] for doc in documents],10).embeddings1112# Add the user query13query = \"How do I stay connected to what's happening at the company?\"1415# Embed the query16query_emb = co.embed(17 model=\"embed-v4.0\",18 embedding_types=[\"float\", \"int8\"],19 input_type=\"search_query\",20 texts=[query],21).embeddings\n```\n\nExample:\n```text\n1# Compute dot product similarity and display results2return_results(query_emb.float, doc_emb.float, faqs_long)\n```\n\nExample:\n```text\n1# Compute dot product similarity and display results2return_results(query_emb.int8, doc_emb.int8, documents)\n```\n\nExample:\n```text\nRank: 1Score: 563583Document: {'text': 'Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.'}Rank: 2Score: 508692Document: {'text': 'Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.'}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.328Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":11,"totalLines":58,"estimatedTokens":1659}}139{"id":"doc-master_reranking_with_cohere_models_cohere-726f01d6","source":"documentation","title":"Master Reranking with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/reranking-with-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# pip install cohere23import cohere45# Get your free API key: https://dashboard.cohere.com/api-keys6co = cohere.ClientV2(api_key=\"COHERE_API_KEY\")\n```\n\nExample:\n```text\n1# Define the documents2faqs = [3 {4 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"5 },6 {7 \"text\": \"Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.\"8 },9 {10 \"text\": \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\"11 },12 {13 \"text\": \"Performance Reviews Frequency: We conduct informal check-ins every quarter and formal performance reviews twice a year.\"14 },15]\n```\n\nExample:\n```text\n1# Add the user query2query = \"Are there fitness-related perks?\"34# Rerank the documents5results = co.rerank(6 model=\"rerank-v4.0-pro\",7 query=query,8 documents=faqs,9 top_n=1,10)1112print(results)\n```\n\nExample:\n```text\nid='2fa5bc0d-28aa-4c99-8355-7de78dbf3c86' results=[RerankResponseResultsItem(document=None, index=2, relevance_score=0.01798621), RerankResponseResultsItem(document=None, index=3, relevance_score=8.463939e-06)] meta=ApiMeta(api_version=ApiMetaApiVersion(version='1', is_deprecated=None, is_experimental=None), billed_units=ApiMetaBilledUnits(input_tokens=None, output_tokens=None, search_units=1.0, classifications=None), tokens=None, warnings=None)\n```\n\nExample:\n```text\n1# Display the reranking results2def return_results(results, documents):3 for idx, result in enumerate(results.results):4 print(f\"Rank: {idx+1}\")5 print(f\"Score: {result.relevance_score}\")6 print(f\"Document: {documents[result.index]}\\n\")789return_results(results, faqs_short)\n```\n\nExample:\n```text\nRank: 1Score: 0.01798621Document: {'text': 'Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.'}Rank: 2Score: 8.463939e-06Document: {'text': 'Performance Reviews Frequency: We conduct informal check-ins every quarter and formal performance reviews twice a year.'}\n```\n\nExample:\n```text\n1# Define the documents2emails = [3 {4 \"from\": \"hr@co1t.com\",5 \"to\": \"david@co1t.com\",6 \"date\": \"2024-06-24\",7 \"subject\": \"A Warm Welcome to Co1t!\",8 \"text\": \"We are delighted to welcome you to the team! As you embark on your journey with us, you'll find attached an agenda to guide you through your first week.\",9 },10 {11 \"from\": \"it@co1t.com\",12 \"to\": \"david@co1t.com\",13 \"date\": \"2024-06-24\",14 \"subject\": \"Setting Up Your IT Needs\",15 \"text\": \"Greetings! To ensure a seamless start, please refer to the attached comprehensive guide, which will assist you in setting up all your work accounts.\",16 },17 {18 \"from\": \"john@co1t.com\",19 \"to\": \"david@co1t.com\",20 \"date\": \"2024-06-24\",21 \"subject\": \"First Week Check-In\",22 \"text\": \"Hello! I hope you're settling in well. Let's connect briefly tomorrow to discuss how your first week has been going. Also, make sure to join us for a welcoming lunch this Thursday at noon—it's a great opportunity to get to know your colleagues!\",23 },24]\n```\n\nExample:\n```text\n1# Convert the documents to YAML format2yaml_docs = [yaml.dump(doc, sort_keys=False) for doc in emails]34# Add the user query5query = \"Any email about check ins?\"67# Rerank the documents8results = co.rerank(9 model=\"rerank-v4.0-pro\",10 query=query,11 documents=yaml_docs,12 top_n=2,13)1415return_results(results, emails)\n```\n\nExample:\n```text\nRank: 1Score: 0.1979091Document: {'from': 'john@co1t.com', 'to': 'david@co1t.com', 'date': '2024-06-24', 'subject': 'First Week Check-In', 'text': \"Hello! I hope you're settling in well. Let's connect briefly tomorrow to discuss how your first week has been going. Also, make sure to join us for a welcoming lunch this Thursday at noon—it's a great opportunity to get to know your colleagues!\"}Rank: 2Score: 9.535461e-05Document: {'from': 'hr@co1t.com', 'to': 'david@co1t.com', 'date': '2024-06-24', 'subject': 'A Warm Welcome to Co1t!', 'text': \"We are delighted to welcome you to the team! As you embark on your journey with us, you'll find attached an agenda to guide you through your first week.\"}\n```\n\nExample:\n```text\n1import pandas as pd2from io import StringIO34# Create a demo CSV file5data = \"\"\"name,role,join_date,email,status6Rebecca Lee,Senior Software Engineer,2024-07-01,rebecca@co1t.com,Full-time7Emma Williams,Product Designer,2024-06-15,emma@co1t.com,Full-time8Michael Jones,Marketing Manager,2024-05-20,michael@co1t.com,Full-time9Amelia Thompson,Sales Representative,2024-05-20,amelia@co1t.com,Part-time10Ethan Davis,Product Designer,2024-05-25,ethan@co1t.com,Contractor\"\"\"11data_csv = StringIO(data)1213# Load the CSV file14df = pd.read_csv(data_csv)15df.head(1)\n```\n\nExample:\n```text\n1# Define the documents2employees = df.to_dict(\"records\")34# Convert the documents to YAML format5yaml_docs = [yaml.dump(doc, sort_keys=False) for doc in employees]67# Add the user query8query = \"Any full-time product designers who joined recently?\"910# Rerank the documents11results = co.rerank(12 model=\"rerank-v4.0-pro\",13 query=query,14 documents=yaml_docs,15 top_n=1,16)17return_results(results, employees)\n```\n\nExample:\n```text\nRank: 1Score: 0.986828Document: {'name': 'Emma Williams', 'role': 'Product Designer', 'join_date': '2024-06-15', 'email': 'emma@co1t.com', 'status': 'Full-time'}\n```\n\nExample:\n```text\n1# Define the query2query = \"هل هناك مزايا تتعلق باللياقة البدنية؟\" # Are there fitness benefits?34# Rerank the documents5results = co.rerank(6 model=\"rerank-v4.0-pro\",7 query=query,8 documents=faqs,9 top_n=1,10)1112return_results(results, faqs)\n```\n\nExample:\n```text\nRank: 1Score: 0.42232594Document: {'text': 'Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.'}Rank: 2Score: 0.00025118678Document: {'text': 'Performance Reviews Frequency: We conduct informal check-ins every quarter and formal performance reviews twice a year.'}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.328Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":14,"totalLines":73,"estimatedTokens":1651}}140{"id":"doc-text_generation_cohere_on_azure_ai_foundry_coher-8a09af64","source":"documentation","title":"Text generation - Cohere on Azure AI Foundry | Cohere","url":"https://docs.cohere.com/docs/cohere-on-azure/azure-ai-text-generation","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# %pip install cohere2import cohere34co = cohere.ClientV2(5 api_key=\"AZURE_API_KEY_CHAT\",6 base_url=\"AZURE_ENDPOINT_CHAT\", # example: \"https://cohere-command-r-plus-08-2024-xyz.eastus.models.ai.azure.com/\"7)\n```\n\nExample:\n```text\n1# Technical support FAQ2faq_tech_support = \"\"\"- Question: How do I set up my new smartphone with my mobile plan?3- Answer:4 - Insert your SIM card into the device.5 - Turn on your phone and follow the on-screen setup instructions.6 - Connect to your mobile network and enter your account details when prompted.7 - Download and install any necessary apps or updates.8 - Contact customer support if you need further assistance.910- Question: My internet connection is slow. How can I improve my mobile data speed?11- Answer:12 - Check your signal strength and move to an area with better coverage.13 - Restart your device and try connecting again.14 - Ensure your data plan is active and has sufficient data.15 - Consider upgrading your plan for faster speeds.1617- Question: I can't connect to my mobile network. What should I do?18- Answer:19 - Check your SIM card is inserted correctly and not damaged.20 - Restart your device and try connecting again.21 - Ensure your account is active and not suspended.22 - Check for any network outages in your area.23 - Contact customer support for further assistance.2425- Question: How do I set up my voicemail?26- Answer:27 - Dial your voicemail access number (usually provided by your carrier).28 - Follow the prompts to set up your voicemail greeting and password.29 - Record your voicemail greeting and save it.30 - Test your voicemail by calling your number and leaving a message.3132- Question: I'm having trouble sending text messages. What could be the issue?33- Answer:34 - Check your signal strength and move to an area with better coverage.35 - Ensure your account has sufficient credit or an active plan.36 - Restart your device and try sending a message again.37 - Check your message settings and ensure they are correct.38 - Contact customer support if the issue persists.\"\"\"\n```\n\nExample:\n```text\n1def generate_text(message):2 response = co.chat(3 model=\"model\", # Pass a dummy string4 messages=[{\"role\": \"user\", \"content\": message}],5 )6 return response\n```\n\nExample:\n```text\n1inquiry = \"I've noticed some fluctuations in my mobile network's performance recently. The connection seems stable most of the time, but every now and then, I experience brief periods of slow data speeds. It happens a few times a day and is quite inconvenient.\"23prompt = f\"\"\"Use the FAQs below to provide a concise response to this customer inquiry.45# Customer inquiry6{inquiry}78# FAQs9{faq_tech_support}\"\"\"1011response = generate_text(prompt)1213print(response.message.content[0].text)\n```\n\nExample:\n```text\n1It's quite common to experience occasional fluctuations in mobile network performance, and there are a few steps you can take to address this issue. 23First, check your signal strength and consider moving to a different location with better coverage. Sometimes, even a small change in position can make a difference. If you find that you're in an area with low signal strength, this could be the primary reason for the slow data speeds. 45Next, try restarting your device. A simple restart can often resolve temporary glitches and improve your connection. After restarting, ensure that your data plan is active and has enough data allocated for your usage. If you're close to reaching your data limit, this could also impact your speeds. 67If the issue persists, it might be worth checking for any network outages in your area. Occasionally, temporary network issues can cause intermittent slowdowns. Contact your mobile network's customer support to inquire about any known issues and to receive further guidance. 89Additionally, consider the age and condition of your device. Older devices or those with outdated software might struggle to maintain consistent data speeds. Ensuring your device is up-to-date and well-maintained can contribute to a better overall network experience. 1011If the problem continues, you may want to explore the option of upgrading your data plan. Higher-tier plans often offer faster speeds and more reliable connections, especially during peak usage times. Contact your mobile provider to discuss the available options and find a plan that better suits your needs.\n```\n\nExample:\n```text\n1prompt = f\"\"\"Summarize this customer inquiry into one short sentence.23Inquiry: {inquiry}\"\"\"45response = generate_text(prompt)67print(response.message.content[0].text)\n```\n\nExample:\n```text\n1A customer is experiencing intermittent slow data speeds on their mobile network several times a day.\n```\n\nExample:\n```text\n1prompt = f\"\"\"Rewrite this customer support agent response into an email format, ready to send to the customer.23If you're experiencing brief periods of slow data speeds or difficulty sending text messages and connecting to your mobile network, here are some troubleshooting steps you can follow:451. Check your signal strength - Move to an area with better coverage.62. Restart your device and try connecting again.73. Ensure your account is active and not suspended.84. Contact customer support for further assistance. (This can include updating your plan for better network performance.)910Did these steps help resolve the issue? Let me know if you need further assistance.\"\"\"1112response = generate_text(prompt)1314print(response.message.content[0].text)\n```\n\nExample:\n```text\n1Subject: Troubleshooting Slow Data Speeds and Network Connection Issues23Dear [Customer's Name],45I hope this email finds you well. I understand that you may be facing some challenges with your mobile network, including slow data speeds and difficulties sending text messages. Here are some recommended troubleshooting steps to help resolve these issues:67- Signal Strength: Check the signal strength on your device and move to a different location if the signal is weak. Moving to an area with better coverage can often improve your connection.89- Restart Your Device: Sometimes, a simple restart can resolve temporary glitches. Please restart your device and then try connecting to the network again.1011- Account Status: Verify that your account is active and in good standing. In some cases, service providers may temporarily suspend accounts due to various reasons, which can impact your network access. 1213- Contact Customer Support: If the issue persists, please reach out to our customer support team for further assistance. Our team can help troubleshoot and provide additional guidance. We can also discuss your current plan and explore options to enhance your network performance if needed.1415I hope these steps will help resolve the issue promptly. Please feel free to reply to this email if you have any further questions or if the problem continues. We are committed to ensuring your satisfaction and providing a seamless network experience.1617Best regards,18[Your Name]19[Customer Support Agent]20[Company Name]\n```\n\nExample:\n```text\n1# Define a system message2system_message = \"\"\"## Task and Context3You are a helpful customer support agent that assists customers of a mobile network service.\"\"\"456# Run the chatbot7def run_chatbot(message, messages=None):8 if messages is None:9 messages = []1011 if \"system\" not in {m.get(\"role\") for m in messages}:12 messages.append({\"role\": \"system\", \"content\": system_message})1314 messages.append({\"role\": \"user\", \"content\": message})1516 response = co.chat(17 model=\"model\", # Pass a dummy string18 messages=messages,19 )2021 messages.append(22 {23 \"role\": \"assistant\",24 \"content\": response.message.content[0].text,25 }26 )2728 print(response.message.content[0].text)2930 return messages\n```\n\nExample:\n```text\n1messages = run_chatbot(2 \"Hi. I've noticed some fluctuations in my mobile network's performance recently.\"3)\n```\n\nExample:\n```text\n1Hello there! I'd be happy to assist you with this issue. Network performance fluctuations can be concerning, and it's important to identify the cause to ensure you have a smooth experience. 23Can you tell me more about the problems you've been experiencing? Are there specific times or locations where the network seems to perform poorly? Any details you can provide will help me understand the situation better and offer potential solutions.\n```\n\nExample:\n```text\n1messages = run_chatbot(2 \"At times, the data speed is very poor. What should I do?\",3 messages,4)\n```\n\nExample:\n```text\n1I'm sorry to hear that you're experiencing slow data speeds. Here are some troubleshooting steps and tips to help improve your network performance:23- **Check Network Coverage:** First, ensure that you are in an area with good network coverage. You can check the coverage map provided by your mobile network service on their website. If you're in a location with known weak signal strength, moving to a different area might improve your data speed.45- **Restart Your Device:** Sometimes, a simple restart of your mobile device can help refresh the network connection. Power off your device, wait for a few moments, and then turn it back on.67- **Check for Network Updates:** Make sure your device is running the latest software and carrier settings. Updates often include improvements and optimizations for network performance. You can check for updates in your device's settings.89- **Manage Network Settings:**10 - *Network Mode:* Try switching to a different network mode (e.g., 4G/LTE, 3G) to see if a specific network band provides better speed.11 - *Airplane Mode:* Toggle Airplane mode on and off to reconnect to the network.12 - *Network Reset:* If the issue persists, you can try resetting your network settings, but note that this will erase saved Wi-Fi passwords.1314- **Contact Customer Support:** If the problem continues, it might be beneficial to contact your mobile network's customer support team. They can check for any known issues in your area and provide further assistance. They might also guide you through advanced troubleshooting steps.1516- **Consider Network Congestion:** Slow data speeds can sometimes occur during peak usage hours when the network is congested. Try using data-intensive apps during off-peak hours to see if that makes a difference.1718- **Check Background Apps:** Certain apps running in the background can consume data and impact speed. Close any unnecessary apps to free up resources.1920If the slow data speed persists despite these measures, it's advisable to reach out to your mobile network provider for further investigation and assistance. They can provide network-specific solutions and ensure you're getting the service you expect.\n```\n\nExample:\n```text\n1messages = run_chatbot(\"Thanks. What else can I check?\", messages)\n```\n\nExample:\n```text\n1ou're welcome! Here are some additional steps and factors to consider:23- **Device Health:** Ensure your device is in good working condition. An older device or one with hardware issues might struggle to maintain a fast data connection. Consider checking for any pending system updates that could optimize your device's performance.45- **SIM Card:** Try removing and reinserting your SIM card to ensure it is properly seated. A loose connection can impact network performance. If the issue persists, it might be worth asking your network provider for a SIM replacement.67- **Network Congestion at Specific Times:** Network speed can vary depending on the time of day. If possible, monitor your data speed during different parts of the day to identify any patterns. This can help determine if network congestion during peak hours is the primary cause.89- **Data Plan and Throttling:** Check your mobile data plan to ensure you haven't exceeded any data limits, which could result in reduced speeds. Some providers throttle speeds after a certain data threshold is reached.1011- **Background Updates and Downloads:** Certain apps might be set to update or download content in the background, consuming data and potentially slowing down your connection. Review your app settings and consider disabling automatic updates or background data usage for apps that don't require real-time updates.1213- **Network Diagnostics Tools:** Some mobile devices have built-in network diagnostics tools that can provide insights into your connection. These tools can help identify issues with signal strength, network latency, and more.1415- **Wi-Fi Calling and Data Usage:** If your device supports Wi-Fi calling, ensure it is enabled. This can offload some data usage from the cellular network, potentially improving speeds.1617- **Network Provider's App:** Download and install your mobile network provider's official app, if available. These apps often provide real-time network status updates and allow you to report issues directly.1819If you've gone through these checks and the problem persists, contacting your network provider's technical support team is the next best step. They can provide further guidance based on your specific situation.\n```\n\nExample:\n```text\n1print(\"Chat history:\")2for message in messages:3 print(message, \"\\n\")\n```\n\nExample:\n```text\n1Chat history:2{'role': 'system', 'content': '## Task and Context\\nYou are a helpful customer support agent that assists customers of a mobile network service.'} 34{'role': 'user', 'content': \"Hi. I've noticed some fluctuations in my mobile network's performance recently.\"} 56{'role': 'assistant', 'content': \"Hello there! I'd be happy to assist you with this issue. Network performance fluctuations can be concerning, and it's important to identify the cause to ensure you have a smooth experience. \\n\\nCan you tell me more about the problems you've been experiencing? Are there specific times or locations where the network seems to perform poorly? Any details you can provide will help me understand the situation better and offer potential solutions.\"} 78{'role': 'user', 'content': 'At times, the data speed is very poor. What should I do?'} 910{'role': 'assistant', 'content': \"I'm sorry to hear that you're experiencing slow data speeds. Here are some troubleshooting steps and tips to help improve your network performance:\\n\\n- **Check Network Coverage:** First, ensure that you are in an area with good network coverage. You can check the coverage map provided by your mobile network service on their website. If you're in a location with known weak signal strength, moving to a different area might improve your data speed.\\n\\n- **Restart Your Device:** Sometimes, a simple restart of your mobile device can help refresh the network connection. Power off your device, wait for a few moments, and then turn it back on.\\n\\n- **Check for Network Updates:** Make sure your device is running the latest software and carrier settings. Updates often include improvements and optimizations for network performance. You can check for updates in your device's settings.\\n\\n- **Manage Network Settings:**\\n - *Network Mode:* Try switching to a different network mode (e.g., 4G/LTE, 3G) to see if a specific network band provides better speed.\\n - *Airplane Mode:* Toggle Airplane mode on and off to reconnect to the network.\\n - *Network Reset:* If the issue persists, you can try resetting your network settings, but note that this will erase saved Wi-Fi passwords.\\n\\n- **Contact Customer Support:** If the problem continues, it might be beneficial to contact your mobile network's customer support team. They can check for any known issues in your area and provide further assistance. They might also guide you through advanced troubleshooting steps.\\n\\n- **Consider Network Congestion:** Slow data speeds can sometimes occur during peak usage hours when the network is congested. Try using data-intensive apps during off-peak hours to see if that makes a difference.\\n\\n- **Check Background Apps:** Certain apps running in the background can consume data and impact speed. Close any unnecessary apps to free up resources.\\n\\nIf the slow data speed persists despite these measures, it's advisable to reach out to your mobile network provider for further investigation and assistance. They can provide network-specific solutions and ensure you're getting the service you expect.\"} 1112{'role': 'user', 'content': 'Thanks. What else can I check?'} 1314{'role': 'assistant', 'content': \"You're welcome! Here are some additional steps and factors to consider:\\n\\n- **Device Health:** Ensure your device is in good working condition. An older device or one with hardware issues might struggle to maintain a fast data connection. Consider checking for any pending system updates that could optimize your device's performance.\\n\\n- **SIM Card:** Try removing and reinserting your SIM card to ensure it is properly seated. A loose connection can impact network performance. If the issue persists, it might be worth asking your network provider for a SIM replacement.\\n\\n- **Network Congestion at Specific Times:** Network speed can vary depending on the time of day. If possible, monitor your data speed during different parts of the day to identify any patterns. This can help determine if network congestion during peak hours is the primary cause.\\n\\n- **Data Plan and Throttling:** Check your mobile data plan to ensure you haven't exceeded any data limits, which could result in reduced speeds. Some providers throttle speeds after a certain data threshold is reached.\\n\\n- **Background Updates and Downloads:** Certain apps might be set to update or download content in the background, consuming data and potentially slowing down your connection. Review your app settings and consider disabling automatic updates or background data usage for apps that don't require real-time updates.\\n\\n- **Network Diagnostics Tools:** Some mobile devices have built-in network diagnostics tools that can provide insights into your connection. These tools can help identify issues with signal strength, network latency, and more.\\n\\n- **Wi-Fi Calling and Data Usage:** If your device supports Wi-Fi calling, ensure it is enabled. This can offload some data usage from the cellular network, potentially improving speeds.\\n\\n- **Network Provider's App:** Download and install your mobile network provider's official app, if available. These apps often provide real-time network status updates and allow you to report issues directly.\\n\\nIf you've gone through these checks and the problem persists, contacting your network provider's technical support team is the next best step. They can provide further guidance based on your specific situation.\"}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.329Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":18,"totalLines":93,"estimatedTokens":4737}}141{"id":"doc-performing_tasks_sequentially_with_cohere_s_rag_-69c811e6","source":"documentation","title":"Performing Tasks Sequentially with Cohere's RAG | Cohere","url":"https://docs.cohere.com/docs/performing-tasks-sequentially","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1! pip install cohere langchain langchain-community pydantic -qq\n```\n\nExample:\n```text\n1import json2import os3import cohere45from tool_def import (6 search_developer_docs,7 search_developer_docs_tool,8 search_internet,9 search_internet_tool,10 search_code_examples,11 search_code_examples_tool,12)1314co = cohere.ClientV2(15 \"COHERE_API_KEY\"16) # Get your free API key: https://dashboard.cohere.com/api-keys1718os.environ[\"TAVILY_API_KEY\"] = (19 \"TAVILY_API_KEY\" # We'll need the Tavily API key to perform internet search. Get your API key: https://app.tavily.com/home20)\n```\n\nExample:\n```text\n1functions_map = {2 \"search_developer_docs\": search_developer_docs,3 \"search_internet\": search_internet,4 \"search_code_examples\": search_code_examples,5}\n```\n\nExample:\n```text\n1tools = [2 search_developer_docs_tool,3 search_internet_tool,4 search_code_examples_tool,5]\n```\n\nExample:\n```text\n1system_message = \"\"\"## Task and Context2You are an assistant who helps developers use Cohere. You are equipped with a number of tools that can provide different types of information. If you can't find the information you need from one tool, you should try other tools if there is a possibility that they could provide the information you need.\"\"\"\n```\n\nExample:\n```text\n1model = \"command-a-plus-05-2026\"234def run_agent(query, messages=None):5 if messages is None:6 messages = []78 if \"system\" not in {m.get(\"role\") for m in messages}:9 messages.append({\"role\": \"system\", \"content\": system_message})1011 # Step 1: get user message12 print(f\"QUESTION:\\n{query}\")13 print(\"=\" * 50)1415 messages.append({\"role\": \"user\", \"content\": query})1617 # Step 2: Generate tool calls (if any)18 response = co.chat(19 model=model, messages=messages, tools=tools, temperature=0.320 )2122 while response.message.tool_calls:2324 print(\"TOOL PLAN:\")25 print(response.message.tool_plan, \"\\n\")26 print(\"TOOL CALLS:\")27 for tc in response.message.tool_calls:28 print(29 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"30 )31 print(\"=\" * 50)3233 messages.append(response.message)3435 # Step 3: Get tool results36 for tc in response.message.tool_calls:37 tool_result = functions_map[tc.function.name](38 **json.loads(tc.function.arguments)39 )40 tool_content = []41 for data in tool_result:42 tool_content.append(43 {44 \"type\": \"document\",45 \"document\": {\"data\": json.dumps(data)},46 }47 )48 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated49 messages.append(50 {51 \"role\": \"tool\",52 \"tool_call_id\": tc.id,53 \"content\": tool_content,54 }55 )5657 # Step 4: Generate response and citations58 response = co.chat(59 model=model,60 messages=messages,61 tools=tools,62 temperature=0.3,63 )6465 messages.append(66 {67 \"role\": \"assistant\",68 \"content\": response.message.content[0].text,69 }70 )7172 # Print final response73 print(\"RESPONSE:\")74 print(response.message.content[0].text)75 print(\"=\" * 50)7677 # Print citations (if any)78 verbose_source = (79 False # Change to True to display the contents of a source80 )81 if response.message.citations:82 print(\"CITATIONS:\\n\")83 for citation in response.message.citations:84 print(85 f\"Start: {citation.start}| End:{citation.end}| Text:'{citation.text}' \"86 )87 print(\"Sources:\")88 for idx, source in enumerate(citation.sources):89 print(f\"{idx+1}. {source.id}\")90 if verbose_source:91 print(f\"{source.tool_output}\")92 print(\"\\n\")9394 return messages\n```\n\nExample:\n```text\n1messages = run_agent(2 \"What's the Cohere feature to reorder search results? Do you have any code examples on that?\"3)\n```\n\nExample:\n```text\n1QUESTION:2What's the Cohere feature to reorder search results? Do you have any code examples on that?3==================================================4TOOL PLAN:5I will search for the Cohere feature to reorder search results. Then I will search for code examples on that. 67TOOL CALLS:8Tool name: search_developer_docs | Parameters: {\"query\":\"reorder search results\"}9==================================================10TOOL PLAN:11I found that the Rerank endpoint is the feature that reorders search results. I will now search for code examples on that. 1213TOOL CALLS:14Tool name: search_code_examples | Parameters: {\"query\":\"rerank endpoint\"}15==================================================16RESPONSE:17The Rerank endpoint is the feature that reorders search results. Unfortunately, I could not find any code examples on that.18==================================================19CITATIONS:2021Start: 4| End:19| Text:'Rerank endpoint' 22Sources:231. search_developer_docs_53tfk9zgwgzt:0\n```\n\nExample:\n```text\n1messages = run_agent(2 \"Who are the CEOs of the companies with the top 3 highest market capitalization.\"3)\n```\n\nExample:\n```text\n1QUESTION:2Who are the CEOs of the companies with the top 3 highest market capitalization.3==================================================4TOOL PLAN:5I will search for the top 3 companies with the highest market capitalization. Then, I will search for the CEOs of those companies. 67TOOL CALLS:8Tool name: search_internet | Parameters: {\"query\":\"top 3 companies with highest market capitalization\"}9==================================================10TOOL PLAN:11The top 3 companies with the highest market capitalization are Apple, Microsoft, and Nvidia. I will now search for the CEOs of these companies. 1213TOOL CALLS:14Tool name: search_internet | Parameters: {\"query\":\"Apple CEO\"}15Tool name: search_internet | Parameters: {\"query\":\"Microsoft CEO\"}16Tool name: search_internet | Parameters: {\"query\":\"Nvidia CEO\"}17==================================================18RESPONSE:19The CEOs of the top 3 companies with the highest market capitalization are:201. Tim Cook of Apple212. Satya Nadella of Microsoft223. Jensen Huang of Nvidia23==================================================24CITATIONS:2526Start: 79| End:87| Text:'Tim Cook' 27Sources:281. search_internet_0f8wyxfc3hmn:0292. search_internet_0f8wyxfc3hmn:1303. search_internet_0f8wyxfc3hmn:2313233Start: 91| End:96| Text:'Apple' 34Sources:351. search_internet_kb9qgs1ps69e:0363738Start: 100| End:113| Text:'Satya Nadella' 39Sources:401. search_internet_wy4mn7286a88:0412. search_internet_wy4mn7286a88:1423. search_internet_wy4mn7286a88:2434445Start: 117| End:126| Text:'Microsoft' 46Sources:471. search_internet_kb9qgs1ps69e:0484950Start: 130| End:142| Text:'Jensen Huang' 51Sources:521. search_internet_q9ahz81npfqz:0532. search_internet_q9ahz81npfqz:1543. search_internet_q9ahz81npfqz:2554. search_internet_q9ahz81npfqz:3565758Start: 146| End:152| Text:'Nvidia' 59Sources:601. search_internet_kb9qgs1ps69e:0\n```\n\nExample:\n```text\n1messages = run_agent(2 \"Who are the authors of the sentence BERT paper?\"3)\n```\n\nExample:\n```text\n1QUESTION:2Who are the authors of the sentence BERT paper?3==================================================4TOOL PLAN:5I will search for the authors of the sentence BERT paper. 67TOOL CALLS:8Tool name: search_developer_docs | Parameters: {\"query\":\"authors of the sentence BERT paper\"}9==================================================10TOOL PLAN:11I was unable to find any information about the authors of the sentence BERT paper. I will now search for 'sentence BERT paper authors'. 1213TOOL CALLS:14Tool name: search_internet | Parameters: {\"query\":\"sentence BERT paper authors\"}15==================================================16RESPONSE:17The authors of the Sentence-BERT paper are Nils Reimers and Iryna Gurevych.18==================================================19CITATIONS:2021Start: 43| End:55| Text:'Nils Reimers' 22Sources:231. search_internet_z8t19852my9q:0242. search_internet_z8t19852my9q:1253. search_internet_z8t19852my9q:2264. search_internet_z8t19852my9q:3275. search_internet_z8t19852my9q:4282930Start: 60| End:75| Text:'Iryna Gurevych.' 31Sources:321. search_internet_z8t19852my9q:0332. search_internet_z8t19852my9q:1343. search_internet_z8t19852my9q:2354. search_internet_z8t19852my9q:3365. search_internet_z8t19852my9q:4\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.330Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":2238}}142{"id":"doc-tool_use_agents_cohere_on_azure_ai_foundry_coher-e45ae568","source":"documentation","title":"Tool use & agents - Cohere on Azure AI Foundry | Cohere","url":"https://docs.cohere.com/docs/cohere-on-azure/azure-ai-tool-use","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# %pip install cohere2import cohere34co = cohere.ClientV2(5 api_key=\"AZURE_API_KEY_CHAT\",6 base_url=\"AZURE_ENDPOINT_CHAT\", # example: \"https://cohere-command-r-plus-08-2024-xyz.eastus.models.ai.azure.com/\"7)\n```\n\nExample:\n```text\n1def search_faqs(query):2 faqs = [3 {4 \"text\": \"Submitting Travel Expenses:\\nSubmit your expenses through our user-friendly finance tool.\"5 },6 {7 \"text\": \"Side Projects Policy:\\nWe encourage you to explore your passions! Just ensure there's no conflict of interest with our business.\"8 },9 {10 \"text\": \"Wellness Benefits:\\nTo promote a healthy lifestyle, we provide gym memberships, on-site yoga classes, and health insurance.\"11 },12 ]13 return faqs141516def search_emails(query):17 emails = [18 {19 \"from\": \"hr@co1t.com\",20 \"to\": \"david@co1t.com\",21 \"date\": \"2024-06-24\",22 \"subject\": \"A Warm Welcome to Co1t, David!\",23 \"text\": \"We are delighted to have you on board. Please find attached your first week's agenda.\",24 },25 {26 \"from\": \"it@co1t.com\",27 \"to\": \"david@co1t.com\",28 \"date\": \"2024-06-24\",29 \"subject\": \"Instructions for IT Setup\",30 \"text\": \"Welcome, David! To get you started, please follow the attached guide to set up your work accounts.\",31 },32 {33 \"from\": \"john@co1t.com\",34 \"to\": \"david@co1t.com\",35 \"date\": \"2024-06-24\",36 \"subject\": \"First Week Check-In\",37 \"text\": \"Hi David, let's chat briefly tomorrow to discuss your first week. Also, come join us for lunch this Thursday at noon to meet everyone!\",38 },39 ]40 return emails414243def create_calendar_event(date: str, time: str, duration: int):44 # You can implement any logic here45 return {46 \"is_success\": True,47 \"message\": f\"Created a {duration} hour long event at {time} on {date}\",48 }495051functions_map = {52 \"search_faqs\": search_faqs,53 \"search_emails\": search_emails,54 \"create_calendar_event\": create_calendar_event,55}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"search_faqs\",6 \"description\": \"Given a user query, searches a company's frequently asked questions (FAQs) list and returns the most relevant matches to the query.\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"query\": {11 \"type\": \"string\",12 \"description\": \"The query from the user\",13 }14 },15 \"required\": [\"query\"],16 },17 },18 },19 {20 \"type\": \"function\",21 \"function\": {22 \"name\": \"search_emails\",23 \"description\": \"Given a user query, searches a person's emails and returns the most relevant matches to the query.\",24 \"parameters\": {25 \"type\": \"object\",26 \"properties\": {27 \"query\": {28 \"type\": \"string\",29 \"description\": \"The query from the user\",30 }31 },32 \"required\": [\"query\"],33 },34 },35 },36 {37 \"type\": \"function\",38 \"function\": {39 \"name\": \"create_calendar_event\",40 \"description\": \"Creates a new calendar event of the specified duration at the specified time and date. A new event cannot be created on the same time as an existing event.\",41 \"parameters\": {42 \"type\": \"object\",43 \"properties\": {44 \"date\": {45 \"type\": \"string\",46 \"description\": \"the date on which the event starts, formatted as mm/dd/yy\",47 },48 \"time\": {49 \"type\": \"string\",50 \"description\": \"the time of the event, formatted using 24h military time formatting\",51 },52 \"duration\": {53 \"type\": \"number\",54 \"description\": \"the number of hours the event lasts for\",55 },56 },57 \"required\": [\"date\", \"time\", \"duration\"],58 },59 },60 },61]\n```\n\nExample:\n```text\n1import json23system_message = \"\"\"## Task and Context4You are an assistant who assists new employees of Co1t with their first week. You respond to their questions and assist them with their needs. Today is Monday, June 24, 2024\"\"\"567def run_assistant(query, messages=None):8 if messages is None:9 messages = []1011 if \"system\" not in {m.get(\"role\") for m in messages}:12 messages.append({\"role\": \"system\", \"content\": system_message})1314 # Step 1: get user message15 print(f\"Question:\\n{query}\")16 print(\"=\" * 50)1718 messages.append({\"role\": \"user\", \"content\": query})1920 # Step 2: Generate tool calls (if any)21 response = co.chat(22 model=\"model\", # Pass a dummy string23 messages=messages,24 tools=tools,25 )2627 while response.message.tool_calls:2829 print(\"Tool plan:\")30 print(response.message.tool_plan, \"\\n\")31 print(\"Tool calls:\")32 for tc in response.message.tool_calls:33 print(34 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"35 )36 print(\"=\" * 50)3738 messages.append(response.message)3940 # Step 3: Get tool results41 for idx, tc in enumerate(response.message.tool_calls):42 tool_result = functions_map[tc.function.name](43 **json.loads(tc.function.arguments)44 )45 tool_content = []46 for data in tool_result:47 tool_content.append(48 {49 \"type\": \"document\",50 \"document\": {\"data\": json.dumps(data)},51 }52 )53 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated54 messages.append(55 {56 \"role\": \"tool\",57 \"tool_call_id\": tc.id,58 \"content\": tool_content,59 }60 )6162 # Step 4: Generate response and citations63 response = co.chat(64 model=\"model\", # Pass a dummy string65 messages=messages,66 tools=tools,67 )6869 messages.append(70 {71 \"role\": \"assistant\",72 \"content\": response.message.content[0].text,73 }74 )7576 # Print final response77 print(\"Response:\")78 print(response.message.content[0].text)79 print(\"=\" * 50)8081 # Print citations (if any)82 if response.message.citations:83 print(\"\\nCITATIONS:\")84 for citation in response.message.citations:85 print(citation, \"\\n\")8687 return messages\n```\n\nExample:\n```text\n1messages = run_assistant(2 \"Any doc on how do I submit travel expenses? Also, any emails about setting up IT access?\"3)\n```\n\nExample:\n```text\n1Question:2Any doc on how do I submit travel expenses? Also, any emails about setting up IT access?3==================================================4Tool plan:5I will search for a document on how to submit travel expenses, and also search for emails about setting up IT access. 67Tool calls:8Tool name: search_faqs | Parameters: {\"query\":\"how to submit travel expenses\"}9Tool name: search_emails | Parameters: {\"query\":\"setting up IT access\"}10==================================================11Response:12You can submit your travel expenses through the user-friendly finance tool.1314You should have received an email from it@co1t.com with instructions for setting up your IT access.15==================================================1617CITATIONS:18start=48 end=75 text='user-friendly finance tool.' sources=[ToolSource(type='tool', id='search_faqs_wkfggn2680c4:0', tool_output={'text': 'Submitting Travel Expenses:\\nSubmit your expenses through our user-friendly finance tool.'})] type='TEXT_CONTENT' 1920start=105 end=176 text='email from it@co1t.com with instructions for setting up your IT access.' sources=[ToolSource(type='tool', id='search_emails_8n0cvsh5xknt:1', tool_output={'date': '2024-06-24', 'from': 'it@co1t.com', 'subject': 'Instructions for IT Setup', 'text': 'Welcome, David! To get you started, please follow the attached guide to set up your work accounts.', 'to': 'david@co1t.com'})] type='TEXT_CONTENT'\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.331Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":6,"totalLines":33,"estimatedTokens":2253}}143{"id":"doc-routing_queries_to_data_sources_cohere-5524200b","source":"documentation","title":"Routing Queries to Data Sources | Cohere","url":"https://docs.cohere.com/docs/routing-queries-to-data-sources","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1! pip install cohere langchain langchain-community pydantic -qq\n```\n\nExample:\n```text\n1import json2import os3import cohere45from tool_def import (6 search_developer_docs,7 search_developer_docs_tool,8 search_internet,9 search_internet_tool,10 search_code_examples,11 search_code_examples_tool,12)1314co = cohere.ClientV2(15 \"COHERE_API_KEY\"16) # Get your free API key: https://dashboard.cohere.com/api-keys1718os.environ[\"TAVILY_API_KEY\"] = (19 \"TAVILY_API_KEY\" # We'll need the Tavily API key to perform internet search. Get your API key: https://app.tavily.com/home20)\n```\n\nExample:\n```text\n1functions_map = {2 \"search_developer_docs\": search_developer_docs,3 \"search_internet\": search_internet,4 \"search_code_examples\": search_code_examples,5}\n```\n\nExample:\n```text\n1tools = [2 search_developer_docs_tool,3 search_internet_tool,4 search_code_examples_tool,5]\n```\n\nExample:\n```text\n1system_message = \"\"\"## Task and Context2You are an assistant who helps developers use Cohere. You are equipped with a number of tools that can provide different types of information. If you can't find the information you need from one tool, you should try other tools if there is a possibility that they could provide the information you need.\"\"\"\n```\n\nExample:\n```text\n1model = \"command-a-plus-05-2026\"234def run_agent(query, messages=None):5 if messages is None:6 messages = []78 if \"system\" not in {m.get(\"role\") for m in messages}:9 messages.append({\"role\": \"system\", \"content\": system_message})1011 # Step 1: get user message12 print(f\"QUESTION:\\n{query}\")13 print(\"=\" * 50)1415 messages.append({\"role\": \"user\", \"content\": query})1617 # Step 2: Generate tool calls (if any)18 response = co.chat(19 model=model, messages=messages, tools=tools, temperature=0.320 )2122 while response.message.tool_calls:2324 print(\"TOOL PLAN:\")25 print(response.message.tool_plan, \"\\n\")26 print(\"TOOL CALLS:\")27 for tc in response.message.tool_calls:28 print(29 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"30 )31 print(\"=\" * 50)3233 messages.append(response.message)3435 # Step 3: Get tool results36 for tc in response.message.tool_calls:37 tool_result = functions_map[tc.function.name](38 **json.loads(tc.function.arguments)39 )40 tool_content = []41 for data in tool_result:42 tool_content.append(43 {44 \"type\": \"document\",45 \"document\": {\"data\": json.dumps(data)},46 }47 )48 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated49 messages.append(50 {51 \"role\": \"tool\",52 \"tool_call_id\": tc.id,53 \"content\": tool_content,54 }55 )5657 # Step 4: Generate response and citations58 response = co.chat(59 model=model,60 messages=messages,61 tools=tools,62 temperature=0.3,63 )6465 messages.append(66 {67 \"role\": \"assistant\",68 \"content\": response.message.content[0].text,69 }70 )7172 # Print final response73 print(\"RESPONSE:\")74 print(response.message.content[0].text)75 print(\"=\" * 50)7677 # Print citations (if any)78 verbose_source = (79 False # Change to True to display the contents of a source80 )81 if response.message.citations:82 print(\"CITATIONS:\\n\")83 for citation in response.message.citations:84 print(85 f\"Start: {citation.start}| End:{citation.end}| Text:'{citation.text}' \"86 )87 print(\"Sources:\")88 for idx, source in enumerate(citation.sources):89 print(f\"{idx+1}. {source.id}\")90 if verbose_source:91 print(f\"{source.tool_output}\")92 print(\"\\n\")9394 return messages\n```\n\nExample:\n```text\n1messages = run_agent(\"How many languages does Embed support?\")\n```\n\nExample:\n```text\n1QUESTION:2How many languages does Embed support?3==================================================4TOOL PLAN:5I will search the Cohere developer documentation for 'how many languages does Embed support'. 67TOOL CALLS:8Tool name: search_developer_docs | Parameters: {\"query\":\"how many languages does Embed support\"}9==================================================10RESPONSE:11The Embed endpoint supports over 100 languages.12==================================================13CITATIONS:1415Start: 28| End:47| Text:'over 100 languages.' 16Sources:171. search_developer_docs_gwt5g55gjc3w:2\n```\n\nExample:\n```text\n1messages = run_agent(\"How to set up the Notion API.\")\n```\n\nExample:\n```text\n1QUESTION:2How to set up the Notion API.3==================================================4TOOL PLAN:5I will search for 'Notion API setup' to find out how to set up the Notion API. 67TOOL CALLS:8Tool name: search_internet | Parameters: {\"query\":\"Notion API setup\"}9==================================================10RESPONSE:11To set up the Notion API, you need to create a new integration in Notion's integrations dashboard. You can do this by navigating to https://www.notion.com/my-integrations and clicking '+ New integration'.1213Once you've done this, you'll need to get your API secret by visiting the Configuration tab. You should keep your API secret just that – a secret! You can refresh your secret if you accidentally expose it.1415Next, you'll need to give your integration page permissions. To do this, you'll need to pick or create a Notion page, then click on the ... More menu in the top-right corner of the page. Scroll down to + Add Connections, then search for your integration and select it. You'll then need to confirm the integration can access the page and all of its child pages.1617If your API requests are failing, you should confirm you have given the integration permission to the page you are trying to update.1819You can also create a Notion API integration and get your internal integration token. You'll then need to create a .env file and add environmental variables, get your Notion database ID and add your integration to your database.2021For more information on what you can build with Notion's API, you can refer to this guide.22==================================================23CITATIONS:2425Start: 38| End:62| Text:'create a new integration' 26Sources:271. search_internet_cwabyfc5mn8c:0282. search_internet_cwabyfc5mn8c:2293031Start: 75| End:98| Text:'integrations dashboard.' 32Sources:331. search_internet_cwabyfc5mn8c:2343536Start: 132| End:170| Text:'https://www.notion.com/my-integrations' 37Sources:381. search_internet_cwabyfc5mn8c:0394041Start: 184| End:203| Text:''+ New integration'' 42Sources:431. search_internet_cwabyfc5mn8c:0442. search_internet_cwabyfc5mn8c:2454647Start: 244| End:263| Text:'get your API secret' 48Sources:491. search_internet_cwabyfc5mn8c:2505152Start: 280| End:298| Text:'Configuration tab.' 53Sources:541. search_internet_cwabyfc5mn8c:2555657Start: 310| End:351| Text:'keep your API secret just that – a secret' 58Sources:591. search_internet_cwabyfc5mn8c:2606162Start: 361| End:411| Text:'refresh your secret if you accidentally expose it.' 63Sources:641. search_internet_cwabyfc5mn8c:2656667Start: 434| End:473| Text:'give your integration page permissions.' 68Sources:691. search_internet_cwabyfc5mn8c:2707172Start: 501| End:529| Text:'pick or create a Notion page' 73Sources:741. search_internet_cwabyfc5mn8c:2757677Start: 536| End:599| Text:'click on the ... More menu in the top-right corner of the page.' 78Sources:791. search_internet_cwabyfc5mn8c:2808182Start: 600| End:632| Text:'Scroll down to + Add Connections' 83Sources:841. search_internet_cwabyfc5mn8c:2858687Start: 639| End:681| Text:'search for your integration and select it.' 88Sources:891. search_internet_cwabyfc5mn8c:2909192Start: 702| End:773| Text:'confirm the integration can access the page and all of its child pages.' 93Sources:941. search_internet_cwabyfc5mn8c:2959697Start: 783| End:807| Text:'API requests are failing' 98Sources:991. search_internet_cwabyfc5mn8c:2100101102Start: 820| End:907| Text:'confirm you have given the integration permission to the page you are trying to update.' 103Sources:1041. search_internet_cwabyfc5mn8c:2105106107Start: 922| End:953| Text:'create a Notion API integration' 108Sources:1091. search_internet_cwabyfc5mn8c:1110111112Start: 958| End:994| Text:'get your internal integration token.' 113Sources:1141. search_internet_cwabyfc5mn8c:1115116117Start: 1015| End:1065| Text:'create a .env file and add environmental variables' 118Sources:1191. search_internet_cwabyfc5mn8c:1120121122Start: 1067| End:1094| Text:'get your Notion database ID' 123Sources:1241. search_internet_cwabyfc5mn8c:1125126127Start: 1099| End:1137| Text:'add your integration to your database.' 128Sources:1291. search_internet_cwabyfc5mn8c:1130131132Start: 1223| End:1229| Text:'guide.' 133Sources:1341. search_internet_cwabyfc5mn8c:3\n```\n\nExample:\n```text\n1messages = run_agent(2 \"Any tutorials that are relevant for enterprises?\"3)\n```\n\nExample:\n```text\n1QUESTION:2Any tutorials that are relevant for enterprises?3==================================================4TOOL PLAN:5I will search for 'enterprise tutorials' in the code examples and tutorials tool. 67TOOL CALLS:8Tool name: search_code_examples | Parameters: {\"query\":\"enterprise tutorials\"}9==================================================10RESPONSE:11I found a tutorial called 'Advanced Document Parsing For Enterprises'.12==================================================13CITATIONS:1415Start: 26| End:69| Text:''Advanced Document Parsing For Enterprises'' 16Sources:171. search_code_examples_jhh40p32wxpw:4\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.332Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":2558}}144{"id":"doc-querying_structured_data_tables_cohere-a1fee850","source":"documentation","title":"Querying Structured Data (Tables) | Cohere","url":"https://docs.cohere.com/docs/querying-structured-data-tables","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1! pip install cohere pandas -qq\n```\n\nExample:\n```text\n1import json2import os3import cohere45co = cohere.ClientV2(6 \"COHERE_API_KEY\"7) # Get your free API key: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1import pandas as pd23df = pd.read_csv(\"evaluation_results.csv\")45df.head()\n```\n\nExample:\n```text\n1from tool_def import (2 analyze_evaluation_results,3 analyze_evaluation_results_tool,4)\n```\n\nExample:\n```text\n1functions_map = {2 \"analyze_evaluation_results\": analyze_evaluation_results3}\n```\n\nExample:\n```text\n1analyze_evaluation_results_tool = {2 \"type\": \"function\",3 \"function\": {4 \"name\": \"analyze_evaluation_results\",5 \"description\": \"Generate Python code using the pandas library to analyze evaluation results from a dataframe called `evaluation_results`. The dataframe has columns 'usecase','run','score','temperature','tokens', and 'latency'. You must start with `import pandas as pd` and read a CSV file called `evaluation_results.csv` into the `evaluation_results` dataframe.\",6 \"parameters\": {7 \"type\": \"object\",8 \"properties\": {9 \"code\": {10 \"type\": \"string\",11 \"description\": \"Executable Python code\",12 }13 },14 \"required\": [\"code\"],15 },16 },17}\n```\n\nExample:\n```text\n1tools = [analyze_evaluation_results_tool]\n```\n\nExample:\n```text\n1system_message = \"\"\"## Task and Context2ou are an assistant who helps developers analyze LLM application evaluation results from a CSV files.\"\"\"\n```\n\nExample:\n```text\n1model = \"command-a-plus-05-2026\"234def run_agent(query, messages=None):5 if messages is None:6 messages = []78 if \"system\" not in {m.get(\"role\") for m in messages}:9 messages.append({\"role\": \"system\", \"content\": system_message})1011 # Step 1: get user message12 print(f\"Question:\\n{query}\")13 print(\"=\" * 50)1415 messages.append({\"role\": \"user\", \"content\": query})1617 # Step 2: Generate tool calls (if any)18 response = co.chat(19 model=model, messages=messages, tools=tools, temperature=0.320 )2122 while response.message.tool_calls:2324 print(\"TOOL PLAN:\")25 print(response.message.tool_plan, \"\\n\")26 print(\"TOOL CALLS:\")27 for tc in response.message.tool_calls:28 if tc.function.name == \"analyze_evaluation_results\":29 print(f\"Tool name: {tc.function.name}\")30 tool_call_prettified = print(31 \"\\n\".join(32 f\" {line}\"33 for line_num, line in enumerate(34 json.loads(tc.function.arguments)[35 \"code\"36 ].splitlines()37 )38 )39 )40 print(tool_call_prettified)41 else:42 print(43 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"44 )45 print(\"=\" * 50)4647 messages.append(response.message)4849 # Step 3: Get tool results50 for tc in response.message.tool_calls:51 tool_result = functions_map[tc.function.name](52 **json.loads(tc.function.arguments)53 )54 tool_content = [55 {56 \"type\": \"document\",57 \"document\": {\"data\": json.dumps(tool_result)},58 }59 ]6061 messages.append(62 {63 \"role\": \"tool\",64 \"tool_call_id\": tc.id,65 \"content\": tool_content,66 }67 )6869 # Step 4: Generate response and citations70 response = co.chat(71 model=model,72 messages=messages,73 tools=tools,74 temperature=0.3,75 )7677 messages.append(78 {79 \"role\": \"assistant\",80 \"content\": response.message.content[0].text,81 }82 )8384 # Print final response85 print(\"RESPONSE:\")86 print(response.message.content[0].text)87 print(\"=\" * 50)8889 # Print citations (if any)90 verbose_source = (91 False # Change to True to display the contents of a source92 )93 if response.message.citations:94 print(\"CITATIONS:\\n\")95 for citation in response.message.citations:96 print(97 f\"Start: {citation.start}| End:{citation.end}| Text:'{citation.text}' \"98 )99 print(\"Sources:\")100 for idx, source in enumerate(citation.sources):101 print(f\"{idx+1}. {source.id}\")102 if verbose_source:103 print(f\"{source.tool_output}\")104 print(\"\\n\")105106 return messages\n```\n\nExample:\n```text\n1messages = run_agent(\"What's the average evaluation score in run A\")2# Answer: 0.63\n```\n\nExample:\n```text\n1Question:2What's the average evaluation score in run A3==================================================456Python REPL can execute arbitrary code. Use with caution.789TOOL PLAN:10I will write and execute Python code to calculate the average evaluation score in run A. 1112TOOL CALLS:13Tool name: analyze_evaluation_results14 import pandas as pd15 16 df = pd.read_csv(\"evaluation_results.csv\")17 18 # Calculate the average evaluation score in run A19 average_score_run_A = df[df[\"run\"] == \"A\"][\"score\"].mean()20 21 print(f\"Average evaluation score in run A: {average_score_run_A}\")22None23==================================================24RESPONSE:25The average evaluation score in run A is 0.63.26==================================================27CITATIONS:2829Start: 41| End:46| Text:'0.63.' 30Sources:311. analyze_evaluation_results_phqpwwat2hgf:0\n```\n\nExample:\n```text\n1messages = run_agent(2 \"What's the latency of the highest-scoring run for the summarize_article use case?\"3)4# Answer: 4.8\n```\n\nExample:\n```text\n1Question:2What's the latency of the highest-scoring run for the summarize_article use case?3==================================================4TOOL PLAN:5I will write Python code to find the latency of the highest-scoring run for the summarize_article use case. 67TOOL CALLS:8Tool name: analyze_evaluation_results9 import pandas as pd10 11 df = pd.read_csv(\"evaluation_results.csv\")12 13 # Filter for the summarize_article use case14 use_case_df = df[df[\"usecase\"] == \"summarize_article\"]15 16 # Find the highest-scoring run17 highest_score_run = use_case_df.loc[use_case_df[\"score\"].idxmax()]18 19 # Get the latency of the highest-scoring run20 latency = highest_score_run[\"latency\"]21 22 print(f\"Latency of the highest-scoring run: {latency}\")23None24==================================================25RESPONSE:26The latency of the highest-scoring run for the summarize_article use case is 4.8.27==================================================28CITATIONS:2930Start: 77| End:81| Text:'4.8.' 31Sources:321. analyze_evaluation_results_es3hnnnp5pey:0\n```\n\nExample:\n```text\n1messages = run_agent(2 \"Which use case uses the least amount of tokens on average? Show the comparison of all use cases in a markdown table.\"3)4# Answer: extract_names (106.25), draft_email (245.75), summarize_article (355.75)\n```\n\nExample:\n```text\n1Question:2Which use case uses the least amount of tokens on average? Show the comparison of all use cases in a markdown table.3==================================================4TOOL PLAN:5I will use the analyze_evaluation_results tool to generate Python code to find the use case that uses the least amount of tokens on average. I will also generate code to create a markdown table to compare all use cases. 67TOOL CALLS:8Tool name: analyze_evaluation_results9 import pandas as pd10 11 evaluation_results = pd.read_csv(\"evaluation_results.csv\")12 13 # Group by 'usecase' and calculate the average tokens14 avg_tokens_by_usecase = evaluation_results.groupby('usecase')['tokens'].mean()15 16 # Find the use case with the least average tokens17 least_avg_tokens_usecase = avg_tokens_by_usecase.idxmin()18 19 print(f\"Use case with the least average tokens: {least_avg_tokens_usecase}\")20 21 # Create a markdown table comparing average tokens for all use cases22 markdown_table = avg_tokens_by_usecase.reset_index()23 markdown_table.columns = [\"Use Case\", \"Average Tokens\"]24 print(markdown_table.to_markdown(index=False))25None26==================================================27RESPONSE:28The use case that uses the least amount of tokens on average is extract_names.2930Here is a markdown table comparing the average tokens for all use cases:3132| Use Case | Average Tokens |33|:-------------------------|-------------------------------:|34| draft_email | 245.75 |35| extract_names | 106.25 |36| summarize_article | 355.75 |37==================================================38CITATIONS:3940Start: 64| End:78| Text:'extract_names.' 41Sources:421. analyze_evaluation_results_zp68h5304e3v:0434445Start: 156| End:164| Text:'Use Case' 46Sources:471. analyze_evaluation_results_zp68h5304e3v:0484950Start: 167| End:181| Text:'Average Tokens' 51Sources:521. analyze_evaluation_results_zp68h5304e3v:0535455Start: 248| End:259| Text:'draft_email' 56Sources:571. analyze_evaluation_results_zp68h5304e3v:0585960Start: 262| End:268| Text:'245.75' 61Sources:621. analyze_evaluation_results_zp68h5304e3v:0636465Start: 273| End:286| Text:'extract_names' 66Sources:671. analyze_evaluation_results_zp68h5304e3v:0686970Start: 289| End:295| Text:'106.25' 71Sources:721. analyze_evaluation_results_zp68h5304e3v:0737475Start: 300| End:317| Text:'summarize_article' 76Sources:771. analyze_evaluation_results_zp68h5304e3v:0787980Start: 320| End:326| Text:'355.75' 81Sources:821. analyze_evaluation_results_zp68h5304e3v:0\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.332Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":2546}}145{"id":"doc-building_rag_models_with_cohere_cohere-42b12d6b","source":"documentation","title":"Building RAG models with Cohere | Cohere","url":"https://docs.cohere.com/docs/rag-with-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# pip install cohere23import cohere4import numpy as np5import json6from typing import List78# Get your free API key: https://dashboard.cohere.com/api-keys9co = cohere.ClientV2(api_key=\"COHERE_API_KEY\")\n```\n\nExample:\n```text\n1documents = [2 {3 \"data\": {4 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"5 }6 },7 {8 \"data\": {9 \"text\": \"Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.\"10 }11 },12 {13 \"data\": {14 \"text\": \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\"15 }16 },17]\n```\n\nExample:\n```text\n1# Add the user query2query = \"Are there health benefits?\"34# Generate the response5response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[{\"role\": \"user\", \"content\": query}],8 documents=documents,9)1011# Display the response12print(response.message.content[0].text)1314# Display the citations and source documents15if response.message.citations:16 print(\"\\nCITATIONS:\")17 for citation in response.message.citations:18 print(citation, \"\\n\")\n```\n\nExample:\n```text\nYes, we offer gym memberships, on-site yoga classes, and comprehensive health insurance.CITATIONS:start=14 end=88 text='gym memberships, on-site yoga classes, and comprehensive health insurance.' sources=[DocumentSource(type='document', id='doc:2', document={'id': 'doc:2', 'text': 'Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.'})]\n```\n\nExample:\n```text\n1def generate_search_queries(message: str) -> List[str]:23 # Define the query generation tool4 query_gen_tool = [5 {6 \"type\": \"function\",7 \"function\": {8 \"name\": \"internet_search\",9 \"description\": \"Returns a list of relevant document snippets for a textual query retrieved from the internet\",10 \"parameters\": {11 \"type\": \"object\",12 \"properties\": {13 \"queries\": {14 \"type\": \"array\",15 \"items\": {\"type\": \"string\"},16 \"description\": \"a list of queries to search the internet with.\",17 }18 },19 \"required\": [\"queries\"],20 },21 },22 }23 ]2425 # Define a system instruction to optimize search query generation26 instructions = \"Write a search query that will find helpful information for answering the user's question accurately. If you need more than one search query, write a list of search queries. If you decide that a search is very unlikely to find information that would be useful in constructing a response to the user, you should instead directly answer.\"2728 # Generate search queries (if any)29 search_queries = []3031 res = co.chat(32 model=\"command-a-plus-05-2026\",33 messages=[34 {\"role\": \"system\", \"content\": instructions},35 {\"role\": \"user\", \"content\": message},36 ],37 tools=query_gen_tool,38 )3940 if res.message.tool_calls:41 for tc in res.message.tool_calls:42 queries = json.loads(tc.function.arguments)[\"queries\"]43 search_queries.extend(queries)4445 return search_queries\n```\n\nExample:\n```text\n1query = \"How to stay connected with the company, and do you organize team events?\"2queries_for_search = generate_search_queries(query)3print(queries_for_search)\n```\n\nExample:\n```text\n['how to stay connected with the company', 'does the company organize team events']\n```\n\nExample:\n```text\n1query = \"How flexible are the working hours\"2queries_for_search = generate_search_queries(query)3print(queries_for_search)\n```\n\nExample:\n```text\n['how flexible are the working hours at the company']\n```\n\nExample:\n```text\n1query = \"What is 2 + 2\"2queries_for_search = generate_search_queries(query)3print(queries_for_search)\n```\n\nExample:\n```text\n1# Define the documents2faqs_long = [3 {4 \"data\": {5 \"text\": \"Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.\"6 }7 },8 {9 \"data\": {10 \"text\": \"Finding Coffee Spots: For your caffeine fix, head to the break room's coffee machine or cross the street to the café for artisan coffee.\"11 }12 },13 {14 \"data\": {15 \"text\": \"Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!\"16 }17 },18 {19 \"data\": {20 \"text\": \"Working Hours Flexibility: We prioritize work-life balance. While our core hours are 9 AM to 5 PM, we offer flexibility to adjust as needed.\"21 }22 },23 {24 \"data\": {25 \"text\": \"Side Projects Policy: We encourage you to pursue your passions. Just be mindful of any potential conflicts of interest with our business.\"26 }27 },28 {29 \"data\": {30 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"31 }32 },33 {34 \"data\": {35 \"text\": \"Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.\"36 }37 },38 {39 \"data\": {40 \"text\": \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\"41 }42 },43 {44 \"data\": {45 \"text\": \"Performance Reviews Frequency: We conduct informal check-ins every quarter and formal performance reviews twice a year.\"46 }47 },48 {49 \"data\": {50 \"text\": \"Proposing New Ideas: Innovation is welcomed! Share your brilliant ideas at our weekly team meetings or directly with your team lead.\"51 }52 },53]5455# Embed the documents56doc_emb = co.embed(57 model=\"embed-v4.0\",58 input_type=\"search_document\",59 texts=[doc[\"data\"][\"text\"] for doc in faqs_long],60 embedding_types=[\"float\"],61).embeddings.float\n```\n\nExample:\n```text\n1# Add the user query2query = \"How to get to know my teammates\"34# Generate the search query5# Note: For simplicity, we are assuming only one query generated. For actual implementations, you will need to perform search for each query.6queries_for_search = generate_search_queries(query)[0]7print(\"Search query: \", queries_for_search)89# Embed the search query10query_emb = co.embed(11 model=\"embed-v4.0\",12 input_type=\"search_query\",13 texts=[queries_for_search],14 embedding_types=[\"float\"],15).embeddings.float\n```\n\nExample:\n```text\nSearch query: how to get to know teammates\n```\n\nExample:\n```text\n1# Compute dot product similarity and display results2n = 53scores = np.dot(query_emb, np.transpose(doc_emb))[0]4max_idx = np.argsort(-scores)[:n]56retrieved_documents = [faqs_long[item] for item in max_idx]78for rank, idx in enumerate(max_idx):9 print(f\"Rank: {rank+1}\")10 print(f\"Score: {scores[idx]}\")11 print(f\"Document: {retrieved_documents[rank]}\\n\")\n```\n\nExample:\n```text\nRank: 1Score: 0.34212792245283796Document: {'data': {'text': 'Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!'}}Rank: 2Score: 0.2883222063024371Document: {'data': {'text': 'Proposing New Ideas: Innovation is welcomed! Share your brilliant ideas at our weekly team meetings or directly with your team lead.'}}Rank: 3Score: 0.278128283997032Document: {'data': {'text': 'Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.'}}Rank: 4Score: 0.19474858706643985Document: {'data': {'text': \"Finding Coffee Spots: For your caffeine fix, head to the break room's coffee machine or cross the street to the café for artisan coffee.\"}}Rank: 5Score: 0.13713692506528824Document: {'data': {'text': 'Side Projects Policy: We encourage you to pursue your passions. Just be mindful of any potential conflicts of interest with our business.'}}\n```\n\nExample:\n```text\n1# Rerank the documents2results = co.rerank(3 query=queries_for_search,4 documents=[doc[\"data\"][\"text\"] for doc in retrieved_documents],5 top_n=2,6 model=\"rerank-english-v3.0\",7)89# Display the reranking results10for idx, result in enumerate(results.results):11 print(f\"Rank: {idx+1}\")12 print(f\"Score: {result.relevance_score}\")13 print(f\"Document: {retrieved_documents[result.index]}\\n\")1415reranked_documents = [16 retrieved_documents[result.index] for result in results.results17]\n```\n\nExample:\n```text\nRank: 1Score: 0.0020507434Document: {'data': {'text': 'Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.'}}Rank: 2Score: 0.0014158706Document: {'data': {'text': 'Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!'}}\n```\n\nExample:\n```text\n1# Generate the response2response = co.chat(3 model=\"command-a-plus-05-2026\",4 messages=[{\"role\": \"user\", \"content\": query}],5 documents=reranked_documents,6)78# Display the response9print(response.message.content[0].text)1011# Display the citations and source documents12if response.message.citations:13 print(\"\\nCITATIONS:\")14 for citation in response.message.citations:15 print(citation, \"\\n\")\n```\n\nExample:\n```text\nYou can get to know your teammates by joining relevant Slack channels and engaging in team-building activities. These activities include monthly outings and weekly game nights. You are also welcome to suggest new activity ideas.CITATIONS:start=38 end=69 text='joining relevant Slack channels' sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Joining Slack Channels: You will receive an invite via email. Be sure to join relevant channels to stay informed and engaged.'})] start=86 end=111 text='team-building activities.' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!'})] start=137 end=176 text='monthly outings and weekly game nights.' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!'})] start=201 end=228 text='suggest new activity ideas.' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Team-Building Activities: We foster team spirit with monthly outings and weekly game nights. Feel free to suggest new activity ideas anytime!'})]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.334Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":19,"totalLines":98,"estimatedTokens":2912}}146{"id":"doc-delete_a_dataset_cohere-a93f5c51","source":"documentation","title":"Delete a Dataset | Cohere","url":"https://docs.cohere.com/reference/delete-dataset","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45# delete dataset6co.datasets.delete(\"id\")\n```\n\nExample:\n```text\n1{}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.334Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":78}}147{"id":"doc-cancel_an_embed_job_cohere-81cda228","source":"documentation","title":"Cancel an Embed Job | Cohere","url":"https://docs.cohere.com/reference/cancel-embed-job","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45# cancel an embed job6co.embed_jobs.cancel(\"job_id\")\n```\n\nExample:\n```text\n1{}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.334Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":81}}148{"id":"doc-building_a_generative_ai_agent_with_cohere_coher-4ad38a9b","source":"documentation","title":"Building a Generative AI Agent with Cohere | Cohere","url":"https://docs.cohere.com/docs/building-an-agent-with-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# pip install cohere23import cohere4import json56# Get your free API key: https://dashboard.cohere.com/api-keys7co = cohere.ClientV2(api_key=\"COHERE_API_KEY\")\n```\n\nExample:\n```text\n1# Create the tools2def search_faqs(query):3 faqs = [4 {5 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"6 },7 {8 \"text\": \"Working from Abroad: Working remotely from another country is possible. Simply coordinate with your manager and ensure your availability during core hours.\"9 },10 ]11 return faqs121314def search_emails(query):15 emails = [16 {17 \"from\": \"it@co1t.com\",18 \"to\": \"david@co1t.com\",19 \"date\": \"2024-06-24\",20 \"subject\": \"Setting Up Your IT Needs\",21 \"text\": \"Greetings! To ensure a seamless start, please refer to the attached comprehensive guide, which will assist you in setting up all your work accounts.\",22 },23 {24 \"from\": \"john@co1t.com\",25 \"to\": \"david@co1t.com\",26 \"date\": \"2024-06-24\",27 \"subject\": \"First Week Check-In\",28 \"text\": \"Hello! I hope you're settling in well. Let's connect briefly tomorrow to discuss how your first week has been going. Also, make sure to join us for a welcoming lunch this Thursday at noon—it's a great opportunity to get to know your colleagues!\",29 },30 ]31 return emails323334def create_calendar_event(date: str, time: str, duration: int):35 # You can implement any logic here36 return {37 \"is_success\": True,38 \"message\": f\"Created a {duration} hour long event at {time} on {date}\",39 }404142functions_map = {43 \"search_faqs\": search_faqs,44 \"search_emails\": search_emails,45 \"create_calendar_event\": create_calendar_event,46}\n```\n\nExample:\n```text\n1# Define the tools2tools = [3 {4 \"type\": \"function\",5 \"function\": {6 \"name\": \"search_faqs\",7 \"description\": \"Given a user query, searches a company's frequently asked questions (FAQs) list and returns the most relevant matches to the query.\",8 \"parameters\": {9 \"type\": \"object\",10 \"properties\": {11 \"query\": {12 \"type\": \"string\",13 \"description\": \"The query from the user\",14 }15 },16 \"required\": [\"query\"],17 },18 },19 },20 {21 \"type\": \"function\",22 \"function\": {23 \"name\": \"search_emails\",24 \"description\": \"Given a user query, searches a person's emails and returns the most relevant matches to the query.\",25 \"parameters\": {26 \"type\": \"object\",27 \"properties\": {28 \"query\": {29 \"type\": \"string\",30 \"description\": \"The query from the user\",31 }32 },33 \"required\": [\"query\"],34 },35 },36 },37 {38 \"type\": \"function\",39 \"function\": {40 \"name\": \"create_calendar_event\",41 \"description\": \"Creates a new calendar event of the specified duration at the specified time and date. A new event cannot be created on the same time as an existing event.\",42 \"parameters\": {43 \"type\": \"object\",44 \"properties\": {45 \"date\": {46 \"type\": \"string\",47 \"description\": \"the date on which the event starts, formatted as mm/dd/yy\",48 },49 \"time\": {50 \"type\": \"string\",51 \"description\": \"the time of the event, formatted using 24h military time formatting\",52 },53 \"duration\": {54 \"type\": \"number\",55 \"description\": \"the number of hours the event lasts for\",56 },57 },58 \"required\": [\"date\", \"time\", \"duration\"],59 },60 },61 },62]\n```\n\nExample:\n```text\n1# Create custom system message2system_message = \"\"\"## Task and Context3You are an assistant who assist new employees of Co1t with their first week. You respond to their questions and assist them with their needs. Today is Monday, June 24, 2024\"\"\"456# Step 1: Get user message7message = \"Is there any message about getting setup with IT?\"89# Add the system and user messages to the chat history10messages = [11 {\"role\": \"system\", \"content\": system_message},12 {\"role\": \"user\", \"content\": message},13]1415# Step 2: Tool planning and calling16response = co.chat(17 model=\"command-a-plus-05-2026\", messages=messages, tools=tools18)1920if response.message.tool_calls:21 print(\"Tool plan:\")22 print(response.message.tool_plan, \"\\n\")23 print(\"Tool calls:\")24 for tc in response.message.tool_calls:25 print(26 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"27 )2829 # Append tool calling details to the chat history30 messages.append(response.message)\n```\n\nExample:\n```text\nTool plan:I will search the user's emails for any messages about getting set up with IT. Tool calls:Tool name: search_emails | Parameters: {\"query\":\"IT setup\"}\n```\n\nExample:\n```text\n1# Step 3: Tool execution2for tc in response.message.tool_calls:3 tool_result = functions_map[tc.function.name](4 **json.loads(tc.function.arguments)5 )6 tool_content = []7 for data in tool_result:8 tool_content.append(9 {10 \"type\": \"document\",11 \"document\": {\"data\": json.dumps(data)},12 }13 )14 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated15 # Append tool results to the chat history16 messages.append(17 {18 \"role\": \"tool\",19 \"tool_call_id\": tc.id,20 \"content\": tool_content,21 }22 )2324 print(\"Tool results:\")25 for result in tool_content:26 print(result)\n```\n\nExample:\n```text\nTool results:{'type': 'document', 'document': {'data': '{\"from\": \"it@co1t.com\", \"to\": \"david@co1t.com\", \"date\": \"2024-06-24\", \"subject\": \"Setting Up Your IT Needs\", \"text\": \"Greetings! To ensure a seamless start, please refer to the attached comprehensive guide, which will assist you in setting up all your work accounts.\"}'}}{'type': 'document', 'document': {'data': '{\"from\": \"john@co1t.com\", \"to\": \"david@co1t.com\", \"date\": \"2024-06-24\", \"subject\": \"First Week Check-In\", \"text\": \"Hello! I hope you\\'re settling in well. Let\\'s connect briefly tomorrow to discuss how your first week has been going. Also, make sure to join us for a welcoming lunch this Thursday at noon\\\\u2014it\\'s a great opportunity to get to know your colleagues!\"}'}}\n```\n\nExample:\n```text\n1# Step 4: Response and citation generation2response = co.chat(3 model=\"command-a-plus-05-2026\", messages=messages, tools=tools4)56# Append assistant response to the chat history7messages.append(8 {\"role\": \"assistant\", \"content\": response.message.content[0].text}9)1011# Print final response12print(\"Response:\")13print(response.message.content[0].text)14print(\"=\" * 50)1516# Print citations (if any)17if response.message.citations:18 print(\"\\nCITATIONS:\")19 for citation in response.message.citations:20 print(citation, \"\\n\")\n```\n\nExample:\n```text\nResponse:Yes, there is an email from it@co1t.com with the subject 'Setting Up Your IT Needs'. It includes an attached guide to help you set up your work accounts.==================================================CITATIONS:start=17 end=83 text=\"email from it@co1t.com with the subject 'Setting Up Your IT Needs'\" sources=[ToolSource(type='tool', id='search_emails_wqs498sp2d07:0', tool_output={'date': '2024-06-24', 'from': 'it@co1t.com', 'subject': 'Setting Up Your IT Needs', 'text': 'Greetings! To ensure a seamless start, please refer to the attached comprehensive guide, which will assist you in setting up all your work accounts.', 'to': 'david@co1t.com'})] start=100 end=153 text='attached guide to help you set up your work accounts.' sources=[ToolSource(type='tool', id='search_emails_wqs498sp2d07:0', tool_output={'date': '2024-06-24', 'from': 'it@co1t.com', 'subject': 'Setting Up Your IT Needs', 'text': 'Greetings! To ensure a seamless start, please refer to the attached comprehensive guide, which will assist you in setting up all your work accounts.', 'to': 'david@co1t.com'})]\n```\n\nExample:\n```text\n1model = \"command-a-plus-05-2026\"23system_message = \"\"\"## Task and Context4You are an assistant who assists new employees of Co1t with their first week. You respond to their questions and assist them with their needs. Today is Monday, June 24, 2024\"\"\"567def run_assistant(query, messages=None):8 if messages is None:9 messages = []1011 if \"system\" not in {m.get(\"role\") for m in messages}:12 messages.append({\"role\": \"system\", \"content\": system_message})1314 # Step 1: get user message15 print(f\"Question:\\n{query}\")16 print(\"=\" * 50)1718 messages.append({\"role\": \"user\", \"content\": query})1920 # Step 2: Generate tool calls (if any)21 response = co.chat(model=model, messages=messages, tools=tools)2223 while response.message.tool_calls:2425 print(\"Tool plan:\")26 print(response.message.tool_plan, \"\\n\")27 print(\"Tool calls:\")28 for tc in response.message.tool_calls:29 print(30 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"31 )32 print(\"=\" * 50)3334 messages.append(response.message)3536 # Step 3: Get tool results37 for idx, tc in enumerate(response.message.tool_calls):38 tool_result = functions_map[tc.function.name](39 **json.loads(tc.function.arguments)40 )41 tool_content = []42 for data in tool_result:43 tool_content.append(44 {45 \"type\": \"document\",46 \"document\": {\"data\": json.dumps(data)},47 }48 )49 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated50 messages.append(51 {52 \"role\": \"tool\",53 \"tool_call_id\": tc.id,54 \"content\": tool_content,55 }56 )5758 # Step 4: Generate response and citations59 response = co.chat(60 model=model, messages=messages, tools=tools61 )6263 messages.append(64 {65 \"role\": \"assistant\",66 \"content\": response.message.content[0].text,67 }68 )6970 # Print final response71 print(\"Response:\")72 print(response.message.content[0].text)73 print(\"=\" * 50)7475 # Print citations (if any)76 if response.message.citations:77 print(\"\\nCITATIONS:\")78 for citation in response.message.citations:79 print(citation, \"\\n\")8081 return messages\n```\n\nExample:\n```text\n1messages = run_assistant(2 \"Can you check if there are any lunch invites, and for those days, create a one-hour event on my calendar at 12PM.\"3)\n```\n\nExample:\n```text\nQuestion:Can you check if there are any lunch invites, and for those days, create a one-hour event on my calendar at 12PM.==================================================Tool plan:I will first search the user's emails for lunch invites. Then, I will create a one-hour event on the user's calendar at 12PM for each day that the user has a lunch invite. Tool calls:Tool name: search_emails | Parameters: {\"query\":\"lunch invites\"}==================================================Tool plan:I have found one lunch invite for Thursday at noon. I will now create a one-hour event on the user's calendar for Thursday at noon. Tool calls:Tool name: create_calendar_event | Parameters: {\"date\":\"06/27/24\",\"duration\":1,\"time\":\"12:00\"}==================================================Response:I found one lunch invite for Thursday, June 27, 2024. I have created a one-hour event on your calendar for that day at 12pm.==================================================CITATIONS:start=29 end=53 text='Thursday, June 27, 2024.' sources=[ToolSource(type='tool', id='search_emails_1dxqzwragh9g:1', tool_output={'date': '2024-06-24', 'from': 'john@co1t.com', 'subject': 'First Week Check-In', 'text': \"Hello! I hope you're settling in well. Let's connect briefly tomorrow to discuss how your first week has been going. Also, make sure to join us for a welcoming lunch this Thursday at noon—it's a great opportunity to get to know your colleagues!\", 'to': 'david@co1t.com'})] start=71 end=85 text='one-hour event' sources=[ToolSource(type='tool', id='create_calendar_event_w11caj6hmqz2:0', tool_output={'content': '\"is_success\"'})] start=119 end=124 text='12pm.' sources=[ToolSource(type='tool', id='search_emails_1dxqzwragh9g:1', tool_output={'date': '2024-06-24', 'from': 'john@co1t.com', 'subject': 'First Week Check-In', 'text': \"Hello! I hope you're settling in well. Let's connect briefly tomorrow to discuss how your first week has been going. Also, make sure to join us for a welcoming lunch this Thursday at noon—it's a great opportunity to get to know your colleagues!\", 'to': 'david@co1t.com'})]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.335Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":3457}}149{"id":"doc-get_dataset_usage_cohere-ac2077d6","source":"documentation","title":"Get Dataset Usage | Cohere","url":"https://docs.cohere.com/reference/get-dataset-usage","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45# get usage6response = co.datasets.get_usage()78print(response)\n```\n\nExample:\n```text\n1{2 \"organization_usage\": 52428800003}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.335Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":92}}150{"id":"doc-tokenize_cohere-547d21f3","source":"documentation","title":"Tokenize | Cohere","url":"https://docs.cohere.com/reference/tokenize","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45response = co.tokenize(6 text=\"tokenize me! :D\", model=\"command-a-03-2025\"7) # optional8print(response)\n```\n\nExample:\n```text\n1{2 \"tokens\": [3 10002,4 2261,5 2012,6 8,7 2792,8 439 ],10 \"token_strings\": [11 \"token\",12 \"ize\",13 \" me\",14 \"!\",15 \" :\",16 \"D\"17 ],18 \"meta\": {19 \"api_version\": {20 \"version\": \"1\"21 }22 }23}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.336Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":155}}151{"id":"doc-querying_structured_data_sql_cohere-3e54abfe","source":"documentation","title":"Querying Structured Data (SQL) | Cohere","url":"https://docs.cohere.com/docs/querying-structured-data-sql","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1! pip install cohere pandas -qq\n```\n\nExample:\n```text\n1import json2import os3import cohere4import sqlite35import pandas as pd67co = cohere.ClientV2(8 \"COHERE_API_KEY\"9) # Get your free API key: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1# Create a connection to a new SQLite database (or connect to an existing one)2conn = sqlite3.connect(\"evaluation_results.db\")3cursor = conn.cursor()45# Execute the CREATE TABLE command6cursor.execute(7 \"\"\"8CREATE TABLE evaluation_results (9 usecase TEXT,10 run TEXT,11 score FLOAT,12 temperature FLOAT,13 tokens INTEGER,14 latency FLOAT15)16\"\"\"17)1819# Execute the INSERT commands20data = [21 (\"extract_names\", \"A\", 0.5, 0.3, 103, 1.12),22 (\"draft_email\", \"A\", 0.6, 0.3, 252, 2.5),23 (\"summarize_article\", \"A\", 0.8, 0.3, 350, 4.2),24 (\"extract_names\", \"B\", 0.2, 0.3, 101, 2.85),25 (\"draft_email\", \"B\", 0.4, 0.3, 230, 3.2),26 (\"summarize_article\", \"B\", 0.6, 0.3, 370, 4.2),27 (\"extract_names\", \"C\", 0.7, 0.3, 101, 2.22),28 (\"draft_email\", \"C\", 0.5, 0.3, 221, 2.5),29 (\"summarize_article\", \"C\", 0.1, 0.3, 361, 3.9),30 (\"extract_names\", \"D\", 0.7, 0.5, 120, 3.2),31 (\"draft_email\", \"D\", 0.8, 0.5, 280, 3.4),32 (\"summarize_article\", \"D\", 0.9, 0.5, 342, 4.8),33]3435cursor.executemany(36 \"INSERT INTO evaluation_results VALUES (?,?,?,?,?,?)\", data37)3839# Commit the changes and close the connection40conn.commit()41conn.close()\n```\n\nExample:\n```text\n1def sql_table_query(query: str) -> dict:2 \"\"\"3 Execute an SQL query on the evaluation_results table and return the result as a dictionary.45 Args:6 query (str): SQL query to execute on the evaluation_results table78 Returns:9 dict: Result of the SQL query10 \"\"\"11 try:12 # Connect to the SQLite database13 conn = sqlite3.connect(\"evaluation_results.db\")1415 # Execute the query and fetch the results into a DataFrame16 df = pd.read_sql_query(query, conn)1718 # Close the connection19 conn.close()2021 # Convert DataFrame to dictionary22 result_dict = df.to_dict(orient=\"records\")2324 return result_dict2526 except sqlite3.Error as e:27 print(f\"An error occurred: {e}\")28 return str(e)29 except Exception as e:30 print(f\"An unexpected error occurred: {e}\")31 return str(e)323334functions_map = {\"sql_table_query\": sql_table_query}\n```\n\nExample:\n```text\n1result = sql_table_query(2 \"SELECT * FROM evaluation_results WHERE usecase = 'extract_names'\"3)4print(result)\n```\n\nExample:\n```text\n1[{'usecase': 'extract_names', 'run': 'A', 'score': 0.5, 'temperature': 0.3, 'tokens': 103, 'latency': 1.12}, {'usecase': 'extract_names', 'run': 'B', 'score': 0.2, 'temperature': 0.3, 'tokens': 101, 'latency': 2.85}, {'usecase': 'extract_names', 'run': 'C', 'score': 0.7, 'temperature': 0.3, 'tokens': 101, 'latency': 2.22}, {'usecase': 'extract_names', 'run': 'D', 'score': 0.7, 'temperature': 0.5, 'tokens': 120, 'latency': 3.2}]\n```\n\nExample:\n```text\n1sql_table_query_tool = {2 \"type\": \"function\",3 \"function\": {4 \"name\": \"sql_table_query\",5 \"description\": \"Execute an SQL query on the evaluation_results table in the SQLite database. The table has columns 'usecase', 'run', 'score', 'temperature', 'tokens', and 'latency'.\",6 \"parameters\": {7 \"type\": \"object\",8 \"properties\": {9 \"query\": {10 \"type\": \"string\",11 \"description\": \"SQL query to execute on the evaluation_results table\",12 }13 },14 \"required\": [\"query\"],15 },16 },17}1819tools = [sql_table_query_tool]\n```\n\nExample:\n```text\n1system_message = \"\"\"## Task and Context2You are an assistant who helps developers analyze LLM application evaluation results from a SQLite database. The database contains a table named 'evaluation_results' with the following schema:34- usecase (TEXT): The type of task being evaluated5- run (TEXT): The identifier for a specific evaluation run6- score (REAL): The performance score of the run7- temperature (REAL): The temperature setting used for the LLM8- tokens (INTEGER): The number of tokens used in the run9- latency (REAL): The time taken for the run in seconds1011You can use SQL queries to analyze this data and provide insights to the developers.\"\"\"\n```\n\nExample:\n```text\n1model = \"command-a-plus-05-2026\"234def run_agent(query, messages=None):5 if messages is None:6 messages = []78 if \"system\" not in {m.get(\"role\") for m in messages}:9 messages.append({\"role\": \"system\", \"content\": system_message})1011 # Step 1: get user message12 print(f\"Question:\\n{query}\")13 print(\"=\" * 50)1415 messages.append({\"role\": \"user\", \"content\": query})1617 # Step 2: Generate tool calls (if any)18 response = co.chat(19 model=model, messages=messages, tools=tools, temperature=0.320 )2122 while response.message.tool_calls:2324 print(\"Tool plan:\")25 print(response.message.tool_plan, \"\\n\")26 print(\"Tool calls:\")27 for tc in response.message.tool_calls:28 # print(f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\")29 if tc.function.name == \"analyze_evaluation_results\":30 print(f\"Tool name: {tc.function.name}\")31 tool_call_prettified = print(32 \"\\n\".join(33 f\" {line}\"34 for line_num, line in enumerate(35 json.loads(tc.function.arguments)[36 \"code\"37 ].splitlines()38 )39 )40 )41 print(tool_call_prettified)42 else:43 print(44 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"45 )46 print(\"=\" * 50)4748 messages.append(response.message)4950 # Step 3: Get tool results51 for tc in response.message.tool_calls:52 tool_result = functions_map[tc.function.name](53 **json.loads(tc.function.arguments)54 )55 tool_content = [56 {57 \"type\": \"document\",58 \"document\": {\"data\": json.dumps(tool_result)},59 }60 ]6162 messages.append(63 {64 \"role\": \"tool\",65 \"tool_call_id\": tc.id,66 \"content\": tool_content,67 }68 )6970 # Step 4: Generate response and citations71 response = co.chat(72 model=model,73 messages=messages,74 tools=tools,75 temperature=0.3,76 )7778 messages.append(79 {80 \"role\": \"assistant\",81 \"content\": response.message.content[0].text,82 }83 )8485 # Print final response86 print(\"Response:\")87 print(response.message.content[0].text)88 print(\"=\" * 50)8990 # Print citations (if any)91 verbose_source = (92 False # Change to True to display the contents of a source93 )94 if response.message.citations:95 print(\"CITATIONS:\\n\")96 for citation in response.message.citations:97 print(98 f\"Start: {citation.start}| End:{citation.end}| Text:'{citation.text}' \"99 )100 print(\"Sources:\")101 for idx, source in enumerate(citation.sources):102 print(f\"{idx+1}. {source.id}\")103 if verbose_source:104 print(f\"{source.tool_output}\")105 print(\"\\n\")106107 return messages\n```\n\nExample:\n```text\n1messages = run_agent(\"What's the average evaluation score in run A\")2# Answer: 0.63\n```\n\nExample:\n```text\n1Question:2What's the average evaluation score in run A3==================================================4Tool plan:5I will query the connected SQL database to find the average evaluation score in run A. 67Tool calls:8Tool name: sql_table_query | Parameters: {\"query\":\"SELECT AVG(score) AS average_score\\r\\nFROM evaluation_results\\r\\nWHERE run = 'A';\"}9==================================================10Response:11The average evaluation score in run A is 0.63.12==================================================13CITATIONS:1415Start: 41| End:46| Text:'0.63.' 16Sources:171. sql_table_query_97h16txpbeqs:0\n```\n\nExample:\n```text\n1messages = run_agent(2 \"What's the latency of the highest-scoring run for the summarize_article use case?\"3)4# Answer: 4.8\n```\n\nExample:\n```text\n1Question:2What's the latency of the highest-scoring run for the summarize_article use case?3==================================================4Tool plan:5I will query the connected SQL database to find the latency of the highest-scoring run for the summarize_article use case.67I will filter the data for the summarize_article use case and order the results by score in descending order. I will then return the latency of the first result. 89Tool calls:10Tool name: sql_table_query | Parameters: {\"query\":\"SELECT latency\\r\\nFROM evaluation_results\\r\\nWHERE usecase = 'summarize_article'\\r\\nORDER BY score DESC\\r\\nLIMIT 1;\"}11==================================================12Response:13The latency of the highest-scoring run for the summarize_article use case is 4.8.14==================================================15CITATIONS:1617Start: 77| End:81| Text:'4.8.' 18Sources:191. sql_table_query_ekswkn14ra34:0\n```\n\nExample:\n```text\n1messages = run_agent(2 \"Which use case uses the least amount of tokens on average? Show the comparison of all use cases in a markdown table.\"3)4# Answer: extract_names (106.25), draft_email (245.75), summarize_article (355.75)\n```\n\nExample:\n```text\n1Question:2Which use case uses the least amount of tokens on average? Show the comparison of all use cases in a markdown table.3==================================================4Tool plan:5I will query the connected SQL database to find the average number of tokens used for each use case. I will then present this information in a markdown table. 67Tool calls:8Tool name: sql_table_query | Parameters: {\"query\":\"SELECT usecase, AVG(tokens) AS avg_tokens\\nFROM evaluation_results\\nGROUP BY usecase\\nORDER BY avg_tokens ASC;\"}9==================================================10Response:11Here is a markdown table showing the average number of tokens used for each use case:1213| Use Case | Average Tokens |14|---|---|15| extract_names | 106.25 |16| draft_email | 245.75 |17| summarize_article | 355.75 |1819The use case that uses the least amount of tokens on average is **extract_names**.20==================================================21CITATIONS:2223Start: 129| End:142| Text:'extract_names' 24Sources:251. sql_table_query_50yjx2cecqx1:0262728Start: 145| End:151| Text:'106.25' 29Sources:301. sql_table_query_50yjx2cecqx1:0313233Start: 156| End:167| Text:'draft_email' 34Sources:351. sql_table_query_50yjx2cecqx1:0363738Start: 170| End:176| Text:'245.75' 39Sources:401. sql_table_query_50yjx2cecqx1:0414243Start: 181| End:198| Text:'summarize_article' 44Sources:451. sql_table_query_50yjx2cecqx1:0464748Start: 201| End:207| Text:'355.75' 49Sources:501. sql_table_query_50yjx2cecqx1:0515253Start: 277| End:290| Text:'extract_names' 54Sources:551. sql_table_query_50yjx2cecqx1:0\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.336Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":2916}}152{"id":"doc-list_embed_jobs_cohere-771cdadc","source":"documentation","title":"List Embed Jobs | Cohere","url":"https://docs.cohere.com/reference/list-embed-jobs","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45# list embed jobs6response = co.embed_jobs.list()78print(response)\n```\n\nExample:\n```text\n1{2 \"embed_jobs\": [3 {4 \"job_id\": \"e7a1f3b2-4c9d-4f8a-9b2e-3d5f7a1c2b4e\",5 \"status\": \"processing\",6 \"created_at\": \"2024-01-15T09:30:00Z\",7 \"input_dataset_id\": \"dataset_987654321\",8 \"model\": \"embed-multilingual-v2.0\",9 \"truncate\": \"START\",10 \"name\": \"User123 Text Embedding Job\",11 \"output_dataset_id\": \"dataset_123456789\",12 \"meta\": {13 \"api_version\": {14 \"version\": \"1.0.0\",15 \"is_deprecated\": false,16 \"is_experimental\": false17 },18 \"billed_units\": {19 \"images\": 0,20 \"input_tokens\": 1500,21 \"image_tokens\": 0,22 \"output_tokens\": 1536,23 \"search_units\": 0,24 \"classifications\": 025 },26 \"tokens\": {27 \"input_tokens\": 1500,28 \"output_tokens\": 153629 },30 \"cached_tokens\": 0,31 \"warnings\": []32 }33 }34 ]35}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.336Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":314}}153{"id":"doc-create_a_transcription_cohere-38b54d0e","source":"documentation","title":"Create a transcription | Cohere","url":"https://docs.cohere.com/reference/create-audio-transcription","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nThe language of the input audio, supplied in ISO-639-1 format.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45transcription = co.audio.transcriptions.create(6 model=\"cohere-transcribe-03-2026\",7 language=\"en\",8 file=open(\"./sample.wav\", \"rb\"),9)1011print(transcription)\n```\n\nExample:\n```text\n1{2 \"text\": \"Hello, this is a sample transcription of the audio file.\"3}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.336Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":144}}154{"id":"doc-errors_status_codes_and_description_cohere-d5ca0acf","source":"documentation","title":"Errors (status codes and description) | Cohere","url":"https://docs.cohere.com/reference/errors","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.337Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}155{"id":"doc-fetch_an_embed_job_cohere-75ba661e","source":"documentation","title":"Fetch an Embed Job | Cohere","url":"https://docs.cohere.com/reference/get-embed-job","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45# get embed job6response = co.embed_jobs.get(\"job_id\")78print(response)\n```\n\nExample:\n```text\n1{2 \"job_id\": \"e3f1c9a2-7b4d-4f3a-9c2e-1a2b3c4d5e6f\",3 \"status\": \"processing\",4 \"created_at\": \"2024-01-15T09:30:00Z\",5 \"input_dataset_id\": \"dataset-7890abcde123\",6 \"model\": \"embed-multilingual-v2.0\",7 \"truncate\": \"START\",8 \"name\": \"User Text Embedding Job April 2024\",9 \"output_dataset_id\": \"dataset-4567fghij890\",10 \"meta\": {11 \"api_version\": {12 \"version\": \"1.0.0\",13 \"is_deprecated\": false,14 \"is_experimental\": false15 },16 \"billed_units\": {17 \"images\": 0,18 \"input_tokens\": 1250,19 \"image_tokens\": 0,20 \"output_tokens\": 1250,21 \"search_units\": 0,22 \"classifications\": 023 },24 \"tokens\": {25 \"input_tokens\": 1250,26 \"output_tokens\": 125027 },28 \"cached_tokens\": 300,29 \"warnings\": [30 \"Input text truncated due to length limit.\"31 ]32 }33}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.337Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":295}}156{"id":"doc-create_an_embed_job_cohere-734f4803","source":"documentation","title":"Create an Embed Job | Cohere","url":"https://docs.cohere.com/reference/create-embed-job","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nThis API launches an async Embed job for a Dataset of type embed-input. The result of a completed embed job is new Dataset of type embed-output, which contains the original text entries and the corresponding embeddings.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45# start an embed job6job = co.embed_jobs.create(7 dataset_id=\"my-dataset-id\", input_type=\"search_document\", model=\"embed-english-v3.0\"8)910# poll the server until the job is complete11response = co.wait(job)1213print(response)\n```\n\nExample:\n```text\n1{2 \"job_id\": \"job-67890abcdef\",3 \"meta\": {4 \"api_version\": {5 \"version\": \"1.0.0\",6 \"is_deprecated\": false,7 \"is_experimental\": false8 },9 \"billed_units\": {10 \"images\": 0,11 \"input_tokens\": 1500,12 \"image_tokens\": 0,13 \"output_tokens\": 1024,14 \"search_units\": 0,15 \"classifications\": 016 },17 \"tokens\": {18 \"input_tokens\": 1500,19 \"output_tokens\": 102420 },21 \"cached_tokens\": 0,22 \"warnings\": []23 }24}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.337Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":301}}157{"id":"doc-get_a_dataset_cohere-8eb8b4fd","source":"documentation","title":"Get a Dataset | Cohere","url":"https://docs.cohere.com/reference/get-dataset","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nRetrieve a dataset by ID. See ‘Datasets’ for more information.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45# get dataset6response = co.datasets.get(id=\"<<datasetId>>\")78print(response)\n```\n\nExample:\n```text\n1{2 \"dataset\": {3 \"id\": \"dset-9f8b7c6a5e4d3f21\",4 \"name\": \"Customer Support Chat Logs\",5 \"created_at\": \"2024-01-15T09:30:00Z\",6 \"updated_at\": \"2024-04-10T12:45:00Z\",7 \"dataset_type\": \"chat-finetune-input\",8 \"validation_status\": \"validated\",9 \"validation_error\": \"\",10 \"schema\": \"{\\\"type\\\":\\\"record\\\",\\\"name\\\":\\\"ChatExample\\\",\\\"fields\\\":[{\\\"name\\\":\\\"context\\\",\\\"type\\\":\\\"string\\\"},{\\\"name\\\":\\\"response\\\",\\\"type\\\":\\\"string\\\"}]}\",11 \"required_fields\": [12 \"context\",13 \"response\"14 ],15 \"preserve_fields\": [16 \"metadata\",17 \"timestamp\"18 ],19 \"dataset_parts\": [20 {21 \"id\": \"part-001\",22 \"name\": \"chat_logs_jan.csv\",23 \"url\": \"https://storage.cohere.com/datasets/dset-9f8b7c6a5e4d3f21/part-001.csv\",24 \"index\": 0,25 \"size_bytes\": 2048576,26 \"num_rows\": 15000,27 \"original_url\": \"https://originalsource.com/chat_logs_jan.csv\",28 \"samples\": [29 \"{\\\"context\\\":\\\"Hello, I need help with my order.\\\",\\\"response\\\":\\\"Sure, can you provide your order ID?\\\"}\",30 \"{\\\"context\\\":\\\"My internet is not working.\\\",\\\"response\\\":\\\"Have you tried restarting your router?\\\"}\"31 ]32 }33 ],34 \"validation_warnings\": [35 \"Some rows contain missing timestamps\",36 \"Detected inconsistent newline characters\"37 ],38 \"parse_info\": {},39 \"metrics\": {40 \"finetune_dataset_metrics\": {41 \"trainable_token_count\": 1250000,42 \"total_examples\": 15000,43 \"train_examples\": 12000,44 \"train_size_bytes\": 1800000,45 \"eval_examples\": 3000,46 \"eval_size_bytes\": 450000,47 \"reranker_data_metrics\": {48 \"num_train_queries\": 0,49 \"num_train_relevant_passages\": 0,50 \"num_train_hard_negatives\": 0,51 \"num_eval_queries\": 0,52 \"num_eval_relevant_passages\": 0,53 \"num_eval_hard_negatives\": 054 },55 \"chat_data_metrics\": {56 \"num_train_turns\": 48000,57 \"num_eval_turns\": 12000,58 \"preamble\": \"Customer support chat logs for fine-tuning chat models.\"59 },60 \"classify_data_metrics\": {61 \"label_metrics\": [62 {63 \"total_examples\": 15000,64 \"label\": \"support\",65 \"samples\": [66 \"How do I reset my password?\",67 \"My shipment is delayed.\"68 ]69 }70 ]71 }72 }73 }74 }75}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.337Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":725}}158{"id":"doc-list_datasets_cohere-07fdb371","source":"documentation","title":"List Datasets | Cohere","url":"https://docs.cohere.com/reference/list-datasets","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45# get list of datasets6response = co.datasets.list()78print(response)\n```\n\nExample:\n```text\n1{2 \"datasets\": [3 {4 \"id\": \"dset-9f8b7c6a5e4d3f21\",5 \"name\": \"Customer Support Chat Logs\",6 \"created_at\": \"2024-01-15T09:30:00Z\",7 \"updated_at\": \"2024-04-10T12:45:00Z\",8 \"dataset_type\": \"chat-finetune-input\",9 \"validation_status\": \"validated\",10 \"validation_error\": \"\",11 \"schema\": \"{\\\"type\\\":\\\"record\\\",\\\"name\\\":\\\"ChatExample\\\",\\\"fields\\\":[{\\\"name\\\":\\\"prompt\\\",\\\"type\\\":\\\"string\\\"},{\\\"name\\\":\\\"response\\\",\\\"type\\\":\\\"string\\\"}]}\",12 \"required_fields\": [13 \"prompt\",14 \"response\"15 ],16 \"preserve_fields\": [17 \"metadata\",18 \"timestamp\"19 ],20 \"dataset_parts\": [21 {22 \"id\": \"part-001\",23 \"name\": \"chat_logs_part1.csv\",24 \"url\": \"https://storage.cohere.com/datasets/dset-9f8b7c6a5e4d3f21/part1.csv\",25 \"index\": 0,26 \"size_bytes\": 2048576,27 \"num_rows\": 50000,28 \"original_url\": \"https://originalsource.com/chat_logs_part1.csv\",29 \"samples\": [30 \"How can I reset my password?,To reset your password, click on 'Forgot Password' at login.\",31 \"What are your support hours?,Our support team is available 24/7.\"32 ]33 }34 ],35 \"validation_warnings\": [36 \"Some entries missing optional metadata fields\"37 ],38 \"parse_info\": {},39 \"metrics\": {40 \"finetune_dataset_metrics\": {41 \"trainable_token_count\": 1250000,42 \"total_examples\": 50000,43 \"train_examples\": 40000,44 \"train_size_bytes\": 1638400,45 \"eval_examples\": 10000,46 \"eval_size_bytes\": 409600,47 \"reranker_data_metrics\": {48 \"num_train_queries\": 0,49 \"num_train_relevant_passages\": 0,50 \"num_train_hard_negatives\": 0,51 \"num_eval_queries\": 0,52 \"num_eval_relevant_passages\": 0,53 \"num_eval_hard_negatives\": 054 },55 \"chat_data_metrics\": {56 \"num_train_turns\": 120000,57 \"num_eval_turns\": 30000,58 \"preamble\": \"Customer support chat logs for fine-tuning conversational AI.\"59 },60 \"classify_data_metrics\": {61 \"label_metrics\": [62 {63 \"total_examples\": 25000,64 \"label\": \"billing\",65 \"samples\": [66 \"I was charged twice for my subscription.\",67 \"How do I update my billing information?\"68 ]69 },70 {71 \"total_examples\": 25000,72 \"label\": \"technical_support\",73 \"samples\": [74 \"My app crashes when I open it.\",75 \"How do I reset my device?\"76 ]77 }78 ]79 }80 }81 }82 }83 ]84}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.337Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":799}}159{"id":"doc-delete_a_connector_cohere-23f7e1e6","source":"documentation","title":"Delete a Connector | Cohere","url":"https://docs.cohere.com/reference/delete-connector","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nDelete a connector by ID. See ‘Connectors’ for more information.\n\nExample:\n```text\n1import cohere23co = cohere.Client()4co.connectors.delete(\"test-id\")\n```\n\nExample:\n```text\n1{}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.338Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":92}}160{"id":"doc-retrieval_augmented_generation_rag_cohere_on_azu-704c069f","source":"documentation","title":"Retrieval augmented generation (RAG) - Cohere on Azure AI Foundry | Cohere","url":"https://docs.cohere.com/docs/cohere-on-azure/azure-ai-rag","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# %pip install cohere hnswlib unstructured23import cohere45co_chat = cohere.ClientV2(6 api_key=\"AZURE_API_KEY_CHAT\",7 base_url=\"AZURE_ENDPOINT_CHAT\", # example: \"https://cohere-command-r-plus-08-2024-xyz.eastus.models.ai.azure.com/\"8)910co_embed = cohere.ClientV2(11 api_key=\"AZURE_API_KEY_EMBED\",12 base_url=\"AZURE_ENDPOINT_EMBED\", # example: \"https://embed-v-4-0-xyz.eastus.models.ai.azure.com/\"13)1415co_rerank = cohere.ClientV2(16 api_key=\"AZURE_API_KEY_RERANK\",17 base_url=\"AZURE_ENDPOINT_RERANK\", # example: \"https://cohere-rerank-v3-multilingual-xyz.eastus.models.ai.azure.com/\"18)\n```\n\nExample:\n```text\n1documents = [2 {3 \"title\": \"Tall penguins\",4 \"text\": \"Emperor penguins are the tallest.\",5 },6 {7 \"title\": \"Penguin habitats\",8 \"text\": \"Emperor penguins only live in Antarctica.\",9 },10 {11 \"title\": \"What are animals?\",12 \"text\": \"Animals are different from plants.\",13 },14]\n```\n\nExample:\n```text\n1message = \"What are the tallest living penguins?\"23response = co_chat.chat(4 model=\"model\", # Pass a dummy string5 messages=[{\"role\": \"user\", \"content\": message}],6 documents=[{\"data\": doc} for doc in documents],7)89print(\"\\nRESPONSE:\\n\")10print(response.message.content[0].text)1112if response.message.citations:13 print(\"\\nCITATIONS:\\n\")14 for citation in response.message.citations:15 print(citation)\n```\n\nExample:\n```text\n1RESPONSE:23The tallest living penguins are the Emperor penguins. They only live in Antarctica.45CITATIONS:67start=36 end=53 text='Emperor penguins.' sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Emperor penguins are the tallest.', 'title': 'Tall penguins'})] type=None8start=59 end=83 text='only live in Antarctica.' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Emperor penguins only live in Antarctica.', 'title': 'Penguin habitats'})] type=None\n```\n\nExample:\n```text\n1import uuid2import yaml3import hnswlib4from typing import List, Dict5from unstructured.partition.html import partition_html6from unstructured.chunking.title import chunk_by_title\n```\n\nExample:\n```text\n1raw_documents = [2 {3 \"title\": \"Crafting Effective Prompts\",4 \"url\": \"https://docs.cohere.com/docs/crafting-effective-prompts\",5 },6 {7 \"title\": \"Advanced Prompt Engineering Techniques\",8 \"url\": \"https://docs.cohere.com/docs/advanced-prompt-engineering-techniques\",9 },10 {11 \"title\": \"Prompt Truncation\",12 \"url\": \"https://docs.cohere.com/docs/prompt-truncation\",13 },14 {15 \"title\": \"Preambles\",16 \"url\": \"https://docs.cohere.com/docs/preambles\",17 },18]\n```\n\nExample:\n```text\n1class Vectorstore:23 def __init__(self, raw_documents: List[Dict[str, str]]):4 self.raw_documents = raw_documents5 self.docs = []6 self.docs_embs = []7 self.retrieve_top_k = 108 self.rerank_top_k = 39 self.load_and_chunk()10 self.embed()11 self.index()1213 def load_and_chunk(self) -> None:14 \"\"\"15 Loads the text from the sources and chunks the HTML content.16 \"\"\"17 print(\"Loading documents...\")1819 for raw_document in self.raw_documents:20 elements = partition_html(url=raw_document[\"url\"])21 chunks = chunk_by_title(elements)22 for chunk in chunks:23 self.docs.append(24 {25 \"data\": {26 \"title\": raw_document[\"title\"],27 \"text\": str(chunk),28 \"url\": raw_document[\"url\"],29 }30 }31 )3233 def embed(self) -> None:34 \"\"\"35 Embeds the document chunks using the Cohere API.36 \"\"\"37 print(\"Embedding document chunks...\")3839 batch_size = 9040 self.docs_len = len(self.docs)41 for i in range(0, self.docs_len, batch_size):42 batch = self.docs[i : min(i + batch_size, self.docs_len)]43 texts = [item[\"data\"][\"text\"] for item in batch]44 docs_embs_batch = co_embed.embed(45 texts=texts,46 model=\"embed-v4.0\",47 input_type=\"search_document\",48 embedding_types=[\"float\"],49 ).embeddings.float50 self.docs_embs.extend(docs_embs_batch)5152 def index(self) -> None:53 \"\"\"54 Indexes the document chunks for efficient retrieval.55 \"\"\"56 print(\"Indexing document chunks...\")5758 self.idx = hnswlib.Index(space=\"ip\", dim=1024)59 self.idx.init_index(60 max_elements=self.docs_len, ef_construction=512, M=6461 )62 self.idx.add_items(63 self.docs_embs, list(range(len(self.docs_embs)))64 )6566 print(67 f\"Indexing complete with {self.idx.get_current_count()} document chunks.\"68 )6970 def retrieve(self, query: str) -> List[Dict[str, str]]:71 \"\"\"72 Retrieves document chunks based on the given query.7374 Parameters:75 query (str): The query to retrieve document chunks for.7677 Returns:78 List[Dict[str, str]]: A list of dictionaries representing the retrieved document chunks, with 'title', 'text', and 'url' keys.79 \"\"\"8081 # Dense retrieval82 query_emb = co_embed.embed(83 texts=[query],84 model=\"embed-v4.0\",85 input_type=\"search_query\",86 embedding_types=[\"float\"],87 ).embeddings.float8889 doc_ids = self.idx.knn_query(90 query_emb, k=self.retrieve_top_k91 )[0][0]9293 # Reranking94 docs_to_rerank = [95 self.docs[doc_id][\"data\"] for doc_id in doc_ids96 ]97 yaml_docs = [98 yaml.dump(doc, sort_keys=False) for doc in docs_to_rerank99 ]100 rerank_results = co_rerank.rerank(101 query=query,102 documents=yaml_docs,103 model=\"model\", # Pass a dummy string104 top_n=self.rerank_top_k,105 )106107 doc_ids_reranked = [108 doc_ids[result.index] for result in rerank_results.results109 ]110111 docs_retrieved = []112 for doc_id in doc_ids_reranked:113 docs_retrieved.append(self.docs[doc_id][\"data\"])114115 return docs_retrieved\n```\n\nExample:\n```text\n1# Create an instance of the Vectorstore class with the given sources2vectorstore = Vectorstore(raw_documents)\n```\n\nExample:\n```text\n1Loading documents...2Embedding document chunks...3Indexing document chunks...4Indexing complete with 137 document chunks.\n```\n\nExample:\n```text\n1vectorstore.retrieve(\"Prompting by giving examples\")\n```\n\nExample:\n```text\n1[{'title': 'Advanced Prompt Engineering Techniques',2 'text': 'Few-shot Prompting\\n\\nUnlike the zero-shot examples above, few-shot prompting is a technique that provides a model with examples of the task being performed before asking the specific question to be answered. We can steer the LLM toward a high-quality solution by providing a few relevant and diverse examples in the prompt. Good examples condition the model to the expected response type and style.',3 'url': 'https://docs.cohere.com/docs/advanced-prompt-engineering-techniques'},4 {'title': 'Crafting Effective Prompts',5 'text': 'Incorporating Example Outputs\\n\\nLLMs respond well when they have specific examples to work from. For example, instead of asking for the salient points of the text and using bullet points “where appropriate”, give an example of what the output should look like.',6 'url': 'https://docs.cohere.com/docs/crafting-effective-prompts'},7 {'title': 'Advanced Prompt Engineering Techniques',8 'text': 'In addition to giving correct examples, including negative examples with a clear indication of why they are wrong can help the LLM learn to distinguish between correct and incorrect responses. Ordering the examples can also be important; if there are patterns that could be picked up on that are not relevant to the correctness of the question, the model may incorrectly pick up on those instead of the semantics of the question itself.',9 'url': 'https://docs.cohere.com/docs/advanced-prompt-engineering-techniques'}]\n```\n\nExample:\n```text\n1def run_chatbot(query, messages=None):2 if messages is None:3 messages = []45 messages.append({\"role\": \"user\", \"content\": query})67 # Retrieve document chunks and format8 documents = vectorstore.retrieve(query)9 documents_formatted = []10 for doc in documents:11 documents_formatted.append({\"data\": doc})1213 # Use document chunks to respond14 response = co_chat.chat(15 model=\"model\", # Pass a dummy string16 messages=messages,17 documents=documents_formatted,18 )1920 # Print the chatbot response, citations, and documents21 print(\"\\nRESPONSE:\\n\")22 print(response.message.content[0].text)2324 if response.message.citations:25 print(\"\\nCITATIONS:\\n\")26 for citation in response.message.citations:27 print(\"-\" * 20)28 print(29 \"start:\",30 citation.start,31 \"end:\",32 citation.end,33 \"text:\",34 citation.text,35 )36 print(\"SOURCES:\")37 print(citation.sources)3839 # Add assistant response to messages40 messages.append(41 {42 \"role\": \"assistant\",43 \"content\": response.message.content[0].text,44 }45 )4647 return messages\n```\n\nExample:\n```text\n1messages = run_chatbot(\"Hello, I have a question\")\n```\n\nExample:\n```text\n1RESPONSE:23Hello there! How can I help you today?\n```\n\nExample:\n```text\n1messages = run_chatbot(\"How to provide examples in prompts\", messages)\n```\n\nExample:\n```text\nRESPONSE:There are a few ways to provide examples in prompts.One way is to provide a few relevant and diverse examples in the prompt. This can help steer the LLM towards a high-quality solution. Good examples condition the model to the expected response type and style.Another way is to provide specific examples to work from. For example, instead of asking for the salient points of the text and using bullet points “where appropriate”, give an example of what the output should look like.In addition to giving correct examples, including negative examples with a clear indication of why they are wrong can help the LLM learn to distinguish between correct and incorrect responses.CITATIONS:--------------------start: 68 end: 126 text: provide a few relevant and diverse examples in the prompt.SOURCES:[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Few-shot Prompting\\n\\nUnlike the zero-shot examples above, few-shot prompting is a technique that provides a model with examples of the task being performed before asking the specific question to be answered. We can steer the LLM toward a high-quality solution by providing a few relevant and diverse examples in the prompt. Good examples condition the model to the expected response type and style.', 'title': 'Advanced Prompt Engineering Techniques', 'url': 'https://docs.cohere.com/docs/advanced-prompt-engineering-techniques'})]--------------------start: 136 end: 187 text: help steer the LLM towards a high-quality solution.SOURCES:[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Few-shot Prompting\\n\\nUnlike the zero-shot examples above, few-shot prompting is a technique that provides a model with examples of the task being performed before asking the specific question to be answered. We can steer the LLM toward a high-quality solution by providing a few relevant and diverse examples in the prompt. Good examples condition the model to the expected response type and style.', 'title': 'Advanced Prompt Engineering Techniques', 'url': 'https://docs.cohere.com/docs/advanced-prompt-engineering-techniques'})]--------------------start: 188 end: 262 text: Good examples condition the model to the expected response type and style.SOURCES:[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Few-shot Prompting\\n\\nUnlike the zero-shot examples above, few-shot prompting is a technique that provides a model with examples of the task being performed before asking the specific question to be answered. We can steer the LLM toward a high-quality solution by providing a few relevant and diverse examples in the prompt. Good examples condition the model to the expected response type and style.', 'title': 'Advanced Prompt Engineering Techniques', 'url': 'https://docs.cohere.com/docs/advanced-prompt-engineering-techniques'})]--------------------start: 282 end: 321 text: provide specific examples to work from.SOURCES:[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Incorporating Example Outputs\\n\\nLLMs respond well when they have specific examples to work from. For example, instead of asking for the salient points of the text and using bullet points “where appropriate”, give an example of what the output should look like.', 'title': 'Crafting Effective Prompts', 'url': 'https://docs.cohere.com/docs/crafting-effective-prompts'})]--------------------start: 335 end: 485 text: instead of asking for the salient points of the text and using bullet points “where appropriate”, give an example of what the output should look like.SOURCES:[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Incorporating Example Outputs\\n\\nLLMs respond well when they have specific examples to work from. For example, instead of asking for the salient points of the text and using bullet points “where appropriate”, give an example of what the output should look like.', 'title': 'Crafting Effective Prompts', 'url': 'https://docs.cohere.com/docs/crafting-effective-prompts'})]--------------------start: 527 end: 679 text: including negative examples with a clear indication of why they are wrong can help the LLM learn to distinguish between correct and incorrect responses.SOURCES:[DocumentSource(type='document', id='doc:2', document={'id': 'doc:2', 'text': 'In addition to giving correct examples, including negative examples with a clear indication of why they are wrong can help the LLM learn to distinguish between correct and incorrect responses. Ordering the examples can also be important; if there are patterns that could be picked up on that are not relevant to the correctness of the question, the model may incorrectly pick up on those instead of the semantics of the question itself.', 'title': 'Advanced Prompt Engineering Techniques', 'url': 'https://docs.cohere.com/docs/advanced-prompt-engineering-techniques'})]\n```\n\nExample:\n```text\n1messages = run_chatbot(2 \"What do you know about 5G networks?\", messages3)\n```\n\nExample:\n```text\n1RESPONSE:23I'm sorry, I could not find any information about 5G networks.\n```\n\nExample:\n```text\n1for message in messages:2 print(message, \"\\n\")\n```\n\nExample:\n```text\n1{'role': 'user', 'content': 'Hello, I have a question'} 23{'role': 'assistant', 'content': 'Hello! How can I help you today?'} 45{'role': 'user', 'content': 'How to provide examples in prompts'} 67{'role': 'assistant', 'content': 'There are a few ways to provide examples in prompts.\\n\\nOne way is to provide a few relevant and diverse examples in the prompt. This can help steer the LLM towards a high-quality solution. Good examples condition the model to the expected response type and style.\\n\\nAnother way is to provide specific examples to work from. For example, instead of asking for the salient points of the text and using bullet points “where appropriate”, give an example of what the output should look like.\\n\\nIn addition to giving correct examples, including negative examples with a clear indication of why they are wrong can help the LLM learn to distinguish between correct and incorrect responses.'} 89{'role': 'user', 'content': 'What do you know about 5G networks?'} 1011{'role': 'assistant', 'content': \"I'm sorry, I could not find any information about 5G networks.\"}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.339Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":20,"totalLines":103,"estimatedTokens":4121}}161{"id":"doc-get_a_model_cohere-36f063e2","source":"documentation","title":"Get a Model | Cohere","url":"https://docs.cohere.com/reference/get-model","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from cohere import Client23client = Client()45response = client.models.get(6 model=\"command-a-03-2025\",7)8print(response)\n```\n\nExample:\n```text\n1{2 \"name\": \"command-a-03-2025\",3 \"is_deprecated\": false,4 \"endpoints\": [5 \"chat\",6 \"generate\"7 ],8 \"finetuned\": true,9 \"context_length\": 2048,10 \"tokenizer_url\": \"https://models.cohere.com/tokenizers/command-a-03-2025.json\",11 \"default_endpoints\": [12 \"chat\"13 ],14 \"features\": [15 \"few-shot-learning\",16 \"instruction-following\"17 ],18 \"sampling_defaults\": {19 \"temperature\": 0.75,20 \"k\": 40,21 \"p\": 0.9,22 \"frequency_penalty\": 0.5,23 \"presence_penalty\": 0.3,24 \"max_tokens_per_doc\": 25625 }26}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.339Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":224}}162{"id":"doc-create_a_dataset_cohere-b49882d9","source":"documentation","title":"Create a Dataset | Cohere","url":"https://docs.cohere.com/reference/create-dataset","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nCreate a dataset by uploading a file. See ‘Dataset Creation’ for more information.\n\nList of names of fields that will be persisted in the Dataset. By default the Dataset will retain only the required fields indicated in the schema for the corresponding Dataset type. For example, datasets of type embed-input will drop all fields other than the required text field. If any of the fields in keep_fields are missing from the uploaded file, Dataset validation will fail.\n\nList of names of fields that will be persisted in the Dataset. By default the Dataset will retain only the required fields indicated in the schema for the corresponding Dataset type. For example, Datasets of type embed-input will drop all fields other than the required text field. If any of the fields in optional_fields are missing from the uploaded file, Dataset validation will pass.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45# upload a dataset6my_dataset = co.datasets.create(7 name=\"embed-dataset\",8 data=open(\"./embed.jsonl\", \"rb\"),9 type=\"embed-input\",10)1112# wait for validation to complete13response = co.wait(my_dataset)1415print(response)\n```\n\nExample:\n```text\n1{2 \"id\": \"ds_123e4567-e89b-12d3-a456-426614174000\"3}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.339Z","totalSectionsIncluded":4,"totalCodeBlocksIncluded":2,"totalLines":19,"estimatedTokens":352}}163{"id":"doc-authorize_with_oauth_cohere-efe9cbca","source":"documentation","title":"Authorize with oAuth | Cohere","url":"https://docs.cohere.com/reference/oauthauthorize-connector","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nAuthorize the connector with the given ID for the connector oauth app. See ‘Connector Authentication’ for more information.\n\nExample:\n```text\n1import cohere23co = cohere.Client()4response = co.connectors.o_auth_authorize(5 connector_id=\"test-id\", after_token_redirect=\"https://test.com\"6)7print(response)\n```\n\nExample:\n```text\n1{2 \"redirect_url\": \"https://auth.example.com/oauth2/authorize?client_id=abc123&response_type=code&redirect_uri=https%3A%2F%2Fapp.example.com%2Fcallback&scope=read_profile\"3}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.339Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":174}}164{"id":"doc-list_models_cohere-e3876d48","source":"documentation","title":"List Models | Cohere","url":"https://docs.cohere.com/reference/list-models","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()4response = co.models.list()5print(response)\n```\n\nExample:\n```text\n1{2 \"models\": [3 {4 \"name\": \"command-xlarge-nightly\",5 \"is_deprecated\": false,6 \"endpoints\": [7 \"generate\",8 \"chat\"9 ],10 \"finetuned\": false,11 \"context_length\": 2048,12 \"tokenizer_url\": \"https://cohere.com/tokenizer/command-xlarge-nightly.json\",13 \"default_endpoints\": [14 \"generate\"15 ],16 \"features\": [17 \"text-generation\",18 \"chat-completions\"19 ],20 \"sampling_defaults\": {21 \"temperature\": 0.75,22 \"k\": 40,23 \"p\": 0.9,24 \"frequency_penalty\": 0,25 \"presence_penalty\": 0,26 \"max_tokens_per_doc\": 30027 }28 }29 ],30 \"next_page_token\": \"eyJwYWdlIjoxfQ==\"31}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.339Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":255}}165{"id":"doc-deletes_a_fine_tuned_model_cohere-6ddf66ff","source":"documentation","title":"Deletes a fine-tuned model. | Cohere","url":"https://docs.cohere.com/reference/deletefinetunedmodel","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()4co.finetuning.delete_finetuned_model(\"test-id\")\n```\n\nExample:\n```text\n1{}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.340Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":79}}166{"id":"doc-detokenize_cohere-f7aaa0ff","source":"documentation","title":"Detokenize | Cohere","url":"https://docs.cohere.com/reference/detokenize","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45response = co.detokenize(6 tokens=[8466, 5169, 2594, 8, 2792, 43], model=\"command-a-03-2025\" # optional7)8print(response)\n```\n\nExample:\n```text\n1{2 \"text\": \"tokenize me! :D\",3 \"meta\": {4 \"api_version\": {5 \"version\": \"1\"6 }7 }8}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.340Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":122}}167{"id":"doc-list_connectors_cohere-0a6e0ffe","source":"documentation","title":"List Connectors | Cohere","url":"https://docs.cohere.com/reference/list-connectors","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nReturns a list of connectors ordered by descending creation date (newer first). See ‘Managing your Connector’ for more information.\n\nExample:\n```text\n1import cohere23co = cohere.Client()4response = co.connectors.list()5print(response)\n```\n\nExample:\n```text\n1{2 \"connectors\": [3 {4 \"id\": \"salesforce-integration\",5 \"name\": \"Salesforce CRM Connector\",6 \"created_at\": \"2024-04-10T14:22:00Z\",7 \"updated_at\": \"2024-04-15T11:45:00Z\",8 \"organization_id\": \"org_9f8e7d6c5b4a3\",9 \"description\": \"Connector to integrate Salesforce CRM data for customer insights.\",10 \"url\": \"https://api.salesforce.com/v1/documents/search\",11 \"excludes\": [12 \"internal_notes\"13 ],14 \"auth_type\": \"oauth\",15 \"oauth\": {16 \"authorize_url\": \"https://login.salesforce.com/services/oauth2/authorize\",17 \"token_url\": \"https://login.salesforce.com/services/oauth2/token\",18 \"client_id\": \"sf-client-12345\",19 \"client_secret\": \"encrypted-secret-placeholder\",20 \"scope\": \"api refresh_token\"21 },22 \"auth_status\": \"valid\",23 \"active\": true,24 \"continue_on_failure\": true25 }26 ],27 \"total_count\": 128}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.340Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":344}}168{"id":"doc-returns_a_fine_tuned_model_by_id_cohere-d251299c","source":"documentation","title":"Returns a fine-tuned model by ID. | Cohere","url":"https://docs.cohere.com/reference/getfinetunedmodel","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()4response = co.finetuning.get_finetuned_model(\"test-id\")5print(response)\n```\n\nExample:\n```text\n1{2 \"finetuned_model\": {3 \"name\": \"chat-ft\",4 \"settings\": {5 \"base_model\": {6 \"base_type\": \"BASE_TYPE_CHAT\",7 \"name\": \"medium\",8 \"version\": \"14.2.0\",9 \"strategy\": \"STRATEGY_TFEW\"10 },11 \"dataset_id\": \"my-dataset-d701tr\",12 \"hyperparameters\": {13 \"early_stopping_patience\": 6,14 \"early_stopping_threshold\": 0.01,15 \"train_batch_size\": 16,16 \"train_epochs\": 1,17 \"learning_rate\": 0.0118 }19 },20 \"id\": \"fee37446-7fc7-42f9-a026-c6ba2fcc422d\",21 \"creator_id\": \"7a317d97-4d05-427d-9396-f31b9fb92c55\",22 \"organization_id\": \"6bdca3d5-3eae-4de0-ac34-786d8063b7ee\",23 \"status\": \"STATUS_READY\",24 \"created_at\": \"2024-01-15T09:30:00Z\",25 \"updated_at\": \"2024-01-15T09:30:00Z\",26 \"completed_at\": \"2024-01-15T09:30:00Z\"27 }28}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.340Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":292}}169{"id":"doc-chat_with_streaming_cohere-b9490d42","source":"documentation","title":"Chat with Streaming | Cohere","url":"https://docs.cohere.com/reference/chat-stream","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nGenerates a text response to a user message. To learn how to use the Chat API and RAG follow our Text Generation guides. Follow the Migration Guide for instructions on moving from API v1 to API v2.\n\nThe name of a compatible Cohere model.\n\nA list of chat messages in chronological order, representing a conversation between the user and the model. Messages can be from User, Assistant, Tool and System roles. Learn more about messages and roles in the Chat API guide.\n\nA list of tools (functions) available to the model. The model response may contain ‘tool_calls’ to the specified tools. Learn more in the Tool Use guide.\n\nConfiguration for forcing the model output to adhere to the specified format. Supported on Command R, Command R+ and newer models. The model can be forced into outputting JSON objects by setting { \"type\": \"json_object\" }. A JSON Schema can optionally be provided, to ensure a specific structure. using { \"type\": \"json_object\" } your message should always explicitly instruct the model to generate a JSON (eg: “Generate a JSON …”) . Otherwise the model may end up getting stuck generating an infinite stream of characters and eventually run out of context length. json_schema is not specified, the generated object can have up to 5 layers of nesting. parameter is not supported when used in combinations with the documents or tools parameters.\n\nUsed to select the safety instruction inserted into the prompt. Defaults to CONTEXTUAL. When OFF is specified, the safety instruction will be omitted. Safety modes are not yet configurable in combination with tools and documents parameters. parameter is only compatible newer Cohere models, starting with Command R 08-2024 and Command R+ 08-2024. and newer models only support \"CONTEXTUAL\" and \"STRICT\" modes.\n\nThe maximum number of output tokens the model will generate in the response. If not set, max_tokens defaults to the model’s maximum output token limit. You can find the maximum output token limits for each model in the model documentation. a low value may result in incomplete generations. In such cases, the finish_reason field in the response will be set to \"MAX_TOKENS\". max_tokens is set higher than the model’s maximum output token limit, the generation will be capped at that model-specific maximum limit.\n\nUsed to control whether or not the model will be forced to use a tool when answering. When REQUIRED is specified, the model will be forced to use at least one of the user-defined tools, and the tools parameter must be passed in the request. When NONE is specified, the model will be forced not to use one of the specified tools, and give a direct response. If tool_choice isn’t specified, then the model is free to choose whether to use the specified tools or not. parameter is only compatible with models Command-r7b and newer.\n\nConfiguration for reasoning features.\n\nWhen set to true, tool calls in the Assistant message will be forced to follow the tool definition strictly. Learn more in the Structured Outputs (Tools) guide. first few requests with a new set of tools will take longer to process.\n\nGenerates a text response to a user message. To learn how to use the Chat API and RAG follow our Text Generation guides. Follow the Migration Guide for instructions on moving from API v1 to API v2.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45response = co.chat_stream(6 model=\"command-a-plus-05-2026\",7 messages=[{\"role\": \"user\", \"content\": \"Tell me about LLMs\"}],8)910for event in response:11 if event.type == \"content-delta\":12 print(event.delta.message.content.text, end=\"\")\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.341Z","totalSectionsIncluded":12,"totalCodeBlocksIncluded":1,"totalLines":30,"estimatedTokens":948}}170{"id":"doc-using_cohere_models_via_the_openai_sdk_cohere-1d5d1747","source":"documentation","title":"Using Cohere models via the OpenAI SDK | Cohere","url":"https://docs.cohere.com/docs/compatibility-api","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$pip install openai\n```\n\nExample:\n```text\n1from openai import OpenAI23client = OpenAI(4 base_url=\"https://api.cohere.ai/compatibility/v1\",5 api_key=\"COHERE_API_KEY\",6)\n```\n\nExample:\n```text\n1from openai import OpenAI23client = OpenAI(4 base_url=\"https://api.cohere.ai/compatibility/v1\",5 api_key=\"COHERE_API_KEY\",6)78completion = client.chat.completions.create(9 model=\"command-a-plus-05-2026\",10 messages=[11 {12 \"role\": \"user\",13 \"content\": \"Write a haiku about recursion in programming.\",14 },15 ],16)1718print(completion.choices[0].message)\n```\n\nExample:\n```text\n1ChatCompletionMessage(content=\"Recursive loops,\\nUnraveling code's depths,\\nEndless, yet complete.\", refusal=None, role='assistant', audio=None, function_call=None, tool_calls=None)\n```\n\nExample:\n```text\n1from openai import OpenAI23client = OpenAI(4 base_url=\"https://api.cohere.ai/compatibility/v1\",5 api_key=\"COHERE_API_KEY\",6)78stream = client.chat.completions.create(9 model=\"command-a-plus-05-2026\",10 messages=[11 {12 \"role\": \"user\",13 \"content\": \"Write a haiku about recursion in programming.\",14 },15 ],16 stream=True,17)1819for chunk in stream:20 print(chunk.choices[0].delta.content or \"\", end=\"\")\n```\n\nExample:\n```text\n1Recursive call,2Unraveling, line by line,3Solving, then again.\n```\n\nExample:\n```text\n1from openai import OpenAI23client = OpenAI(4 base_url=\"https://api.cohere.ai/compatibility/v1\",5 api_key=\"COHERE_API_KEY\",6)78completion = client.chat.completions.create(9 messages=[10 {11 \"role\": \"developer\",12 \"content\": \"You must respond in the style of a pirate.\",13 },14 {15 \"role\": \"user\",16 \"content\": \"What's 2 + 2.\",17 },18 {19 \"role\": \"assistant\",20 \"content\": \"Arrr, matey! 2 + 2 be 4, just like a doubloon in the sea!\",21 },22 {23 \"role\": \"user\",24 \"content\": \"Add 30 to that.\",25 },26 ],27 model=\"command-a-plus-05-2026\",28)2930print(completion.choices[0].message)\n```\n\nExample:\n```text\n1ChatCompletionMessage(content='Aye aye, captain! 4 + 30 be 34, a treasure to behold!', refusal=None, role='assistant', audio=None, function_call=None, tool_calls=None)\n```\n\nExample:\n```text\n1from openai import OpenAI23client = OpenAI(4 base_url=\"https://api.cohere.ai/compatibility/v1\",5 api_key=\"COHERE_API_KEY\",6)78completion = client.beta.chat.completions.parse(9 model=\"command-a-plus-05-2026\",10 messages=[11 {12 \"role\": \"user\",13 \"content\": \"Generate a JSON describing a book.\",14 }15 ],16 response_format={17 \"type\": \"json_object\",18 \"schema\": {19 \"type\": \"object\",20 \"properties\": {21 \"title\": {\"type\": \"string\"},22 \"author\": {\"type\": \"string\"},23 \"publication_year\": {\"type\": \"integer\"},24 },25 \"required\": [\"title\", \"author\", \"publication_year\"],26 },27 },28)2930print(completion.choices[0].message.content)\n```\n\nExample:\n```text\n{ \"title\": \"The Great Gatsby\", \"author\": \"F. Scott Fitzgerald\", \"publication_year\": 1925}\n```\n\nExample:\n```text\n1from openai import OpenAI23client = OpenAI(4 base_url=\"https://api.cohere.ai/compatibility/v1\",5 api_key=\"COHERE_API_KEY\",6)78tools = [9 {10 \"type\": \"function\",11 \"function\": {12 \"name\": \"get_flight_info\",13 \"description\": \"Get flight information between two cities or airports\",14 \"parameters\": {15 \"type\": \"object\",16 \"properties\": {17 \"loc_origin\": {18 \"type\": \"string\",19 \"description\": \"The departure airport, e.g. MIA\",20 },21 \"loc_destination\": {22 \"type\": \"string\",23 \"description\": \"The destination airport, e.g. NYC\",24 },25 },26 \"required\": [\"loc_origin\", \"loc_destination\"],27 },28 },29 }30]3132messages = [33 {\"role\": \"developer\", \"content\": \"Today is April 30th\"},34 {35 \"role\": \"user\",36 \"content\": \"When is the next flight from Miami to Seattle?\",37 },38 {39 \"role\": \"assistant\",40 \"tool_calls\": [41 {42 \"function\": {43 \"arguments\": '{ \"loc_destination\": \"Seattle\", \"loc_origin\": \"Miami\" }',44 \"name\": \"get_flight_info\",45 },46 \"id\": \"get_flight_info0\",47 \"type\": \"function\",48 }49 ],50 },51 {52 \"role\": \"tool\",53 \"name\": \"get_flight_info\",54 \"tool_call_id\": \"get_flight_info0\",55 \"content\": \"Miami to Seattle, May 1st, 10 AM.\",56 },57]5859completion = client.chat.completions.create(60 model=\"command-a-plus-05-2026\",61 messages=messages,62 tools=tools,63 temperature=0.7,64)6566print(completion.choices[0].message)\n```\n\nExample:\n```text\n1ChatCompletionMessage(content='The next flight from Miami to Seattle is on May 1st, 10 AM.', refusal=None, role='assistant', audio=None, function_call=None, tool_calls=None)\n```\n\nExample:\n```text\n1from openai import OpenAI23client = OpenAI(4 base_url=\"https://api.cohere.ai/compatibility/v1\",5 api_key=COHERE_API_KEY,6)78response = client.embeddings.create(9 input=[\"Hello world!\"],10 model=\"embed-v4.0\",11 encoding_format=\"float\",12)1314print(15 response.data[0].embedding[:5]16) # Display the first 5 dimensions\n```\n\nExample:\n```text\n1[0.0045051575, 0.046905518, 0.025543213, 0.009651184, -0.024993896]\n```\n\nExample:\n```text\n1from openai import OpenAI23client = OpenAI(4 base_url=\"https://api.cohere.ai/compatibility/v1\",5 api_key=COHERE_API_KEY,6)78response = client.audio.transcriptions.create(9 model=\"cohere-transcribe-03-2026\",10 language=\"en\",11 file=open(\"./sample.wav\", \"rb\"),12)1314print(response)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.341Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":1578}}171{"id":"doc-get_a_connector_cohere-25afc535","source":"documentation","title":"Get a Connector | Cohere","url":"https://docs.cohere.com/reference/get-connector","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nRetrieve a connector by ID. See ‘Connectors’ for more information.\n\nExample:\n```text\n1import cohere23co = cohere.Client()4response = co.connectors.get(\"test-id\")5print(response)\n```\n\nExample:\n```text\n1{2 \"connector\": {3 \"id\": \"salesforce-crm-connector\",4 \"name\": \"Salesforce CRM Integration\",5 \"created_at\": \"2024-01-15T09:30:00Z\",6 \"updated_at\": \"2024-04-10T14:45:00Z\",7 \"organization_id\": \"org_9f8e7d6c5b4a3\",8 \"description\": \"Connector to integrate Salesforce CRM data for customer insights and support.\",9 \"url\": \"https://api.salesforce.com/v1/documents/search\",10 \"excludes\": [11 \"internal_notes\"12 ],13 \"auth_type\": \"oauth\",14 \"oauth\": {15 \"authorize_url\": \"https://login.salesforce.com/services/oauth2/authorize\",16 \"token_url\": \"https://login.salesforce.com/services/oauth2/token\",17 \"client_id\": \"sf-client-1234567890\",18 \"client_secret\": \"encrypted-secret-not-returned\",19 \"scope\": \"api refresh_token\"20 },21 \"auth_status\": \"valid\",22 \"active\": true,23 \"continue_on_failure\": true24 }25}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.341Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":316}}172{"id":"doc-update_a_connector_cohere-d762eba4","source":"documentation","title":"Update a Connector | Cohere","url":"https://docs.cohere.com/reference/update-connector","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nUpdate a connector by ID. Omitted fields will not be updated. See ‘Managing your Connector’ for more information.\n\nExample:\n```text\n1import cohere23co = cohere.Client()4response = co.connectors.update(5 connector_id=\"test-id\", name=\"new name\", url=\"https://example.com/search\"6)7print(response)\n```\n\nExample:\n```text\n1{2 \"connector\": {3 \"id\": \"connector-12345\",4 \"name\": \"Salesforce Data Connector\",5 \"created_at\": \"2024-01-15T09:30:00Z\",6 \"updated_at\": \"2024-01-15T09:30:00Z\",7 \"organization_id\": \"org-67890\",8 \"description\": \"Connector for integrating Salesforce CRM data.\",9 \"url\": \"https://salesforce.example.com/api/search\",10 \"excludes\": [11 \"password\",12 \"ssn\"13 ],14 \"auth_type\": \"oauth\",15 \"oauth\": {16 \"authorize_url\": \"https://login.salesforce.com/services/oauth2/authorize\",17 \"token_url\": \"https://login.salesforce.com/services/oauth2/token\",18 \"client_id\": \"sf-client-abc123\",19 \"client_secret\": \"SGVsbG8gV29ybGQ=\",20 \"scope\": \"api refresh_token\"21 },22 \"auth_status\": \"valid\",23 \"active\": true,24 \"continue_on_failure\": true25 }26}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.341Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":330}}173{"id":"doc-fetch_history_of_statuses_for_a_fine_tuned_model-e82940d5","source":"documentation","title":"Fetch history of statuses for a fine-tuned model. | Cohere","url":"https://docs.cohere.com/reference/listevents","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nComma separated list of fields. For example: “created_at,name”. The default sorting order is ascending. To specify descending order for a field, append ” desc” to the field name. For example: “created_at desc,name”. Supported sorting (default)\n\nExample:\n```text\n1import cohere23co = cohere.Client()4response = co.finetuning.list_events(finetuned_model_id=\"test-id\")5print(response)\n```\n\nExample:\n```text\n1{2 \"events\": [3 {4 \"user_id\": \"7a317d97-4d05-427d-9396-f31b9fb92c55\",5 \"status\": \"STATUS_QUEUED\",6 \"created_at\": \"2024-01-15T09:30:00Z\"7 },8 {9 \"user_id\": \"7a317d97-4d05-427d-9396-f31b9fb92c55\",10 \"status\": \"STATUS_FINETUNING\",11 \"created_at\": \"2024-01-15T09:30:00Z\"12 },13 {14 \"user_id\": \"7a317d97-4d05-427d-9396-f31b9fb92c55\",15 \"status\": \"STATUS_DEPLOYING_API\",16 \"created_at\": \"2024-01-15T09:30:00Z\"17 },18 {19 \"status\": \"STATUS_READY\",20 \"created_at\": \"2024-01-15T09:30:00Z\"21 }22 ],23 \"total_size\": 524}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.342Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":296}}174{"id":"doc-create_a_connector_cohere-a4c24ad7","source":"documentation","title":"Create a Connector | Cohere","url":"https://docs.cohere.com/reference/create-connector","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nCreates a new connector. The connector is tested during registration and will cancel registration when the test is unsuccessful. See ‘Creating and Deploying a Connector’ for more information.\n\nExample:\n```text\n1import cohere23co = cohere.Client()4response = co.connectors.create(5 name=\"Example connector\",6 url=\"https://connector-example.com/search\",7)8print(response)\n```\n\nExample:\n```text\n1{2 \"connector\": {3 \"id\": \"salesforce-crm-connector-01\",4 \"name\": \"Salesforce CRM Connector\",5 \"created_at\": \"2024-01-15T09:30:00Z\",6 \"updated_at\": \"2024-01-15T09:30:00Z\",7 \"organization_id\": \"org-1234567890\",8 \"description\": \"Connector to integrate Salesforce CRM data for document search and retrieval.\",9 \"url\": \"https://api.salesforce.com/v1/search\",10 \"excludes\": [11 \"internalNotes\"12 ],13 \"auth_type\": \"service_auth\",14 \"oauth\": {15 \"authorize_url\": \"https://login.salesforce.com/services/oauth2/authorize\",16 \"token_url\": \"https://login.salesforce.com/services/oauth2/token\",17 \"client_id\": \"sf-client-abc123\",18 \"client_secret\": \"SGVsbG8gV29ybGQ=\",19 \"scope\": \"api refresh_token\"20 },21 \"auth_status\": \"valid\",22 \"active\": true,23 \"continue_on_failure\": true24 }25}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.342Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":360}}175{"id":"doc-trains_and_deploys_a_fine_tuned_model_cohere-a169de78","source":"documentation","title":"Trains and deploys a fine-tuned model. | Cohere","url":"https://docs.cohere.com/reference/createfinetunedmodel","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from cohere.finetuning import (2 BaseModel,3 FinetunedModel,4 Hyperparameters,5 Settings,6 WandbConfig,7)8import cohere910co = cohere.Client()11hp = Hyperparameters(12 early_stopping_patience=10,13 early_stopping_threshold=0.001,14 train_batch_size=16,15 train_epochs=1,16 learning_rate=0.01,17)18wnb_config = WandbConfig(19 project=\"test-project\",20 api_key=\"<<wandbApiKey>>\",21 entity=\"test-entity\",22)23finetuned_model = co.finetuning.create_finetuned_model(24 request=FinetunedModel(25 name=\"test-finetuned-model\",26 settings=Settings(27 base_model=BaseModel(28 base_type=\"BASE_TYPE_CHAT\",29 ),30 dataset_id=\"my-dataset-id\",31 hyperparameters=hp,32 wandb=wnb_config,33 ),34 )35)36print(finetuned_model)\n```\n\nExample:\n```text\n1{2 \"finetuned_model\": {3 \"name\": \"customer-support-chatbot-v1\",4 \"settings\": {5 \"base_model\": {6 \"base_type\": \"BASE_TYPE_CHAT\",7 \"name\": \"cohere-chat-2024\",8 \"version\": \"v2.1.0\",9 \"strategy\": \"STRATEGY_VANILLA\"10 },11 \"dataset_id\": \"customer-support-dataset-2024\",12 \"hyperparameters\": {13 \"early_stopping_patience\": 5,14 \"early_stopping_threshold\": 0.0005,15 \"train_batch_size\": 32,16 \"train_epochs\": 3,17 \"learning_rate\": 0.005,18 \"lora_alpha\": 16,19 \"lora_rank\": 4,20 \"lora_target_modules\": \"LORA_TARGET_MODULES_QKVO\"21 },22 \"multi_label\": false,23 \"wandb\": {24 \"project\": \"customer-support-finetuning\",25 \"api_key\": \"wandb_api_key_1234567890abcdef\",26 \"entity\": \"cohere-team\"27 }28 },29 \"id\": \"ftm_123e4567-e89b-12d3-a456-426614174000\",30 \"creator_id\": \"user_789a1234-b56c-78d9-e012-3456789abcde\",31 \"organization_id\": \"org_456b7890-c12d-34e5-f678-901234567890\",32 \"status\": \"STATUS_FINETUNING\",33 \"created_at\": \"2024-01-15T09:30:00Z\",34 \"updated_at\": \"2024-01-20T14:45:00Z\",35 \"completed_at\": \"2024-01-25T18:00:00Z\",36 \"last_used\": \"2024-02-01T12:00:00Z\"37 }38}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.342Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":580}}176{"id":"doc-lists_fine_tuned_models_cohere-b50a4e42","source":"documentation","title":"Lists fine-tuned models. | Cohere","url":"https://docs.cohere.com/reference/listfinetunedmodels","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nComma separated list of fields. For example: “created_at,name”. The default sorting order is ascending. To specify descending order for a field, append ” desc” to the field name. For example: “created_at desc,name”. Supported sorting (default)\n\nExample:\n```text\n1import cohere23co = cohere.Client()4response = co.finetuning.list_finetuned_models()5print(response)\n```\n\nExample:\n```text\n1{2 \"finetuned_models\": [3 {4 \"name\": \"chat-ft\",5 \"settings\": {6 \"base_model\": {7 \"base_type\": \"BASE_TYPE_CHAT\",8 \"name\": \"medium\",9 \"version\": \"14.2.0\",10 \"strategy\": \"STRATEGY_TFEW\"11 },12 \"dataset_id\": \"my-dataset-d701tr\",13 \"hyperparameters\": {14 \"early_stopping_patience\": 6,15 \"early_stopping_threshold\": 0.01,16 \"train_batch_size\": 16,17 \"train_epochs\": 1,18 \"learning_rate\": 0.0119 }20 },21 \"id\": \"fee37446-7fc7-42f9-a026-c6ba2fcc422d\",22 \"creator_id\": \"7a317d97-4d05-427d-9396-f31b9fb92c55\",23 \"organization_id\": \"6bdca3d5-3eae-4de0-ac34-786d8063b7ee\",24 \"status\": \"STATUS_READY\",25 \"created_at\": \"2024-01-15T09:30:00Z\",26 \"updated_at\": \"2024-01-15T09:30:00Z\",27 \"completed_at\": \"2024-01-15T09:30:00Z\"28 },29 {30 \"name\": \"rerank-ft\",31 \"settings\": {32 \"base_model\": {33 \"base_type\": \"BASE_TYPE_RERANK\",34 \"name\": \"english\",35 \"version\": \"2.0.0\",36 \"strategy\": \"STRATEGY_VANILLA\"37 },38 \"dataset_id\": \"rerank-dataset-d820xf\"39 },40 \"id\": \"9d927c5e-7598-4772-98b7-cdf2014e8874\",41 \"creator_id\": \"7a317d97-4d05-427d-9396-f31b9fb92c55\",42 \"organization_id\": \"6bdca3d5-3eae-4de0-ac34-786d8063b7ee\",43 \"status\": \"STATUS_READY\",44 \"created_at\": \"2024-01-15T09:30:00Z\",45 \"updated_at\": \"2024-01-15T09:30:00Z\",46 \"completed_at\": \"2024-01-15T09:30:00Z\"47 }48 ]49}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.342Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":531}}177{"id":"doc-classify_cohere-f5f5a3db","source":"documentation","title":"Classify | Cohere","url":"https://docs.cohere.com/reference/classify","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nThis endpoint makes a prediction about which label fits the specified text inputs best. To make a prediction, Classify uses the provided examples of text + label pairs as a reference. models trained on classification examples don’t require the examples parameter to be passed in explicitly.\n\nA list of up to 96 texts to be classified. Each one must be a non-empty string. There is, however, no consistent, universal limit to the length a particular input can be. We perform classification on the first x tokens of each input, and x varies depending on which underlying model is powering classification. The maximum token length for each model is listed in the “max tokens” column here. default the truncate parameter is set to END, so tokens exceeding the limit will be automatically dropped. This behavior can be disabled by setting truncate to NONE, which will result in validation errors for longer texts.\n\nAn array of examples to provide context to the model. Each example is a text string and its associated label/class. Each unique label requires at least 2 examples associated with it; the maximum number of examples is 2500, and each example has a maximum length of 512 tokens. The values should be structured as {text: \"...\",label: \"...\"}. Models trained on classification examples don’t require the examples parameter to be passed in explicitly.\n\nID of a Fine-tuned Classify model\n\nOne of NONE|START|END to specify how the API will handle inputs longer than the maximum token length. Passing START will discard the start of the input. END will discard the end of the input. In both cases, input is discarded until the remaining input is exactly the maximum input token length for the model. If NONE is selected, when the input exceeds the maximum input token length an error will be returned.\n\nThe ID of a custom playground preset. You can create presets in the playground. If you use a preset, all other parameters become optional, and any included parameters will override the preset’s parameters.\n\nExample:\n```text\n1import cohere2from cohere import ClassifyExample34co = cohere.Client()5examples = [6 ClassifyExample(text=\"Dermatologists don't like her!\", label=\"Spam\"),7 ClassifyExample(text=\"'Hello, open to this?'\", label=\"Spam\"),8 ClassifyExample(text=\"I need help please wire me $1000 right now\", label=\"Spam\"),9 ClassifyExample(text=\"Nice to know you ;)\", label=\"Spam\"),10 ClassifyExample(text=\"Please help me?\", label=\"Spam\"),11 ClassifyExample(text=\"Your parcel will be delivered today\", label=\"Not spam\"),12 ClassifyExample(13 text=\"Review changes to our Terms and Conditions\", label=\"Not spam\"14 ),15 ClassifyExample(text=\"Weekly sync notes\", label=\"Not spam\"),16 ClassifyExample(text=\"'Re: Follow up from today's meeting'\", label=\"Not spam\"),17 ClassifyExample(text=\"Pre-read for tomorrow\", label=\"Not spam\"),18]19inputs = [20 \"Confirm your email address\",21 \"hey i need u to send some $\",22]23response = co.classify(24 model=\"<YOUR-FINE-TUNED-MODEL-ID>\",25 inputs=inputs,26 examples=examples,27)28print(response)\n```\n\nExample:\n```text\n1{2 \"id\": \"86886163-b3f3-4e36-8554-60eca7696216\",3 \"classifications\": [4 {5 \"id\": \"842d12fe-934b-4b71-82c2-c581eca00718\",6 \"predictions\": [7 \"Not spam\"8 ],9 \"confidences\": [10 0.566159811 ],12 \"labels\": {13 \"Not spam\": {14 \"confidence\": 0.566159815 },16 \"Spam\": {17 \"confidence\": 0.4338402518 }19 },20 \"classification_type\": \"single-label\",21 \"input\": \"Confirm your email address\",22 \"prediction\": \"Not spam\",23 \"confidence\": 0.566159824 },25 {26 \"id\": \"e1a39b3e-1ecd-41d2-be75-90ed726f7b9e\",27 \"predictions\": [28 \"Spam\"29 ],30 \"confidences\": [31 0.990981132 ],33 \"labels\": {34 \"Not spam\": {35 \"confidence\": 0.00901888336 },37 \"Spam\": {38 \"confidence\": 0.990981139 }40 },41 \"classification_type\": \"single-label\",42 \"input\": \"hey i need u to send some $\",43 \"prediction\": \"Spam\",44 \"confidence\": 0.990981145 }46 ],47 \"meta\": {48 \"api_version\": {49 \"version\": \"1\"50 },51 \"billed_units\": {52 \"classifications\": 253 }54 }55}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.343Z","totalSectionsIncluded":7,"totalCodeBlocksIncluded":2,"totalLines":25,"estimatedTokens":1123}}178{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-6ef7fc6b","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/fine-tuning","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.343Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}179{"id":"doc-retrieve_training_metrics_for_fine_tuned_models_-03765656","source":"documentation","title":"Retrieve training metrics for fine-tuned models. | Cohere","url":"https://docs.cohere.com/reference/listtrainingstepmetrics","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client()4train_step_metrics = co.finetuning.list_training_step_metrics(5 finetuned_model_id=\"test-id\"6)7print(train_step_metrics)\n```\n\nExample:\n```text\n1{2 \"step_metrics\": [3 {4 \"created_at\": \"2024-01-15T09:30:00Z\",5 \"metrics\": {6 \"accuracy\": 0.4557601809501648,7 \"cross_entropy\": 4.264331340789795,8 \"generation_accuracy\": 0.4557601809501648,9 \"generation_cross_entropy\": 4.264331340789795,10 \"step\": 011 }12 },13 {14 \"created_at\": \"2024-01-15T09:30:00Z\",15 \"step_number\": 9,16 \"metrics\": {17 \"accuracy\": 0.7393720149993896,18 \"cross_entropy\": 0.7702581286430359,19 \"generation_accuracy\": 0.7393720149993896,20 \"generation_cross_entropy\": 0.7702581286430359,21 \"step\": 922 }23 }24 ]25}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.343Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":261}}180{"id":"doc-updates_a_fine_tuned_model_cohere-1f32c3ff","source":"documentation","title":"Updates a fine-tuned model. | Cohere","url":"https://docs.cohere.com/reference/updatefinetunedmodel","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from cohere.finetuning import (2 BaseModel,3 Settings,4)5import cohere67co = cohere.Client()8finetuned_model = co.finetuning.update_finetuned_model(9 id=\"test-id\",10 name=\"new name\",11 settings=Settings(12 base_model=BaseModel(13 base_type=\"BASE_TYPE_CHAT\",14 ),15 dataset_id=\"my-dataset-id\",16 ),17)1819print(finetuned_model)\n```\n\nExample:\n```text\n1{2 \"finetuned_model\": {3 \"name\": \"Customer Support Chatbot\",4 \"settings\": {5 \"base_model\": {6 \"base_type\": \"BASE_TYPE_CHAT\",7 \"name\": \"cohere-chat-1\",8 \"version\": \"v1.2.3\",9 \"strategy\": \"STRATEGY_VANILLA\"10 },11 \"dataset_id\": \"customer-support-dataset-2024\",12 \"hyperparameters\": {13 \"early_stopping_patience\": 3,14 \"early_stopping_threshold\": 0.01,15 \"train_batch_size\": 16,16 \"train_epochs\": 5,17 \"learning_rate\": 0.0005,18 \"lora_alpha\": 32,19 \"lora_rank\": 8,20 \"lora_target_modules\": \"LORA_TARGET_MODULES_QKVO\"21 },22 \"multi_label\": false,23 \"wandb\": {24 \"project\": \"customer-support-finetuning\",25 \"api_key\": \"wandb_api_key_1234567890abcdef\",26 \"entity\": \"cohere-team\"27 }28 },29 \"id\": \"ftm-1234567890abcdef\",30 \"creator_id\": \"user-9876543210fedcba\",31 \"organization_id\": \"org-1122334455aabbcc\",32 \"status\": \"STATUS_READY\",33 \"created_at\": \"2024-01-15T09:30:00Z\",34 \"updated_at\": \"2024-02-20T14:45:00Z\",35 \"completed_at\": \"2024-02-15T12:00:00Z\",36 \"last_used\": \"2024-04-10T08:15:00Z\"37 }38}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.343Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":445}}181{"id":"doc-calendar_agent_with_native_multi_step_tool_coher-d6d7ffa1","source":"documentation","title":"Calendar Agent with Native Multi Step Tool | Cohere","url":"https://docs.cohere.com/page/calendar-agent","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# !pip install cohere==5.5.3\n```\n\nExample:\n```text\n1# Instantiate the Cohere client23import cohere4import os56COHERE_API_KEY = os.environ[\"COHERE_API_KEY\"]7co = cohere.Client(api_key=COHERE_API_KEY)\n```\n\nExample:\n```text\n1# Define the tools23import json45def list_calendar_events(date: str):6 events = '[{\"start\": \"14:00\", \"end\": \"15:00\"}, {\"start\": \"15:00\", \"end\": \"16:00\"}, {\"start\": \"17:00\", \"end\": \"18:00\"}]'7 print(f\"Listing events: {events}\")8 return events910def create_calendar_event(date: str, time: str, duration: int):11 print(f\"Creating a {duration} hour long event at {time} on {date}\")12 return True1314list_calendar_events_tool = {15 \"name\": \"list_calendar_events\",16 \"description\": \"returns a list of calendar events for the specified date, including the start time and end time for each event\",17 \"parameter_definitions\": {18 \"date\": {19 \"description\": \"the date to list events for, formatted as mm/dd/yy\",20 \"type\": \"str\",21 \"required\": True22 }23 }24}2526create_calendar_event_tool = {27 \"name\": \"create_calendar_event_tool\",28 \"description\": \"creates a calendar event of the specified duration at the specified time and date\",29 \"parameter_definitions\": {30 \"date\": {31 \"description\": \"the date on which the event starts, formatted as mm/dd/yy\",32 \"type\": \"str\",33 \"required\": True34 },35 \"time\": {36 \"description\": \"the time of the event, formatted using 24h military time formatting\",37 \"type\": \"str\",38 \"required\": True39 },40 \"duration\": {41 \"description\": \"the number of hours the event lasts for\",42 \"type\": \"float\",43 \"required\": True44 }45 }46}4748# helper function for routing to the correct tool49def invoke_tool(tool_call: cohere.ToolCall):50 if tool_call.name == list_calendar_events_tool[\"name\"]:51 date = tool_call.parameters[\"date\"]52 return [{53 \"events\": list_calendar_events(date)54 }]55 elif tool_call.name == create_calendar_event_tool[\"name\"]:56 date = tool_call.parameters[\"date\"]57 time = tool_call.parameters[\"time\"]58 duration = tool_call.parameters[\"duration\"]5960 return [{61 \"is_success\": create_calendar_event(date, time, duration)62 }]63 else:64 raise f\"Unknown tool name '{tool_call.name}'\"\n```\n\nExample:\n```text\n1# Check what tools the model wants to use and how to use them2res = co.chat(3 model=\"command-a-03-2025\",4 preamble=\"Today is Thursday, may 23, 2024\",5 message=\"book an hour long appointment for the first available free slot after 3pm\",6 force_single_step=False,7 tools=[list_calendar_events_tool, create_calendar_event_tool])89while res.tool_calls:10 print(res.text) # This will be an observation and a plan with next steps1112 # invoke the recommended tools13 tool_results = []14 for call in res.tool_calls:15 tool_results.append({\"call\": call, \"outputs\": invoke_tool(call)})1617 # send back the tool results18 res = co.chat(19 model=\"command-a-03-2025\",20 chat_history=res.chat_history,21 message=\"\",22 force_single_step=False,23 tools=[list_calendar_events_tool, create_calendar_event_tool],24 tool_results=tool_results,25 )2627print(res.text) # print the final answer\n```\n\nExample:\n```text\nI will check the user's calendar for today after 3pm and book an hour-long appointment in the first available slot.Listing events: [{\"start\": \"14:00\", \"end\": \"15:00\"}, {\"start\": \"15:00\", \"end\": \"16:00\"}, {\"start\": \"17:00\", \"end\": \"18:00\"}]The user has events scheduled from 2pm to 4pm and from 5pm to 6pm. I will book an hour-long appointment from 4pm to 5pm.Creating a 1 hour long event at 16:00 on 05/23/2024I've booked an hour-long appointment for you today from 4pm to 5pm.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.343Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":5,"totalLines":28,"estimatedTokens":982}}182{"id":"doc-basic_semantic_search_with_cohere_models_cohere-a9fbcdc0","source":"documentation","title":"Basic Semantic Search with Cohere Models | Cohere","url":"https://docs.cohere.com/page/basic-semantic-search","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1#!pip install --upgrade cohere\n```\n\nExample:\n```text\n1#@title Import libraries (Run this cell to execute required code) {display-mode: \"form\"}23import cohere4import numpy as np5import re6import pandas as pd7from tqdm import tqdm8from datasets import load_dataset9import umap10import altair as alt11from sklearn.metrics.pairwise import cosine_similarity12from annoy import AnnoyIndex13import warnings14warnings.filterwarnings('ignore')15pd.set_option('display.max_colwidth', None)\n```\n\nExample:\n```text\n1model_name = \"embed-v4.0\"2api_key = \"\"3input_type_embed = \"search_document\"45co = cohere.Client(api_key)\n```\n\nExample:\n```text\n1dataset = load_dataset(\"trec\", split=\"train\")23df = pd.DataFrame(dataset)[:1000]45df.head(10)\n```\n\nExample:\n```text\n1embeds = co.embed(texts=list(df['text']),2 model=model_name,3 input_type=input_type_embed).embeddings\n```\n\nExample:\n```text\n1embeds = np.array(embeds)2embeds.shape\n```\n\nExample:\n```text\n(1000, 4096)\n```\n\nExample:\n```text\n1search_index = AnnoyIndex(embeds.shape[1], 'angular')2for i in range(len(embeds)):3 search_index.add_item(i, embeds[i])45search_index.build(10) # 10 trees6search_index.save('test.ann')\n```\n\nExample:\n```text\nTrue\n```\n\nExample:\n```text\n1example_id = 9223similar_item_ids = search_index.get_nns_by_item(example_id,10,4 include_distances=True)5results = pd.DataFrame(data={'texts': df.iloc[similar_item_ids[0]]['text'],6 'distance': similar_item_ids[1]}).drop(example_id)78print(f\"Question:'{df.iloc[example_id]['text']}'\\nNearest neighbors:\")9results\n```\n\nExample:\n```text\nQuestion:'What are bear and bull markets ?'Nearest neighbors:\n```\n\nExample:\n```text\n1query = \"What is the tallest mountain in the world?\"2input_type_query = \"search_query\"34query_embed = co.embed(texts=[query],5 model=model_name,6 input_type=input_type_query).embeddings78similar_item_ids = search_index.get_nns_by_vector(query_embed[0],10,9 include_distances=True)10query_results = pd.DataFrame(data={'texts': df.iloc[similar_item_ids[0]]['text'],11 'distance': similar_item_ids[1]})121314print(f\"Query:'{query}'\\nNearest neighbors:\")15print(query_results) # NOTE: Your results might look slightly different to ours.\n```\n\nExample:\n```text\nQuery:'What is the tallest mountain in the world?'Nearest neighbors:\n```\n\nExample:\n```text\n1#@title Plot the archive {display-mode: \"form\"}23reducer = umap.UMAP(n_neighbors=20)4umap_embeds = reducer.fit_transform(embeds)5df_explore = pd.DataFrame(data={'text': df['text']})6df_explore['x'] = umap_embeds[:,0]7df_explore['y'] = umap_embeds[:,1]89chart = alt.Chart(df_explore).mark_circle(size=60).encode(10 x=#'x',11 alt.X('x',12 scale=alt.Scale(zero=False)13 ),14 y=15 alt.Y('y',16 scale=alt.Scale(zero=False)17 ),18 tooltip=['text']19).properties(20 width=700,21 height=40022)23chart.interactive()\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.344Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":14,"totalLines":73,"estimatedTokens":812}}183{"id":"doc-migrating_from_api_v1_to_api_v2_cohere-78d2a44d","source":"documentation","title":"Migrating From API v1 to API v2 | Cohere","url":"https://docs.cohere.com/docs/migrating-v1-to-v2","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install -U cohere23import cohere45# instantiating the old client6co_v1 = cohere.Client(api_key=\"<YOUR API KEY>\")78# instantiating the new client9co_v2 = cohere.ClientV2(api_key=\"<YOUR API KEY>\")\n```\n\nExample:\n```text\n1res = co_v1.chat(2 model=\"command-a-03-2025\",3 preamble=\"You respond in concise sentences.\",4 chat_history=[5 {\"role\": \"user\", \"message\": \"Hello\"},6 {7 \"role\": \"chatbot\",8 \"message\": \"Hi, how can I help you today?\",9 },10 ],11 message=\"I'm joining a new startup called Co1t today. Could you help me write a one-sentence introduction message to my teammates?\",12)1314print(res.text)\n```\n\nExample:\n```text\nExcited to join the team at Co1t, where I look forward to contributing my skills and collaborating with everyone to drive innovation and success.\n```\n\nExample:\n```text\n1res = co_v2.chat(2 model=\"command-a-plus-05-2026\",3 messages=[4 {5 \"role\": \"system\",6 \"content\": \"You respond in concise sentences.\",7 },8 {\"role\": \"user\", \"content\": \"Hello\"},9 {10 \"role\": \"assistant\",11 \"content\": \"Hi, how can I help you today?\",12 },13 {14 \"role\": \"user\",15 \"content\": \"I'm joining a new startup called Co1t today. Could you help me write a one-sentence introduction message to my teammates.\",16 },17 ],18)1920print(res.message.content[0].text)\n```\n\nExample:\n```text\nExcited to join the team at Co1t, bringing my passion for innovation and a background in [your expertise] to contribute to the company's success!\n```\n\nExample:\n```text\n1res = co_v1.chat(model=\"command-a-03-2025\", message=\"What is 2 + 2\")23print(res.text)\n```\n\nExample:\n```text\nThe answer is 4.\n```\n\nExample:\n```text\n1res = co_v2.chat(2 model=\"command-a-plus-05-2026\",3 messages=[{\"role\": \"user\", \"content\": \"What is 2 + 2\"}],4)56print(res.message.content[0].text)\n```\n\nExample:\n```text\n1message = \"I'm joining a new startup called Co1t today. Could you help me write a one-sentence introduction message to my teammates.\"23res = co_v1.chat_stream(model=\"command-a-03-2025\", message=message)45for chunk in res:6 if chunk.event_type == \"text-generation\":7 print(chunk.text, end=\"\")\n```\n\nExample:\n```text\n\"Hi, I'm [your name] and I'm thrilled to join the Co1t team today as a [your role], eager to contribute my skills and ideas to help drive innovation and success for our startup!\"\n```\n\nExample:\n```text\n1message = \"I'm joining a new startup called Co1t today. Could you help me write a one-sentence introduction message to my teammates.\"23res = co_v2.chat_stream(4 model=\"command-a-plus-05-2026\",5 messages=[{\"role\": \"user\", \"content\": message}],6)78for chunk in res:9 if chunk:10 if chunk.type == \"content-delta\":11 print(chunk.delta.message.content.text, end=\"\")\n```\n\nExample:\n```text\n\"Hi everyone, I'm thrilled to join the Co1t team today and look forward to contributing my skills and ideas to drive innovation and success!\"\n```\n\nExample:\n```text\n1# Define the documents2documents_v1 = [3 {4 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"5 },6 {7 \"text\": \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\"8 },9]1011# The user query12message = \"Are there fitness-related benefits?\"1314# Generate the response15res_v1 = co_v1.chat(16 model=\"command-a-03-2025\",17 message=message,18 documents=documents_v1,19)2021print(res_v1.text)\n```\n\nExample:\n```text\nYes, there are fitness-related benefits. We offer gym memberships, on-site yoga classes, and comprehensive health insurance.\n```\n\nExample:\n```text\n1# Define the documents2documents_v2 = [3 {4 \"data\": {5 \"text\": \"Reimbursing Travel Expenses: Easily manage your travel expenses by submitting them through our finance tool. Approvals are prompt and straightforward.\"6 }7 },8 {9 \"data\": {10 \"text\": \"Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.\"11 }12 },13]1415# The user query16message = \"Are there fitness-related benefits?\"1718# Generate the response19res_v2 = co_v2.chat(20 model=\"command-a-plus-05-2026\",21 messages=[{\"role\": \"user\", \"content\": message}],22 documents=documents_v2,23)2425print(res_v2.message.content[0].text)\n```\n\nExample:\n```text\nYes, we offer gym memberships, on-site yoga classes, and comprehensive health insurance.\n```\n\nExample:\n```text\n1documents_v2 = [2 # List of objects with data string3 {4 \"id\": \"123\",5 \"data\": \"I love penguins. they are fluffy\",6 },7 # List of objects with data object8 {9 \"id\": \"456\",10 \"data\": {11 \"text\": \"I love penguins. they are fluffy\",12 \"author\": \"Abdullah\",13 \"create_date\": \"09021989\",14 },15 },16 # List of strings17 \"just a string\",18]\n```\n\nExample:\n```text\n1# Yes, there are fitness-related benefits. We offer gym memberships, on-site yoga classes, and comprehensive health insurance.23print(res_v1.citations)4print(res_v1.documents)\n```\n\nExample:\n```text\n[ChatCitation(start=50, end=124, text='gym memberships, on-site yoga classes, and comprehensive health insurance.', document_ids=['doc_1'])][{'id': 'doc_1', 'text': 'Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.'}]\n```\n\nExample:\n```text\n1# Yes, we offer gym memberships, on-site yoga classes, and comprehensive health insurance.23print(res_v2.message.citations)\n```\n\nExample:\n```text\n[Citation(start=14, end=88, text='gym memberships, on-site yoga classes, and comprehensive health insurance.', sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.'})])]\n```\n\nExample:\n```text\n1res_v1 = co_v1.chat(2 message=\"who won euro 2024\",3 connectors=[{\"id\": \"web-search\"}],4)56print(res_v1.text)\n```\n\nExample:\n```text\nSpain won the UEFA Euro 2024, defeating England 2-1 in the final.\n```\n\nExample:\n```text\n1# Any search engine can be used. This example uses the Tavily API.2from tavily import TavilyClient34tavily_client = TavilyClient(api_key=os.environ[\"TAVILY_API_KEY\"])567# Create a web search function8def web_search(queries: list[str]) -> list[dict]:910 documents = []1112 for query in queries:13 response = tavily_client.search(query, max_results=2)1415 results = [16 {17 \"title\": r[\"title\"],18 \"content\": r[\"content\"],19 \"url\": r[\"url\"],20 }21 for r in response[\"results\"]22 ]2324 for idx, result in enumerate(results):25 document = {\"id\": str(idx), \"data\": result}26 documents.append(document)2728 return documents293031# Define the web search tool32web_search_tool = [33 {34 \"type\": \"function\",35 \"function\": {36 \"name\": \"web_search\",37 \"description\": \"Returns a list of relevant document snippets for a textual query retrieved from the internet\",38 \"parameters\": {39 \"type\": \"object\",40 \"properties\": {41 \"queries\": {42 \"type\": \"array\",43 \"items\": {\"type\": \"string\"},44 \"description\": \"a list of queries to search the internet with.\",45 }46 },47 \"required\": [\"queries\"],48 },49 },50 }51]5253# The user query54query = \"who won euro 2024\"5556# Define a system message to optimize search query generation57instructions = \"Write a search query that will find helpful information for answering the user's question accurately. If you need more than one search query, write a list of search queries. If you decide that a search is very unlikely to find information that would be useful in constructing a response to the user, you should instead directly answer.\"5859messages = [60 {\"role\": \"system\", \"content\": instructions},61 {\"role\": \"user\", \"content\": query},62]6364model = \"command-a-plus-05-2026\"6566# Generate search queries (if any)67response = co_v2.chat(68 model=model, messages=messages, tools=web_search_tool69)7071search_queries = []7273while response.message.tool_calls:7475 print(\"Tool plan:\")76 print(response.message.tool_plan, \"\\n\")77 print(\"Tool calls:\")78 for tc in response.message.tool_calls:79 print(80 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"81 )82 print(\"=\" * 50)8384 messages.append(response.message)8586 # Step 3: Get tool results87 for idx, tc in enumerate(response.message.tool_calls):88 tool_result = web_search(**json.loads(tc.function.arguments))89 tool_content = []90 for data in tool_result:91 tool_content.append(92 {93 \"type\": \"document\",94 \"document\": {\"data\": json.dumps(data)},95 }96 )97 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated98 messages.append(99 {100 \"role\": \"tool\",101 \"tool_call_id\": tc.id,102 \"content\": tool_content,103 }104 )105106 # Step 4: Generate response and citations107 response = co_v2.chat(108 model=model, messages=messages, tools=web_search_tool109 )110111print(response.message.content[0].text)\n```\n\nExample:\n```text\nTool plan:I will search for 'who won euro 2024' to find out who won the competition. Tool calls:Tool name: web_search | Parameters: {\"queries\":[\"who won euro 2024\"]}==================================================Spain won the 2024 European Championship. They beat England in the final, with substitute Mikel Oyarzabal scoring the winning goal.\n```\n\nExample:\n```text\n1message = \"Are there fitness-related benefits?\"23res_v1 = co_v1.chat_stream(4 model=\"command-a-03-2025\",5 message=message,6 documents=documents_v1,7)89for chunk in res_v1:10 if chunk.event_type == \"text-generation\":11 print(chunk.text, end=\"\")12 if chunk.event_type == \"citation-generation\":13 print(f\"\\n{chunk.citations}\")\n```\n\nExample:\n```text\nYes, we offer gym memberships, on-site yoga classes, and comprehensive health insurance as part of our health and wellness benefits.[ChatCitation(start=14, end=87, text='gym memberships, on-site yoga classes, and comprehensive health insurance', document_ids=['doc_1'])][ChatCitation(start=103, end=132, text='health and wellness benefits.', document_ids=['doc_1'])]\n```\n\nExample:\n```text\n1message = \"Are there fitness-related benefits?\"23messages = [{\"role\": \"user\", \"content\": message}]45res_v2 = co_v2.chat_stream(6 model=\"command-a-plus-05-2026\",7 messages=messages,8 documents=documents_v2,9)1011for chunk in res_v2:12 if chunk:13 if chunk.type == \"content-delta\":14 print(chunk.delta.message.content.text, end=\"\")15 if chunk.type == \"citation-start\":16 print(f\"\\n{chunk.delta.message.citations}\")\n```\n\nExample:\n```text\nYes, we offer gym memberships, on-site yoga classes, and comprehensive health insurance.start=14 end=88 text='gym memberships, on-site yoga classes, and comprehensive health insurance.' sources=[DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': 'Health and Wellness Benefits: We care about your well-being and offer gym memberships, on-site yoga classes, and comprehensive health insurance.'})]\n```\n\nExample:\n```text\n1def get_weather(location):2 return {\"temperature\": \"20C\"}345functions_map = {\"get_weather\": get_weather}67tools_v1 = [8 {9 \"name\": \"get_weather\",10 \"description\": \"Gets the weather of a given location\",11 \"parameter_definitions\": {12 \"location\": {13 \"description\": \"The location to get weather, example: San Francisco, CA\",14 \"type\": \"str\",15 \"required\": True,16 }17 },18 },19]\n```\n\nExample:\n```text\n1def get_weather(location):2 return [{\"temperature\": \"20C\"}]3 # You can return a list of objects e.g. [{\"url\": \"abc.com\", \"text\": \"...\"}, {\"url\": \"xyz.com\", \"text\": \"...\"}]456functions_map = {\"get_weather\": get_weather}78tools_v2 = [9 {10 \"type\": \"function\",11 \"function\": {12 \"name\": \"get_weather\",13 \"description\": \"gets the weather of a given location\",14 \"parameters\": {15 \"type\": \"object\",16 \"properties\": {17 \"location\": {18 \"type\": \"string\",19 \"description\": \"the location to get weather, example: San Fransisco, CA\",20 }21 },22 \"required\": [\"location\"],23 },24 },25 },26]\n```\n\nExample:\n```text\n1message = \"What's the weather in Toronto?\"23res_v1 = co_v1.chat(4 model=\"command-a-03-2025\", message=message, tools=tools_v15)67print(res_v1.tool_calls)\n```\n\nExample:\n```text\n[ToolCall(name='get_weather', parameters={'location': 'Toronto'})]\n```\n\nExample:\n```text\n1messages = [2 {\"role\": \"user\", \"content\": \"What's the weather in Toronto?\"}3]45res_v2 = co_v2.chat(6 model=\"command-a-plus-05-2026\", messages=messages, tools=tools_v27)89if res_v2.message.tool_calls:10 messages.append(res_v2.message)1112 print(res_v2.message.tool_calls)\n```\n\nExample:\n```text\n[ToolCallV2(id='get_weather_k88p0m8504w5', type='function', function=ToolCallV2Function(name='get_weather', arguments='{\"location\":\"Toronto\"}'))]\n```\n\nExample:\n```text\n1tool_results = [2 {3 \"call\": {4 \"name\": \"<tool name>\",5 \"parameters\": {\"<param name>\": \"<param value>\"},6 },7 \"outputs\": [{\"<key>\": \"<value>\"}],8 },9]\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"tool\",4 \"tool_call_id\": \"123\",5 \"content\": [6 {7 \"type\": \"document\",8 \"document\": {9 \"id\": \"123\",10 \"data\": {\"<key>\": \"<value>\"},11 },12 }13 ],14 }15]\n```\n\nExample:\n```text\n1tool_content_v1 = []2if res_v1.tool_calls:3 for tc in res_v1.tool_calls:4 tool_call = {\"name\": tc.name, \"parameters\": tc.parameters}5 tool_result = functions_map[tc.name](**tc.parameters)6 tool_content_v1.append(7 {\"call\": tool_call, \"outputs\": [tool_result]}8 )910res_v1 = co_v1.chat(11 model=\"command-a-03-2025\",12 message=\"\",13 tools=tools_v1,14 tool_results=tool_content_v1,15 chat_history=res_v1.chat_history,16)1718print(res_v1.text)\n```\n\nExample:\n```text\nIt is currently 20°C in Toronto.\n```\n\nExample:\n```text\n1if res_v2.message.tool_calls:2 for tc in res_v2.message.tool_calls:3 tool_result = functions_map[tc.function.name](4 **json.loads(tc.function.arguments)5 )6 tool_content_v2 = []7 for data in tool_result:8 tool_content_v2.append(9 {10 \"type\": \"document\",11 \"document\": {\"data\": json.dumps(data)},12 }13 )14 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated15 messages.append(16 {17 \"role\": \"tool\",18 \"tool_call_id\": tc.id,19 \"content\": tool_content_v2,20 }21 )2223res_v2 = co_v2.chat(24 model=\"command-a-plus-05-2026\", messages=messages, tools=tools_v225)2627print(res_v2.message.content[0].text)\n```\n\nExample:\n```text\nIt's 20°C in Toronto.\n```\n\nExample:\n```text\n1print(res_v1.citations)2print(res_v1.documents)\n```\n\nExample:\n```text\n[ChatCitation(start=16, end=20, text='20°C', document_ids=['get_weather:0:2:0'])][{'id': 'get_weather:0:2:0', 'temperature': '20C', 'tool_name': 'get_weather'}]\n```\n\nExample:\n```text\n1print(res_v2.message.citations)\n```\n\nExample:\n```text\n[Citation(start=5, end=9, text='20°C', sources=[ToolSource(type='tool', id='get_weather_k88p0m8504w5:0', tool_output={'temperature': '20C'})])]\n```\n\nExample:\n```text\n1tool_content_v1 = []2if res_v1.tool_calls:3 for tc in res_v1.tool_calls:4 tool_call = {\"name\": tc.name, \"parameters\": tc.parameters}5 tool_result = functions_map[tc.name](**tc.parameters)6 tool_content_v1.append(7 {\"call\": tool_call, \"outputs\": [tool_result]}8 )910res_v1 = co_v1.chat_stream(11 message=\"\",12 tools=tools_v1,13 tool_results=tool_content_v1,14 chat_history=res_v1.chat_history,15)1617for chunk in res_v1:18 if chunk.event_type == \"text-generation\":19 print(chunk.text, end=\"\")20 if chunk.event_type == \"citation-generation\":21 print(f\"\\n{chunk.citations}\")\n```\n\nExample:\n```text\nIt's 20°C in Toronto.[ChatCitation(start=5, end=9, text='20°C', document_ids=['get_weather:0:2:0', 'get_weather:0:4:0'])]\n```\n\nExample:\n```text\n1if res_v2.message.tool_calls:2 for tc in res_v2.message.tool_calls:3 tool_result = functions_map[tc.function.name](4 **json.loads(tc.function.arguments)5 )6 tool_content_v2 = []7 for data in tool_result:8 tool_content_v2.append(9 {10 \"type\": \"document\",11 \"document\": {\"data\": json.dumps(data)},12 }13 )14 # Optional: add an \"id\" field in the \"document\" object, otherwise IDs are auto-generated15 messages.append(16 {17 \"role\": \"tool\",18 \"tool_call_id\": tc.id,19 \"content\": tool_content_v2,20 }21 )2223res_v2 = co_v2.chat_stream(24 model=\"command-a-plus-05-2026\", messages=messages, tools=tools_v225)2627for chunk in res_v2:28 if chunk:29 if chunk.type == \"content-delta\":30 print(chunk.delta.message.content.text, end=\"\")31 elif chunk.type == \"citation-start\":32 print(f\"\\n{chunk.delta.message.citations}\")\n```\n\nExample:\n```text\nIt's 20°C in Toronto.start=5 end=9 text='20°C' sources=[ToolSource(type='tool', id='get_weather_k88p0m8504w5:0', tool_output={'temperature': '20C'})]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.345Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":49,"totalLines":248,"estimatedTokens":4729}}184{"id":"doc-analyzing_hacker_news_with_cohere_cohere-7ac22ce2","source":"documentation","title":"Analyzing Hacker News with Cohere | Cohere","url":"https://docs.cohere.com/page/analyzing-hacker-news","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1!pip install cohere umap-learn altair annoy bertopic\n```\n\nExample:\n```text\nRequirement already satisfied: cohere in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (5.1.5)Requirement already satisfied: umap-learn in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (0.5.5)Requirement already satisfied: altair in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (5.2.0)Requirement already satisfied: annoy in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (1.17.3)Requirement already satisfied: bertopic in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (0.16.0)Requirement already satisfied: httpx>=0.21.2 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from cohere) (0.27.0)Requirement already satisfied: pydantic>=1.9.2 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from cohere) (2.6.0)Requirement already satisfied: typing_extensions>=4.0.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from cohere) (4.10.0)Requirement already satisfied: numpy>=1.17 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from umap-learn) (1.24.3)Requirement already satisfied: scipy>=1.3.1 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from umap-learn) (1.11.1)Requirement already satisfied: scikit-learn>=0.22 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from umap-learn) (1.3.0)Requirement already satisfied: numba>=0.51.2 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from umap-learn) (0.57.0)Requirement already satisfied: pynndescent>=0.5 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from umap-learn) (0.5.12)Requirement already satisfied: tqdm in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from umap-learn) (4.65.0)Requirement already satisfied: jinja2 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from altair) (3.1.2)Requirement already satisfied: jsonschema>=3.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from altair) (4.17.3)Requirement already satisfied: packaging in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from altair) (23.2)Requirement already satisfied: pandas>=0.25 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from altair) (2.0.3)Requirement already satisfied: toolz in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from altair) (0.12.0)Requirement already satisfied: hdbscan>=0.8.29 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from bertopic) (0.8.33)Requirement already satisfied: sentence-transformers>=0.4.1 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from bertopic) (2.6.1)Requirement already satisfied: plotly>=4.7.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from bertopic) (5.9.0)Requirement already satisfied: cython<3,>=0.27 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from hdbscan>=0.8.29->bertopic) (0.29.37)Requirement already satisfied: joblib>=1.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from hdbscan>=0.8.29->bertopic) (1.2.0)Requirement already satisfied: anyio in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from httpx>=0.21.2->cohere) (3.5.0)Requirement already satisfied: certifi in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from httpx>=0.21.2->cohere) (2023.11.17)Requirement already satisfied: httpcore==1.* in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from httpx>=0.21.2->cohere) (1.0.2)Requirement already satisfied: idna in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from httpx>=0.21.2->cohere) (3.4)Requirement already satisfied: sniffio in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from httpx>=0.21.2->cohere) (1.2.0)Requirement already satisfied: h11<0.15,>=0.13 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from httpcore==1.*->httpx>=0.21.2->cohere) (0.14.0)Requirement already satisfied: attrs>=17.4.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from jsonschema>=3.0->altair) (22.1.0)Requirement already satisfied: pyrsistent!=0.17.0,!=0.17.1,!=0.17.2,>=0.14.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from jsonschema>=3.0->altair) (0.18.0)Requirement already satisfied: llvmlite<0.41,>=0.40.0dev0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from numba>=0.51.2->umap-learn) (0.40.0)Requirement already satisfied: python-dateutil>=2.8.2 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from pandas>=0.25->altair) (2.8.2)Requirement already satisfied: pytz>=2020.1 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from pandas>=0.25->altair) (2023.3.post1)Requirement already satisfied: tzdata>=2022.1 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from pandas>=0.25->altair) (2023.3)Requirement already satisfied: tenacity>=6.2.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from plotly>=4.7.0->bertopic) (8.2.2)Requirement already satisfied: annotated-types>=0.4.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from pydantic>=1.9.2->cohere) (0.6.0)Requirement already satisfied: pydantic-core==2.16.1 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from pydantic>=1.9.2->cohere) (2.16.1)Requirement already satisfied: threadpoolctl>=2.0.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from scikit-learn>=0.22->umap-learn) (2.2.0)Requirement already satisfied: transformers<5.0.0,>=4.32.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from sentence-transformers>=0.4.1->bertopic) (4.39.3)Requirement already satisfied: torch>=1.11.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from sentence-transformers>=0.4.1->bertopic) (2.2.2)Requirement already satisfied: huggingface-hub>=0.15.1 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from sentence-transformers>=0.4.1->bertopic) (0.22.2)Requirement already satisfied: Pillow in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from sentence-transformers>=0.4.1->bertopic) (10.0.1)Requirement already satisfied: MarkupSafe>=2.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from jinja2->altair) (2.1.1)Requirement already satisfied: filelock in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from huggingface-hub>=0.15.1->sentence-transformers>=0.4.1->bertopic) (3.9.0)Requirement already satisfied: fsspec>=2023.5.0 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from huggingface-hub>=0.15.1->sentence-transformers>=0.4.1->bertopic) (2024.3.1)Requirement already satisfied: pyyaml>=5.1 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from huggingface-hub>=0.15.1->sentence-transformers>=0.4.1->bertopic) (6.0)Requirement already satisfied: requests in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from huggingface-hub>=0.15.1->sentence-transformers>=0.4.1->bertopic) (2.31.0)Requirement already satisfied: six>=1.5 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from python-dateutil>=2.8.2->pandas>=0.25->altair) (1.16.0)Requirement already satisfied: sympy in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from torch>=1.11.0->sentence-transformers>=0.4.1->bertopic) (1.11.1)Requirement already satisfied: networkx in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from torch>=1.11.0->sentence-transformers>=0.4.1->bertopic) (3.1)Requirement already satisfied: regex!=2019.12.17 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from transformers<5.0.0,>=4.32.0->sentence-transformers>=0.4.1->bertopic) (2022.7.9)Requirement already satisfied: tokenizers<0.19,>=0.14 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from transformers<5.0.0,>=4.32.0->sentence-transformers>=0.4.1->bertopic) (0.15.2)Requirement already satisfied: safetensors>=0.4.1 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from transformers<5.0.0,>=4.32.0->sentence-transformers>=0.4.1->bertopic) (0.4.2)Requirement already satisfied: charset-normalizer<4,>=2 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from requests->huggingface-hub>=0.15.1->sentence-transformers>=0.4.1->bertopic) (3.3.2)Requirement already satisfied: urllib3<3,>=1.21.1 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from requests->huggingface-hub>=0.15.1->sentence-transformers>=0.4.1->bertopic) (1.26.18)Requirement already satisfied: mpmath>=0.19 in /Users/alexiscook/anaconda3/lib/python3.11/site-packages (from sympy->torch>=1.11.0->sentence-transformers>=0.4.1->bertopic) (1.3.0)\n```\n\nExample:\n```text\n1import cohere2import numpy as np3import pandas as pd4import umap5import altair as alt6from annoy import AnnoyIndex7import warnings8from sklearn.cluster import KMeans9from sklearn.feature_extraction.text import CountVectorizer10from bertopic.vectorizers import ClassTfidfTransformer1112warnings.filterwarnings('ignore')13pd.set_option('display.max_colwidth', None)\n```\n\nExample:\n```text\n1co = cohere.Client(\"COHERE_API_KEY\") # Insert your Cohere API key\n```\n\nExample:\n```text\n1df = pd.read_csv('https://storage.googleapis.com/cohere-assets/blog/text-clustering/data/askhn3k_df.csv', index_col=0)23print(f'Loaded a DataFrame with {len(df)} rows')\n```\n\nExample:\n```text\nLoaded a DataFrame with 3000 rows\n```\n\nExample:\n```text\n1df.head()\n```\n\nExample:\n```text\n1batch_size = 9023embeds_list = []4for i in range(0, len(df), batch_size):5 batch = df[i : min(i + batch_size, len(df))]6 texts = list(batch[\"title\"])7 embs_batch = co.embed(8 texts=texts, model=\"embed-v4.0\", input_type=\"search_document\"9 ).embeddings10 embeds_list.extend(embs_batch)1112embeds = np.array(embeds_list)13embeds.shape\n```\n\nExample:\n```text\n(3000, 1024)\n```\n\nExample:\n```text\n1search_index = AnnoyIndex(embeds.shape[1], 'angular')2for i in range(len(embeds)):3 search_index.add_item(i, embeds[i])45search_index.build(10) # 10 trees6search_index.save('askhn.ann')\n```\n\nExample:\n```text\nTrue\n```\n\nExample:\n```text\n1example_id = 5023similar_item_ids = search_index.get_nns_by_item(example_id,4 10, # Number of results to retrieve5 include_distances=True)6results = pd.DataFrame(data={'post titles': df.iloc[similar_item_ids[0]]['title'],7 'distance': similar_item_ids[1]}).drop(example_id)89print(f\"Query post:'{df.iloc[example_id]['title']}'\\nNearest neighbors:\")10results\n```\n\nExample:\n```text\nQuery post:'Pick startups for YC to fund'Nearest neighbors:\n```\n\nExample:\n```text\n1query = \"How can I improve my knowledge of calculus?\"23query_embed = co.embed(texts=[query],4 model=\"embed-v4.0\",5 truncate=\"RIGHT\",6 input_type=\"search_query\").embeddings78similar_item_ids = search_index.get_nns_by_vector(query_embed[0], 10, include_distances=True)910results = pd.DataFrame(data={'texts': df.iloc[similar_item_ids[0]]['title'],11 'distance': similar_item_ids[1]})12print(f\"Query:'{query}'\\nNearest neighbors:\")13results\n```\n\nExample:\n```text\nQuery:'How can I improve my knowledge of calculus?'Nearest neighbors:\n```\n\nExample:\n```text\n1reducer = umap.UMAP(n_neighbors=100)2umap_embeds = reducer.fit_transform(embeds)\n```\n\nExample:\n```text\n1df['x'] = umap_embeds[:,0]2df['y'] = umap_embeds[:,1]34chart = alt.Chart(df).mark_circle(size=60).encode(5 x=#'x',6 alt.X('x',7 scale=alt.Scale(zero=False),8 axis=alt.Axis(labels=False, ticks=False, domain=False)9 ),10 y=11 alt.Y('y',12 scale=alt.Scale(zero=False),13 axis=alt.Axis(labels=False, ticks=False, domain=False)14 ),15 tooltip=['title']16 ).configure(background=\"#FDF7F0\"17 ).properties(18 width=700,19 height=400,20 title='Ask HN: top 3,000 posts'21 )2223chart.interactive()\n```\n\nExample:\n```text\n1n_clusters = 823kmeans_model = KMeans(n_clusters=n_clusters, random_state=0)4classes = kmeans_model.fit_predict(embeds)\n```\n\nExample:\n```text\n1documents = df['title']2documents = pd.DataFrame({\"Document\": documents,3 \"ID\": range(len(documents)),4 \"Topic\": None})5documents['Topic'] = classes6documents_per_topic = documents.groupby(['Topic'], as_index=False).agg({'Document': ' '.join})7count_vectorizer = CountVectorizer(stop_words=\"english\").fit(documents_per_topic.Document)8count = count_vectorizer.transform(documents_per_topic.Document)9words = count_vectorizer.get_feature_names_out()\n```\n\nExample:\n```text\n1ctfidf = ClassTfidfTransformer().fit_transform(count).toarray()2words_per_class = {label: [words[index] for index in ctfidf[label].argsort()[-10:]] for label in documents_per_topic.Topic}3df['cluster'] = classes4df['keywords'] = df['cluster'].map(lambda topic_num: \", \".join(np.array(words_per_class[topic_num])[:]))\n```\n\nExample:\n```text\n1selection = alt.selection_multi(fields=['keywords'], bind='legend')23chart = alt.Chart(df).transform_calculate(4 url='https://news.ycombinator.com/item?id=' + alt.datum.id5).mark_circle(size=60, stroke='#666', strokeWidth=1, opacity=0.3).encode(6 x=#'x',7 alt.X('x',8 scale=alt.Scale(zero=False),9 axis=alt.Axis(labels=False, ticks=False, domain=False)10 ),11 y=12 alt.Y('y',13 scale=alt.Scale(zero=False),14 axis=alt.Axis(labels=False, ticks=False, domain=False)15 ),16 href='url:N',17 color=alt.Color('keywords:N',18 legend=alt.Legend(columns=1, symbolLimit=0, labelFontSize=14)19 ),20 opacity=alt.condition(selection, alt.value(1), alt.value(0.2)),21 tooltip=['title', 'keywords', 'cluster', 'score', 'descendants']22).properties(23 width=800,24 height=50025).add_selection(26 selection27).configure_legend(labelLimit= 0).configure_view(28 strokeWidth=029).configure(background=\"#FDF7F0\").properties(30 title='Ask HN: Top 3,000 Posts'31)32chart.interactive()\n```\n\nExample:\n```text\nThe common theme of the following words: books, book, read, the, you, are, what, best, in, youris that they all relate to favorite books to read.---The common theme of the following words: startup, company, yc, failedis that they all relate to startup companies and their failures.---The common theme of the following words: freelancer, wants, hired, be, who, seeking, to, 2014, 2020, aprilis that they all relate to hiring for a freelancer to join the team of a startup.---The common theme of the following words: <insert keywords here>is that they all relate to\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.347Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":22,"totalLines":113,"estimatedTokens":3713}}185{"id":"doc-analysis_of_form_10_k_10_q_using_cohere_and_rag_-746ecc46","source":"documentation","title":"Analysis of Form 10-K/10-Q Using Cohere and RAG | Cohere","url":"https://docs.cohere.com/page/analysis-of-financial-forms","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1%%capture2!sudo apt install tesseract-ocr poppler-utils3!pip install \"cohere<5\" langchain llama-index llama-index-embeddings-cohere llama-index-postprocessor-cohere-rerank pytesseract pdf2image\n```\n\nExample:\n```text\n1# Due to compatibility issues, we need to do imports like this2from llama_index.core.schema import TextNode34%%capture5!pip install unstructured\n```\n\nExample:\n```text\n1import cohere2from getpass import getpass34# Set up Cohere client5COHERE_API_KEY = getpass(\"Enter your Cohere API key: \")67# Instantiate a client to communicate with Cohere's API using our Python SDK8co = cohere.Client(COHERE_API_KEY)\n```\n\nExample:\n```text\nEnter your Cohere API key: ··········\n```\n\nExample:\n```text\n1# Using langchain here since they have access to the Unstructured Data Loader powered by unstructured.io2from langchain_community.document_loaders import UnstructuredURLLoader34# Load up Airbnb's 10-K from this past fiscal year (filed in 2024)5# Feel free to fill in some other EDGAR path6url = \"https://www.sec.gov/Archives/edgar/data/1559720/000155972024000006/abnb-20231231.htm\"7loader = UnstructuredURLLoader(urls=[url], headers={\"User-Agent\": \"cohere cohere@cohere.com\"})8documents = loader.load()910edgar_10k = documents[0].page_content1112# Load the document(s) as simple text nodes, to be passed to the tokenization processor13nodes = [TextNode(text=document.page_content, id_=f\"doc_{i}\") for i, document in enumerate(documents)]\n```\n\nExample:\n```text\n[nltk_data] Downloading package averaged_perceptron_tagger to[nltk_data] /root/nltk_data...[nltk_data] Unzipping taggers/averaged_perceptron_tagger.zip.\n```\n\nExample:\n```text\n1from llama_index.core.ingestion import IngestionPipeline2from llama_index.core.node_parser import SentenceSplitter34from transformers import AutoTokenizer56model_id = \"CohereForAI/c4ai-command-r-v01\"7tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)89# TODO: replace with a HF implementation so this is much faster. We'll10# presumably release it when we OS the model11tokenizer_fn = lambda x: tokenizer(x).input_ids if len(x) > 0 else []1213pipeline = IngestionPipeline(14 transformations=[15 SentenceSplitter(chunk_size=512, chunk_overlap=0, tokenizer=tokenizer_fn)16 ]17)1819# Run the pipeline to transform the text20nodes = pipeline.run(nodes=nodes)\n```\n\nExample:\n```text\n/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_token.py:88: UserWarning:The secret `HF_TOKEN` does not exist in your Colab secrets.To authenticate with the Hugging Face Hub, create a token in your settings tab (https://huggingface.co/settings/tokens), set it as secret in your Google Colab and restart your session.You will be able to reuse this secret in all of your notebooks.Please note that authentication is recommended but still optional to access public models or datasets. warnings.warn(tokenizer_config.json: 0%| | 0.00/7.92k [00:00<?, ?B/s]tokenization_cohere_fast.py: 0%| | 0.00/43.7k [00:00<?, ?B/s]configuration_cohere.py: 0%| | 0.00/7.37k [00:00<?, ?B/s]A new version of the following files was downloaded from https://huggingface.co/CohereForAI/c4ai-command-r-v01:- configuration_cohere.py. Make sure to double-check they do not contain any added malicious code. To avoid downloading new versions of the code file, you can pin a revision.A new version of the following files was downloaded from https://huggingface.co/CohereForAI/c4ai-command-r-v01:- tokenization_cohere_fast.py- configuration_cohere.py. Make sure to double-check they do not contain any added malicious code. To avoid downloading new versions of the code file, you can pin a revision.tokenizer.json: 0%| | 0.00/12.8M [00:00<?, ?B/s]special_tokens_map.json: 0%| | 0.00/429 [00:00<?, ?B/s]Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.\n```\n\nExample:\n```text\n1from llama_index.core import Settings, VectorStoreIndex23from llama_index.postprocessor.cohere_rerank import CohereRerank45from llama_index.embeddings.cohere import CohereEmbedding67# Instantiate the embedding model8embed_model = CohereEmbedding(cohere_api_key=COHERE_API_KEY)910# Global settings11Settings.chunk_size = 51212Settings.embed_model = embed_model1314# Create the vector store15index = VectorStoreIndex(nodes)1617retriever = index.as_retriever(similarity_top_k=30) # Change to whatever top_k you want1819# Instantiate the reranker20rerank = CohereRerank(api_key=COHERE_API_KEY, top_n=15)2122# Function `retrieve` is ready, using both Cohere embeddings for similarity search as well as23retrieve = lambda query: rerank.postprocess_nodes(retriever.retrieve(query), query_str=query)\n```\n\nExample:\n```text\n1PROMPT = \"List the overall revenue numbers for 2021, 2022, and 2023 in the 10-K as bullet points, then explain the revenue growth trends.\"23# Get queries to run against our index from the model4r = co.chat(PROMPT, model=\"command-r\", search_queries_only=True)5if r.search_queries:6 queries = [q[\"text\"] for q in r.search_queries]7else:8 print(\"No queries returned by the model\")\n```\n\nExample:\n```text\n1# Convenience function for formatting documents2def format_for_cohere_client(nodes_):3 return [4 {5 \"text\": node.node.text,6 \"llamaindex_id\": node.node.id_,7 }8 for node9 in nodes_10 ]111213documents = []14# Retrieve a set of chunks from the vector index and append them to the list of15# documents that should be included in the final RAG step16for query in queries:17 ret_nodes = retrieve(query)18 documents.extend(format_for_cohere_client(ret_nodes))1920# One final dedpulication step in case multiple queries return the same chunk21documents = [dict(t, id=f\"doc_{i}\") for i, t in enumerate({tuple(d.items()) for d in documents})]\n```\n\nExample:\n```text\n1# Make a request to the model2response = co.chat(3 message=PROMPT,4 model=\"command-r\",5 temperature=0.3,6 documents=documents,7 prompt_truncation=\"AUTO\"8)910print(response.text)\n```\n\nExample:\n```text\nHere are the overall revenue numbers for the years 2021, 2022, and 2023 as bullet points:- 2021: $5,992 million- 2022: $8,399 million- 2023: $9,917 millionRevenue increased by 18% in 2023 compared to 2022, primarily due to a 14% increase in Nights and Experiences Booked, which reached 54.5 million. This, combined with higher average daily rates, resulted in a 16% increase in Gross Booking Value, which reached $10.0 billion.The revenue growth trend demonstrates sustained strong travel demand. On a constant-currency basis, revenue increased by 17% in 2023 compared to the previous year.Other factors influencing the company's financial performance are described outside of the revenue growth trends.\n```\n\nExample:\n```text\n1# Helper function for displaying response WITH citations2def insert_citations(text: str, citations: list[dict]):3 \"\"\"4 A helper function to pretty print citations.5 \"\"\"6 offset = 07 # Process citations in the order they were provided8 for citation in citations:9 # Adjust start/end with offset10 start, end = citation['start'] + offset, citation['end'] + offset11 cited_docs = [doc[4:] for doc in citation[\"document_ids\"]]12 # Shorten citations if they're too long for convenience13 if len(cited_docs) > 3:14 placeholder = \"[\" + \", \".join(cited_docs[:3]) + \"...]\"15 else:16 placeholder = \"[\" + \", \".join(cited_docs) + \"]\"17 # ^ doc[4:] removes the 'doc_' prefix, and leaves the quoted document18 modification = f'{text[start:end]} {placeholder}'19 # Replace the cited text with its bolded version + placeholder20 text = text[:start] + modification + text[end:]21 # Update the offset for subsequent replacements22 offset += len(modification) - (end - start)2324 return text2526print(insert_citations(response.text, response.citations))\n```\n\nExample:\n```text\nHere are the overall revenue numbers for the years 2021, 2022, and 2023 as bullet points:- 2021: $5,992 million [13]- 2022: $8,399 million [13]- 2023: $9,917 million [13]Revenue increased by 18% in 2023 [11] compared to 2022, primarily due to a 14% increase in Nights and Experiences Booked [11], which reached 54.5 million. [11] This, combined with higher average daily rates [11], resulted in a 16% increase in Gross Booking Value [11], which reached $10.0 billion. [11]The revenue growth trend demonstrates sustained strong travel demand. [11] On a constant-currency basis [11], revenue increased by 17% in 2023 [11] compared to the previous year.Other factors [8, 14] influencing the company's financial performance are described outside of the revenue growth trends. [8, 14]\n```\n\nExample:\n```text\n1import pytesseract2from pdf2image import convert_from_path34# pdf2image extracts as a list of PIL.Image objects5# TODO: host this PDF somewhere6pages = convert_from_path(\"/content/uber_10k.pdf\")78# We access the only page in this sample PDF by indexing at 09pages = [pytesseract.image_to_string(page) for page in pages]\n```\n\nExample:\n```text\n1def get_response(prompt, rag):2 if rag:3 # Get queries to run against our index from the model4 r = co.chat(prompt, model=\"command-r\", search_queries_only=True)5 if r.search_queries:6 queries = [q[\"text\"] for q in r.search_queries]7 else:8 print(\"No queries returned by the model\")910 documents = []11 # Retrieve a set of chunks from the vector index and append them to the list of12 # documents that should be included in the final RAG step13 for query in queries:14 ret_nodes = retrieve(query)15 documents.extend(format_for_cohere_client(ret_nodes))1617 # One final dedpulication step in case multiple queries return the same chunk18 documents = [dict(t) for t in {tuple(d.items()) for d in documents}]1920 # Make a request to the model21 response = co.chat(22 message=prompt,23 model=\"command-r\",24 temperature=0.3,25 documents=documents,26 prompt_truncation=\"AUTO\"27 )28 else:29 response = co.chat(30 message=prompt,31 model=\"command-r\",32 temperature=0.3,33 )3435 return response\n```\n\nExample:\n```text\n1prompt_template = \"\"\"# financial form 10-K2{tenk}34# question5{question}\"\"\"67full_context_prompt = prompt_template.format(tenk=edgar_10k, question=PROMPT)\n```\n\nExample:\n```text\n1r1 = get_response(PROMPT, rag=True)2r2 = get_response(full_context_prompt, rag=False)\n```\n\nExample:\n```text\n1def get_price(r):2 return (r.token_count[\"prompt_tokens\"] * 0.5 / 10e6) + (r.token_count[\"response_tokens\"] * 1.5 / 10e6)\n```\n\nExample:\n```text\n1rag_price = get_price(r1)2full_context_price = get_price(r2)34print(f\"RAG is {(full_context_price - rag_price) / full_context_price:.0%} cheaper than full context\")\n```\n\nExample:\n```text\nRAG is 93% cheaper than full context\n```\n\nExample:\n```text\n1%timeit get_response(PROMPT, rag=True)\n```\n\nExample:\n```text\n14.9 s ± 1.4 s per loop (mean ± std. dev. of 7 runs, 1 loop each)\n```\n\nExample:\n```text\n1%timeit get_response(full_context_prompt, rag=False)\n```\n\nExample:\n```text\n22.7 s ± 7.43 s per loop (mean ± std. dev. of 7 runs, 1 loop each)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.349Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":26,"totalLines":133,"estimatedTokens":2898}}186{"id":"doc-getting_started_with_basic_tool_use_cohere-175388b0","source":"documentation","title":"Getting Started with Basic Tool Use | Cohere","url":"https://docs.cohere.com/page/basic-tool-use","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere, json2API_KEY = \"...\" # fill in your Cohere API key here3co = cohere.Client(API_KEY)\n```\n\nExample:\n```text\n1sales_database = {2 '2023-09-28': {3 'total_sales_amount': 5000,4 'total_units_sold': 100,5 },6 '2023-09-29': {7 'total_sales_amount': 10000,8 'total_units_sold': 250,9 },10 '2023-09-30': {11 'total_sales_amount': 8000,12 'total_units_sold': 200,13 }14}1516product_catalog = {17 'Electronics': [18 {'product_id': 'E1001', 'name': 'Smartphone', 'price': 500, 'stock_level': 20},19 {'product_id': 'E1002', 'name': 'Laptop', 'price': 1000, 'stock_level': 15},20 {'product_id': 'E1003', 'name': 'Tablet', 'price': 300, 'stock_level': 25},21 ],22 'Clothing': [23 {'product_id': 'C1001', 'name': 'T-Shirt', 'price': 20, 'stock_level': 100},24 {'product_id': 'C1002', 'name': 'Jeans', 'price': 50, 'stock_level': 80},25 {'product_id': 'C1003', 'name': 'Jacket', 'price': 100, 'stock_level': 40},26 ]27}\n```\n\nExample:\n```text\n1def query_daily_sales_report(day: str) -> dict:2 \"\"\"3 Function to retrieve the sales report for the given day4 \"\"\"5 report = sales_database.get(day, {})6 if report:7 return {8 'date': day,9 'summary': f\"Total Sales Amount: {report['total_sales_amount']}, Total Units Sold: {report['total_units_sold']}\"10 }11 else:12 return {'date': day, 'summary': 'No sales data available for this day.'}131415def query_product_catalog(category: str) -> dict:16 \"\"\"17 Function to retrieve products for the given category18 \"\"\"19 products = product_catalog.get(category, [])20 return {21 'category': category,22 'products': products23 }242526functions_map = {27 \"query_daily_sales_report\": query_daily_sales_report,28 \"query_product_catalog\": query_product_catalog29}\n```\n\nExample:\n```text\n1tools = [2 {3 \"name\": \"query_daily_sales_report\",4 \"description\": \"Connects to a database to retrieve overall sales volumes and sales information for a given day.\",5 \"parameter_definitions\": {6 \"day\": {7 \"description\": \"Retrieves sales data for this day, formatted as YYYY-MM-DD.\",8 \"type\": \"str\",9 \"required\": True10 }11 }12 },13 {14 \"name\": \"query_product_catalog\",15 \"description\": \"Connects to a a product catalog with information about all the products being sold, including categories, prices, and stock levels.\",16 \"parameter_definitions\": {17 \"category\": {18 \"description\": \"Retrieves product information data for all products in this category.\",19 \"type\": \"str\",20 \"required\": True21 }22 }23 }24]\n```\n\nExample:\n```text\n1preamble = \"\"\"2## Task & Context3You help people answer their questions and other requests interactively. You will be asked a very wide array of requests on all kinds of topics. You will be equipped with a wide range of search engines or similar tools to help you, which you use to research your answer. You should focus on serving the user's needs as best you can, which will be wide-ranging.45## Style Guide6Unless the user asks for a different style of answer, you should answer in full sentences, using proper grammar and spelling.7\"\"\"89message = \"Can you provide a sales summary for 29th September 2023, and also give me some details about the products in the 'Electronics' category, for example their prices and stock levels?\"\n```\n\nExample:\n```text\n1response = co.chat(2 message=message,3 tools=tools,4 preamble=preamble,5 model=\"command-a-03-2025\",6)789print(\"The model recommends doing the following tool calls:\")10print(\"\\n\".join(str(tool_call) for tool_call in response.tool_calls))\n```\n\nExample:\n```text\nThe model recommends doing the following tool calls:name='query_daily_sales_report' parameters={'day': '2023-09-29'}name='query_product_catalog' parameters={'category': 'Electronics'}\n```\n\nExample:\n```text\n1tool_results = []2for tool_call in response.tool_calls:3 # here is where you would call the tool recommended by the model, using the parameters recommended by the model4 print(f\"= running tool {tool_call.name}, with parameters: {tool_call.parameters}\")5 output = functions_map[tool_call.name](**tool_call.parameters)6 # store the output in a list7 outputs = [output]8 print(f\"== tool results: {outputs}\")9 # store your tool results in this format10 tool_results.append({11 \"call\": tool_call,12 \"outputs\": outputs13 })1415print(\"Tool results that will be fed back to the model in step 4:\")1617serializable_results = [{\"call\": {\"name\": r[\"call\"].name, \"parameters\": r[\"call\"].parameters}, \"outputs\": r[\"outputs\"]} for r in tool_results]18print(json.dumps(serializable_results, indent=4))\n```\n\nExample:\n```text\n= running tool query_daily_sales_report, with parameters: {'day': '2023-09-29'}== tool results: [{'date': '2023-09-29', 'summary': 'Total Sales Amount: 10000, Total Units Sold: 250'}]= running tool query_product_catalog, with parameters: {'category': 'Electronics'}== tool results: [{'category': 'Electronics', 'products': [{'product_id': 'E1001', 'name': 'Smartphone', 'price': 500, 'stock_level': 20}, {'product_id': 'E1002', 'name': 'Laptop', 'price': 1000, 'stock_level': 15}, {'product_id': 'E1003', 'name': 'Tablet', 'price': 300, 'stock_level': 25}]}]Tool results that will be fed back to the model in step 4:[ { \"call\": { \"name\": \"query_daily_sales_report\", \"parameters\": { \"day\": \"2023-09-29\" } }, \"outputs\": [ { \"date\": \"2023-09-29\", \"summary\": \"Total Sales Amount: 10000, Total Units Sold: 250\" } ] }, { \"call\": { \"name\": \"query_product_catalog\", \"parameters\": { \"category\": \"Electronics\" } }, \"outputs\": [ { \"category\": \"Electronics\", \"products\": [ { \"product_id\": \"E1001\", \"name\": \"Smartphone\", \"price\": 500, \"stock_level\": 20 }, { \"product_id\": \"E1002\", \"name\": \"Laptop\", \"price\": 1000, \"stock_level\": 15 }, { \"product_id\": \"E1003\", \"name\": \"Tablet\", \"price\": 300, \"stock_level\": 25 } ] } ] }]\n```\n\nExample:\n```text\n1response = co.chat(2 message=message,3 tools=tools,4 tool_results=tool_results,5 preamble=preamble,6 model=\"command-a-03-2025\",7 temperature=0.3,8 force_single_step=True9)\n```\n\nExample:\n```text\n1print(\"Final answer:\")2print(response.text)\n```\n\nExample:\n```text\nFinal answer:On the 29th September 2023, the total sales amount was 10,000 and the total units sold was 250.Here are the details of the products in the 'Electronics' category:| Product Name | Price | Product ID | Stock Level ||---|---|---|---|| Smartphone | 500 | E1001 | 20 || Laptop | 1000 | E1002 | 15 || Tablet | 300 | E1003 | 25 |\n```\n\nExample:\n```text\n1print(\"Citations that support the final answer:\")2for cite in response.citations:3 print(cite)\n```\n\nExample:\n```text\nCitations that support the final answer:{'start': 7, 'end': 26, 'text': '29th September 2023', 'document_ids': ['query_daily_sales_report:0:0']}{'start': 32, 'end': 61, 'text': 'total sales amount was 10,000', 'document_ids': ['query_daily_sales_report:0:0']}{'start': 70, 'end': 95, 'text': 'total units sold was 250.', 'document_ids': ['query_daily_sales_report:0:0']}{'start': 168, 'end': 180, 'text': 'Product Name', 'document_ids': ['query_product_catalog:1:0']}{'start': 183, 'end': 188, 'text': 'Price', 'document_ids': ['query_product_catalog:1:0']}{'start': 191, 'end': 201, 'text': 'Product ID', 'document_ids': ['query_product_catalog:1:0']}{'start': 204, 'end': 215, 'text': 'Stock Level', 'document_ids': ['query_product_catalog:1:0']}{'start': 238, 'end': 248, 'text': 'Smartphone', 'document_ids': ['query_product_catalog:1:0']}{'start': 251, 'end': 254, 'text': '500', 'document_ids': ['query_product_catalog:1:0']}{'start': 257, 'end': 262, 'text': 'E1001', 'document_ids': ['query_product_catalog:1:0']}{'start': 265, 'end': 267, 'text': '20', 'document_ids': ['query_product_catalog:1:0']}{'start': 272, 'end': 278, 'text': 'Laptop', 'document_ids': ['query_product_catalog:1:0']}{'start': 281, 'end': 285, 'text': '1000', 'document_ids': ['query_product_catalog:1:0']}{'start': 288, 'end': 293, 'text': 'E1002', 'document_ids': ['query_product_catalog:1:0']}{'start': 296, 'end': 298, 'text': '15', 'document_ids': ['query_product_catalog:1:0']}{'start': 303, 'end': 309, 'text': 'Tablet', 'document_ids': ['query_product_catalog:1:0']}{'start': 312, 'end': 315, 'text': '300', 'document_ids': ['query_product_catalog:1:0']}{'start': 318, 'end': 323, 'text': 'E1003', 'document_ids': ['query_product_catalog:1:0']}{'start': 326, 'end': 328, 'text': '25', 'document_ids': ['query_product_catalog:1:0']}\n```\n\nExample:\n```text\n1def insert_citations_in_order(text, citations):2 \"\"\"3 A helper function to pretty print citations.4 \"\"\"5 offset = 06 document_id_to_number = {}7 citation_number = 08 modified_citations = []910 # Process citations, assigning numbers based on unique document_ids11 for citation in citations:12 citation_numbers = []13 for document_id in sorted(citation[\"document_ids\"]):14 if document_id not in document_id_to_number:15 citation_number += 1 # Increment for a new document_id16 document_id_to_number[document_id] = citation_number17 citation_numbers.append(document_id_to_number[document_id])1819 # Adjust start/end with offset20 start, end = citation['start'] + offset, citation['end'] + offset21 placeholder = ''.join([f'[{number}]' for number in citation_numbers])22 # Bold the cited text and append the placeholder23 modification = f'**{text[start:end]}**{placeholder}'24 # Replace the cited text with its bolded version + placeholder25 text = text[:start] + modification + text[end:]26 # Update the offset for subsequent replacements27 offset += len(modification) - (end - start)2829 # Prepare citations for listing at the bottom, ensuring unique document_ids are listed once30 unique_citations = {number: doc_id for doc_id, number in document_id_to_number.items()}31 citation_list = '\\n'.join([f'[{doc_id}] source: {tool_results[doc_id - 1][\"outputs\"]} \\n based on tool call: {dict(tool_results[doc_id - 1][\"call\"])}' for doc_id, number in sorted(unique_citations.items(), key=lambda item: item[1])])32 text_with_citations = f'{text}\\n\\n{citation_list}'3334 return text_with_citations353637print(insert_citations_in_order(response.text, response.citations))\n```\n\nExample:\n```text\nOn the **29th September 2023**[1], the **total sales amount was 10,000**[1] and the **total units sold was 250.**[1]Here are the details of the products in the 'Electronics' category:| **Product Name**[2] | **Price**[2] | **Product ID**[2] | **Stock Level**[2] ||---|---|---|---|| **Smartphone**[2] | **500**[2] | **E1001**[2] | **20**[2] || **Laptop**[2] | **1000**[2] | **E1002**[2] | **15**[2] || **Tablet**[2] | **300**[2] | **E1003**[2] | **25**[2] |[1] source: [{'date': '2023-09-29', 'summary': 'Total Sales Amount: 10000, Total Units Sold: 250'}] based on tool call: {'name': 'query_daily_sales_report', 'parameters': {'day': '2023-09-29'}}[2] source: [{'category': 'Electronics', 'products': [{'product_id': 'E1001', 'name': 'Smartphone', 'price': 500, 'stock_level': 20}, {'product_id': 'E1002', 'name': 'Laptop', 'price': 1000, 'stock_level': 15}, {'product_id': 'E1003', 'name': 'Tablet', 'price': 300, 'stock_level': 25}]}] based on tool call: {'name': 'query_product_catalog', 'parameters': {'category': 'Electronics'}}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.351Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":16,"totalLines":83,"estimatedTokens":3140}}187{"id":"doc-agentic_multi_stage_rag_with_cohere_tools_api_co-a5419859","source":"documentation","title":"Agentic Multi-Stage RAG with Cohere Tools API | Cohere","url":"https://docs.cohere.com/page/agentic-multi-stage-rag","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import os2from pprint import pprint34import cohere5import pandas as pd6from sklearn.metrics.pairwise import cosine_similarity\n```\n\nExample:\n```text\n1# versions2print('cohere version:', cohere.__version__)\n```\n\nExample:\n```text\ncohere version: 5.5.1\n```\n\nExample:\n```text\n1COHERE_API_KEY = os.environ.get(\"CO_API_KEY\")2COHERE_MODEL = 'command-a-03-2025'3co = cohere.Client(api_key=COHERE_API_KEY)\n```\n\nExample:\n```text\n1documents = [2 {3 \"title\": \"Bicycle law\",4 \"body\": \"\"\"5 Traffic Infractions and fees - For all information related to bicycle traffic infractions such as not wearing a helmet and fee information, please visit Section 3b for more information.6 Riding on the road - When riding on a roadway, a cyclist has all the rights and responsibilities of a vehicle driver (RCW 46.61.755). Bicyclists who violate traffic laws may be ticketed (RCW 46.61.750).7 Roads closed to bicyclists - Some designated sections of the state's limited access highway system may be closed to bicyclists. See the permanent bike restrictions map for more information. In addition, local governments may adopt ordinances banning cycling on specific roads or on sidewalks within business districts.8 Children bicycling - Parents or guardians may not knowingly permit bicycle traffic violations by their ward (RCW 46.61.700).9 Riding side by side - Bicyclists may ride side by side, but not more than two abreast (RCW 46.61.770).10 Riding at night - For night bicycle riding, a white front light (not a reflector) visible for 500 feet and a red rear reflector are required. A red rear light may be used in addition to the required reflector (RCW 46.61.780).11 Shoulder vs. bike lane - Bicyclists may choose to ride on the path, bike lane, shoulder or travel lane as suits their safety needs (RCW 46.61.770).12 Bicycle helmets - Currently, there is no state law requiring helmet use. However, some cities and counties do require helmets. For specific information along with location for bicycle helmet law please reference to section 21a.13 Bicycle equipment - Bicycles must be equipped with a white front light visible for 500 feet and a red rear reflector (RCW 46.61.780). A red rear light may be used in addition to the required reflector.14\"\"\",15 },16 {17 \"title\": \"Bicycle helmet requirement\",18 \"body\": \"Currently, there is no state law requiring helmet use. However, some cities and counties do require helmet use with bicycles. Here is a list of those locations and when the laws were enacted. For specific information along with location for bicycle helmet law please reference to section 21a.\",19 },20 {21 \"title\": \"Section 21a\",22 \"body\": \"\"\"helmet rules by location: These are city and county level rules. The following group must wear helmets.23 Location name | Who is affected | Effective date24 Aberdeen | All ages | 200125 Bainbridge Island | All ages | 200126 Bellevue | All ages | 200127 Bremerton | All ages | 200028 DuPont | All ages | 200829 Eatonville | All ages | 199630 Fircrest | All ages | 199531 Gig Harbor | All ages | 199632 Kent | All ages | 199933 Lynnwood | All ages | 200434 Lakewood | All ages | 199635 Milton | All ages | 199736 Orting | Under 17 | 19973738 For fines and rules, you will be charged in according with Section 3b of the law.39 \"\"\",40 },41 {42 \"title\": \"Section 3b\",43 \"body\": \"\"\"Traffic infraction - A person operating a bicycle upon a roadway or highway shall be subject to the provisions of this chapter relating to traffic infractions.44 1. Stop for people in crosswalks. Every intersection is a crosswalk - It’s the law. Drivers must stop for pedestrians at intersections, whether it’s an unmarked or marked crosswalk, and bicyclists in crosswalks are considered pedestrians. Also, it is illegal to pass another vehicle stopped for someone at a crosswalk. In Washington, the leading action motorists take that results in them hitting someone is a failure to yield to pedestrians.45 2. Put the phone down. Hand-held cell phone use and texting is prohibited for all Washington drivers and may result in a $136 fine for first offense, $235 on the second distracted-driving citation.46 3. Helmets are required for all bicyclists according to the state and municipal laws. If you are in a group required to wear a helmet but do not wear it you can be fined $48. # If you are the parent or legal guardian of a child under 17 and knowingly allow them to ride without a helmet, you can be fined $136.47\"\"\",48 },49]50db = pd.DataFrame(documents)51# comebine title and body52db[\"combined\"] = \"Title: \" + db[\"title\"] + \"\\n\" + \"Body: \" + db[\"body\"]53# generate embedding54embeddings = co.embed(55 texts=db.combined.tolist(), model=\"embed-v4.0\", input_type=\"search_document\"56)57db[\"embeddings\"] = embeddings.embeddings\n```\n\nExample:\n```text\n1db\n```\n\nExample:\n```text\n1def retrieve_documents(query: str, n=1) -> dict:2 \"\"\"3 Function to retrieve documents a given query.45 Steps:6 1. Embed the query7 2. Calculate cosine similarity between the query embedding and the embeddings of the documents8 3. Return the top n documents with the highest similarity scores9 \"\"\"10 query_emb = co.embed(11 texts=[query], model=\"embed-v4.0\", input_type=\"search_query\"12 )1314 similarity_scores = cosine_similarity(15 [query_emb.embeddings[0]], db.embeddings.tolist()16 )17 similarity_scores = similarity_scores[0]1819 top_indices = similarity_scores.argsort()[::-1][:n]20 top_matches = db.iloc[top_indices]2122 return {\"top_matched_document\": top_matches.combined}232425functions_map = {26 \"retrieve_documents\": retrieve_documents,27}2829tools = [30 {31 \"name\": \"retrieve_documents\",32 \"description\": \"given a query, retrieve documents from a database to answer user's question\",33 \"parameter_definitions\": {34 \"query\": {\"description\": \"query\", \"type\": \"str\", \"required\": True}35 },36 }37]\n```\n\nExample:\n```text\n1def simple_rag(query, db):2 \"\"\"3 Given user's query, retrieve top documents and generate response using documents parameter.4 \"\"\"5 top_matched_document = retrieve_documents(query)[\"top_matched_document\"]67 print(\"top_matched_document\", top_matched_document)89 output = co.chat(10 message=query, model=COHERE_MODEL, documents=[top_matched_document]11 )1213 return output.text\n```\n\nExample:\n```text\n1def cohere_agent(2 message: str,3 preamble: str,4 tools: list[dict],5 force_single_step=False,6 verbose: bool = False,7 temperature: float = 0.3,8) -> str:9 \"\"\"10 Function to handle multi-step tool use api.1112 Args:13 message (str): The message to send to the Cohere AI model.14 preamble (str): The preamble or context for the conversation.15 tools (list of dict): List of tools to use in the conversation.16 verbose (bool, optional): Whether to print verbose output. Defaults to False.1718 Returns:19 str: The final response from the call.20 \"\"\"2122 counter = 12324 response = co.chat(25 model=COHERE_MODEL,26 message=message,27 preamble=preamble,28 tools=tools,29 force_single_step=force_single_step,30 temperature=temperature,31 )3233 if verbose:34 print(f\"\\nrunning 0th step.\")35 print(response.text)3637 while response.tool_calls:38 tool_results = []3940 if verbose:41 print(f\"\\nrunning {counter}th step.\")4243 for tool_call in response.tool_calls:44 output = functions_map[tool_call.name](**tool_call.parameters)45 outputs = [output]46 tool_results.append({\"call\": tool_call, \"outputs\": outputs})4748 if verbose:49 print(50 f\"= running tool {tool_call.name}, with parameters: \\n{tool_call.parameters}\"51 )52 print(f\"== tool results:\")53 pprint(output)5455 response = co.chat(56 model=COHERE_MODEL,57 message=\"\",58 chat_history=response.chat_history,59 preamble=preamble,60 tools=tools,61 force_single_step=force_single_step,62 tool_results=tool_results,63 temperature=temperature,64 )6566 if verbose:67 print(response.text)68 counter += 16970 return response.text\n```\n\nExample:\n```text\n1question1 = \"Is there a state level law for wearing helmets?\"\n```\n\nExample:\n```text\n1output = simple_rag(question1, db)2print(output)\n```\n\nExample:\n```text\ntop_matched_document 1 Title: Bicycle helmet requirement\\nBody: Curre...Name: combined, dtype: objectThere is currently no state law requiring the use of helmets when riding a bicycle. However, some cities and counties do require helmet use.\n```\n\nExample:\n```text\n1preamble = \"\"\"2You are an expert assistant that helps users answers question about legal documents and policies.3Use the provided documents to answer questions about an employee's specific situation.4\"\"\"56output = cohere_agent(question1, preamble, tools, verbose=True)\n```\n\nExample:\n```text\nrunning 0th step.I will search for 'state level law for wearing helmets' in the documents provided and write an answer based on what I find.running 1th step.= running tool retrieve_documents, with parameters:{'query': 'state level law for wearing helmets'}== tool results:{'top_matched_document': 1 Title: Bicycle helmet requirement\\nBody: Curre...Name: combined, dtype: object}There is currently no state law requiring helmet use. However, some cities and counties do require helmet use with bicycles.\n```\n\nExample:\n```text\n1question2 = \"I live in orting, do I need to wear a helmet with a bike?\"\n```\n\nExample:\n```text\n1output = simple_rag(question2, db)2print(output)\n```\n\nExample:\n```text\ntop_matched_document 1 Title: Bicycle helmet requirement\\nBody: Curre...Name: combined, dtype: objectIn the state of Washington, there is no law requiring you to wear a helmet when riding a bike. However, some cities and counties do require helmet use, so it is worth checking your local laws.\n```\n\nExample:\n```text\n1preamble = \"\"\"2You are an expert assistant that helps users answers question about legal documents and policies.3Use the provided documents to answer questions about an employee's specific situation.4\"\"\"56output = cohere_agent(question2, preamble, tools, verbose=True)\n```\n\nExample:\n```text\nrunning 0th step.I will search for 'helmet with a bike' and then write an answer.running 1th step.= running tool retrieve_documents, with parameters:{'query': 'helmet with a bike'}== tool results:{'top_matched_document': 1 Title: Bicycle helmet requirement\\nBody: Curre...Name: combined, dtype: object}There is no state law requiring helmet use, however, some cities and counties do require helmet use with bicycles. I cannot find any information about Orting specifically, but you should check with your local authority.\n```\n\nExample:\n```text\n1def reference_extractor(query: str, documents: list[str]) -> str:2 \"\"\"3 Given a query and document, find references to other documents.4 \"\"\"5 prompt = f\"\"\"6 # instruction7 Does the reference document mention any other documents? If so, list them.8 If not, return empty string.910 # user query11 {query}1213 # retrieved documents14 {documents}15 \"\"\"1617 return co.chat(message=prompt, model=COHERE_MODEL, preamble=None).text181920def retrieve_documents(query: str, n=1) -> dict:21 \"\"\"22 Function to retrieve most relevant documents a given query.23 It also returns other references mentioned in the top matched documents.24 \"\"\"25 query_emb = co.embed(26 texts=[query], model=\"embed-v4.0\", input_type=\"search_query\"27 )2829 similarity_scores = cosine_similarity(30 [query_emb.embeddings[0]], db.embeddings.tolist()31 )32 similarity_scores = similarity_scores[0]3334 top_indices = similarity_scores.argsort()[::-1][:n]35 top_matches = db.iloc[top_indices]36 other_references = reference_extractor(query, top_matches.combined.tolist())3738 return {39 \"top_matched_document\": top_matches.combined,40 \"other_references_to_query\": other_references,41 }424344functions_map = {45 \"retrieve_documents\": retrieve_documents,46}4748tools = [49 {50 \"name\": \"retrieve_documents\",51 \"description\": \"given a query, retrieve documents from a database to answer user's question. It also finds references to other documents that should be leveraged to retrieve more documents\",52 \"parameter_definitions\": {53 \"query\": {54 \"description\": \"user's question or question or name of other document sections or references.\",55 \"type\": \"str\",56 \"required\": True,57 }58 },59 }60]\n```\n\nExample:\n```text\n1preamble2 = \"\"\"# Instruction2You are an expert assistant that helps users answer questions about legal documents and policies.34Please follow these steps:51. Using user's query, use `retrieve_documents` tool to retrieve the most relevant document from the database.62. If you see `other_references_to_query` in the tool result, search the mentioned referenced using `retrieve_documents(<other reference=\"\">)` tool to retrieve more documents.73. Keep trying until you find the answer.84. Answer with yes or no as much as you can to answer the question directly.9\"\"\"1011output = cohere_agent(question2, preamble2, tools, verbose=True)\n```\n\nExample:\n```text\nrunning 0th step.I will search for 'Orting' and 'bike helmet' to find the relevant information.running 1th step.= running tool retrieve_documents, with parameters:{'query': 'Orting bike helmet'}== tool results:{'other_references_to_query': 'Section 21a, Section 3b', 'top_matched_document': 0 Title: Bicycle law\\nBody: \\n Riding on ...Name: combined, dtype: object}I have found that there is no state law requiring helmet use, but some cities and counties do require helmets. I will now search for 'Section 21a' to find out if Orting is one of these cities or counties.running 2th step.= running tool retrieve_documents, with parameters:{'query': 'Section 21a'}== tool results:{'other_references_to_query': '- Section 3b', 'top_matched_document': 2 Title: Section 21a\\nBody: helmet rules by loca...Name: combined, dtype: object}Yes, you do need to wear a helmet when riding a bike in Orting if you are under 17.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.352Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":22,"totalLines":113,"estimatedTokens":3726}}188{"id":"doc-semantic_search_with_cohere_embed_jobs_cohere-73c20af6","source":"documentation","title":"Semantic Search with Cohere Embed Jobs | Cohere","url":"https://docs.cohere.com/page/embed-jobs","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import time2import cohere3import hnswlib4co = cohere.Client('COHERE_API_KEY')\n```\n\nExample:\n```text\n1dataset_file_path = \"data/embed_jobs_sample_data.jsonl\" # Full path - https://raw.githubusercontent.com/cohere-ai/cohere-developer-experience/main/notebooks/data/embed_jobs_sample_data.jsonl23ds=co.create_dataset(4\tname='sample_file',5\tdata=open(dataset_file_path, 'rb'),6\tkeep_fields = ['id','wiki_id'],7\tdataset_type=\"embed-input\"8\t)\n```\n\nExample:\n```text\nuploading file, starting validation...sample-file-hca4x0 was uploaded...\n```\n\nExample:\n```text\n1print(ds.await_validation())\n```\n\nExample:\n```text\ncohere.Dataset { id: sample-file-hca4x0 name: sample_file dataset_type: embed-input validation_status: validated created_at: 2024-01-13 02:51:48.215973 updated_at: 2024-01-13 02:51:48.215973 download_urls: [''] validation_error: None validation_warnings: []}\n```\n\nExample:\n```text\n1job = co.create_embed_job(2 dataset_id=ds.id,3 input_type='search_document' ,4 model='embed-english-v3.0',5 embeddings_types=['float'])67job.wait() # poll the server until the job is completed\n```\n\nExample:\n```text\n......\n```\n\nExample:\n```text\n1print(job)\n```\n\nExample:\n```text\ncohere.EmbedJob { job_id: 792bbc1a-561b-48c2-8a97-0c80c1914ea8 status: complete created_at: 2024-01-13T02:53:31.879719Z input_dataset_id: sample-file-hca4x0 output_urls: None model: embed-english-v3.0 truncate: RIGHT percent_complete: 100 output: cohere.Dataset { id: embeded-sample-file-drtjf9 name: embeded-sample-file dataset_type: embed-result validation_status: validated created_at: 2024-01-13 02:53:33.569362 updated_at: 2024-01-13 02:53:33.569362 download_urls: [''] validation_error: None validation_warnings: []}}\n```\n\nExample:\n```text\n1embeddings_file_path = 'embed_jobs_output.csv'2output_dataset=co.get_dataset(job.output.id)3output_dataset.save(filepath=embeddings_file_path, format=\"csv\")\n```\n\nExample:\n```text\n1embeddings=[]2texts=[]3for record in output_dataset:4 embeddings.append(record['embeddings']['float'])5 texts.append(record['text'])\n```\n\nExample:\n```text\n1index = hnswlib.Index(space='ip', dim=1024)2index.init_index(max_elements=len(embeddings), ef_construction=512, M=64)3index.add_items(embeddings,list(range(len(embeddings))))\n```\n\nExample:\n```text\n1query = \"What was the first youtube video about?\"23query_emb=co.embed(4 texts=[query], model=\"embed-english-v3.0\", input_type=\"search_query\"5 ).embeddings67doc_index = index.knn_query(query_emb, k=10)[0][0]89docs_to_rerank = []10for index in doc_index:11 docs_to_rerank.append(texts[index])1213final_result = co.rerank(14 query= query,15 documents=docs_to_rerank,16 model=\"rerank-english-v2.0\",17 top_n=3)\n```\n\nExample:\n```text\n1for idx, r in enumerate(final_result):2 print(f\"Document Rank: {idx + 1}, Document Index: {r.index}\")3 print(f\"Document: {r.document['text']}\")4 print(f\"Relevance Score: {r.relevance_score:.5f}\")5 print(\"\\n\")\n```\n\nExample:\n```text\nDocument Rank: 1, Document Index: 0Document: YouTube began as a venture capital–funded technology startup. Between November 2005 and April 2006, the company raised money from various investors, with Sequoia Capital, $11.5 million, and Artis Capital Management, $8 million, being the largest two. YouTube's early headquarters were situated above a pizzeria and a Japanese restaurant in San Mateo, California. In February 2005, the company activated codice_1. The first video was uploaded April 23, 2005. Titled \"Me at the zoo\", it shows co-founder Jawed Karim at the San Diego Zoo and can still be viewed on the site. In May, the company launched a public beta and by November, a Nike ad featuring Ronaldinho became the first video to reach one million total views. The site launched officially on December 15, 2005, by which time the site was receiving 8 million views a day. Clips at the time were limited to 100 megabytes, as little as 30 seconds of footage.Relevance Score: 0.94815Document Rank: 2, Document Index: 1Document: Karim said the inspiration for YouTube first came from the Super Bowl XXXVIII halftime show controversy when Janet Jackson's breast was briefly exposed by Justin Timberlake during the halftime show. Karim could not easily find video clips of the incident and the 2004 Indian Ocean Tsunami online, which led to the idea of a video-sharing site. Hurley and Chen said that the original idea for YouTube was a video version of an online dating service, and had been influenced by the website Hot or Not. They created posts on Craigslist asking attractive women to upload videos of themselves to YouTube in exchange for a $100 reward. Difficulty in finding enough dating videos led to a change of plans, with the site's founders deciding to accept uploads of any video.Relevance Score: 0.91626Document Rank: 3, Document Index: 2Document: YouTube was not the first video-sharing site on the Internet; Vimeo was launched in November 2004, though that site remained a side project of its developers from CollegeHumor at the time and did not grow much, either. The week of YouTube's launch, NBC-Universal's \"Saturday Night Live\" ran a skit \"Lazy Sunday\" by The Lonely Island. Besides helping to bolster ratings and long-term viewership for \"Saturday Night Live\", \"Lazy Sunday\"'s status as an early viral video helped establish YouTube as an important website. Unofficial uploads of the skit to YouTube drew in more than five million collective views by February 2006 before they were removed when NBCUniversal requested it two months later based on copyright concerns. Despite eventually being taken down, these duplicate uploads of the skit helped popularize YouTube's reach and led to the upload of more third-party content. The site grew rapidly; in July 2006, the company announced that more than 65,000 new videos were being uploaded every day and that the site was receiving 100 million video views per day.Relevance Score: 0.90665\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.353Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":1548}}189{"id":"doc-fueling_generative_content_with_keyword_research-17e7a549","source":"documentation","title":"Fueling Generative Content with Keyword Research | Cohere","url":"https://docs.cohere.com/page/fueling-generative-content","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1! pip install cohere -q\n```\n\nExample:\n```text\n1import cohere2import numpy as np3import pandas as pd4from sklearn.cluster import KMeans56import cohere7co = cohere.Client(\"COHERE_API_KEY\") # Get your API key: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1#@title Enable text wrapping in Google Colab23from IPython.display import HTML, display45def set_css():6 display(HTML('''78 '''))9get_ipython().events.register('pre_run_cell', set_css)\n```\n\nExample:\n```text\n1import wget2wget.download(\"https://raw.githubusercontent.com/cohere-ai/cohere-developer-experience/main/notebooks/data/remote_teams.csv\", \"remote_teams.csv\")\n```\n\nExample:\n```text\n'remote_teams.csv'\n```\n\nExample:\n```text\n1df = pd.read_csv('remote_teams.csv')2df.columns = [\"keyword\",\"volume\"]3df.head()\n```\n\nExample:\n```text\n1def embed_text(texts):2 output = co.embed(3 texts=texts,4 model='embed-v4.0',5 input_type=\"search_document\",6 )7 return output.embeddings89embeds = np.array(embed_text(df['keyword'].tolist()))\n```\n\nExample:\n```text\n1NUM_TOPICS = 42kmeans = KMeans(n_clusters=NUM_TOPICS, random_state=21, n_init=\"auto\").fit(embeds)3df['topic'] = list(kmeans.labels_)4df.head()\n```\n\nExample:\n```text\n1topic_keywords_dict = {topic: list(set(group['keyword'])) for topic, group in df.groupby('topic')}\n```\n\nExample:\n```text\n1def generate_topic_name(keywords):2 # Construct the prompt3 prompt = f\"\"\"Generate a concise topic name that best represents these keywords.\\4Provide just the topic name and not any additional details.56Keywords: {', '.join(keywords)}\"\"\"78 # Call the Cohere API9 response = co.chat(10 model='command-a-03-2025', # Choose the model size11 message=prompt,12 preamble=\"\")1314 # Return the generated text15 return response.text\n```\n\nExample:\n```text\n1topic_name_mapping = {topic: generate_topic_name(keywords) for topic, keywords in topic_keywords_dict.items()}23df['topic_name'] = df['topic'].map(topic_name_mapping)45df.head()\n```\n\nExample:\n```text\n1for topic, name in topic_name_mapping.items():2 print(f\"Topic {topic}: {name}\")\n```\n\nExample:\n```text\nTopic 0: **Effective Leadership and Management of Remote Teams**Topic 1: **Essential Tools and Software for Remote Team Collaboration and Productivity**Topic 2: **Remote Team Engagement and Team Building Activities**Topic 3: **Remote Team Engagement Games**\n```\n\nExample:\n```text\n1TOP_N = 1023top_keywords = (df.groupby('topic')4 .apply(lambda x: x.nlargest(TOP_N, 'volume'))5 .reset_index(drop=True))678content_by_topic = {}9for topic, group in top_keywords.groupby('topic'):10 keywords = ', '.join(list(group['keyword']))11 topic2name = topic2name = dict(df.groupby('topic')['topic_name'].first())12 topic_name = topic2name[topic]13 content_by_topic[topic] = {'topic_name': topic_name, 'keywords': keywords}\n```\n\nExample:\n```text\n1content_by_topic\n```\n\nExample:\n```text\n{0: {'topic_name': '**Effective Leadership and Management of Remote Teams**', 'keywords': 'managing remote teams, remote teams, how to manage remote teams, leading remote teams, managing remote teams best practices, remote teams best practices, scrum remote teams, best practices for managing remote teams, manage remote teams, slack best practices for remote teams'}, 1: {'topic_name': '**Essential Tools and Software for Remote Team Collaboration and Productivity**', 'keywords': 'collaboration tools for remote teams, best collaboration tools for remote teams, tools for remote teams, zapier remote teams, best communication tools for remote teams, free collaboration tools for remote teams, free retrospective tools for remote teams, project management tools for remote teams, best tools for remote teams, collaboration tool for remote teams'}, 2: {'topic_name': '**Remote Team Engagement and Team Building Activities**', 'keywords': 'team building activities for remote teams, team building for remote teams, retrospective ideas for remote teams, team building ideas for remote teams, fun retrospective ideas for remote teams, retro ideas for remote teams, team building exercises for remote teams, trust building exercises for remote teams, activities for remote teams, communication exercises for remote teams'}, 3: {'topic_name': '**Remote Team Engagement Games**', 'keywords': 'online games for remote teams, games for remote teams, retrospective games for remote teams, virtual games for remote teams, agile games for remote teams, fun games for remote teams, icebreaker games for remote teams, best games for remote teams, best online games for remote teams, bingo for remote teams'}}\n```\n\nExample:\n```text\n1def generate_blog_ideas(keywords):2 prompt = f\"\"\"{keywords}\\n\\nThe above is a list of high-traffic keywords obtained from a keyword research tool.3Suggest three blog post ideas that are highly relevant to these keywords.4For each idea, write a one paragraph abstract about the topic.5Use this format:6Blog title: <text>7Abstract: <text>\"\"\"89 response = co.chat(10 model='command-a-03-2025',11 message = prompt)12 return response.text\n```\n\nExample:\n```text\n1for key,value in content_by_topic.items():2 value['ideas'] = generate_blog_ideas(value['keywords'])345for key,value in content_by_topic.items():6 print(f\"Topic Name: {value['topic_name']}\\n\")7 print(f\"Top Keywords: {value['keywords']}\\n\")8 print(f\"Blog Post Ideas: {value['ideas']}\\n\")9 print(\"-\"*50)\n```\n\nExample:\n```text\nTopic Name: **Effective Leadership and Management of Remote Teams**Top Keywords: managing remote teams, remote teams, how to manage remote teams, leading remote teams, managing remote teams best practices, remote teams best practices, scrum remote teams, best practices for managing remote teams, manage remote teams, slack best practices for remote teamsBlog Post Ideas: **Blog Title: Mastering Remote Team Management: Best Practices for Leaders** **Abstract:** Leading a remote team comes with unique challenges, from communication barriers to maintaining team cohesion. This blog post dives into the essential best practices for managing remote teams effectively, including setting clear expectations, leveraging the right tools, and fostering a culture of trust and accountability. Whether you're new to remote leadership or looking to refine your approach, these actionable strategies will help you build a productive and engaged remote workforce.**Blog Title: Scrum for Remote Teams: Adapting Agile Practices for Virtual Collaboration** **Abstract:** Agile methodologies like Scrum are powerful frameworks for project management, but how do they translate to remote teams? This post explores how to adapt Scrum practices for virtual environments, focusing on tools like Slack, Zoom, and Jira to facilitate daily stand-ups, sprint planning, and retrospectives. Learn how to overcome common challenges, such as time zone differences and reduced face-to-face interaction, while maintaining the agility and efficiency Scrum is known for.**Blog Title: Slack Best Practices for Remote Teams: Boosting Communication and Productivity** **Abstract:** Slack has become a cornerstone of remote team communication, but without the right strategies, it can lead to overwhelm and inefficiency. This blog post outlines the best practices for using Slack in remote teams, including channel organization, etiquette guidelines, and integration with other tools. Discover how to create a streamlined communication workflow that keeps your team aligned, reduces noise, and enhances productivity in a remote work setting.--------------------------------------------------Topic Name: **Essential Tools and Software for Remote Team Collaboration and Productivity**Top Keywords: collaboration tools for remote teams, best collaboration tools for remote teams, tools for remote teams, zapier remote teams, best communication tools for remote teams, free collaboration tools for remote teams, free retrospective tools for remote teams, project management tools for remote teams, best tools for remote teams, collaboration tool for remote teamsBlog Post Ideas: **Blog Title: The Ultimate Guide to the Best Collaboration Tools for Remote Teams in 2023** **Abstract:** As remote work continues to rise, finding the right collaboration tools is essential for maintaining productivity and team cohesion. This comprehensive guide explores the top collaboration tools for remote teams, including project management platforms, communication apps, and automation tools like Zapier. Whether you're a small startup or a large enterprise, this post will help you identify the best tools to streamline workflows, enhance communication, and ensure your team stays aligned, regardless of their location. **Blog Title: 10 Free Collaboration Tools Every Remote Team Should Be Using** **Abstract:** Budget constraints shouldn't hinder your remote team's ability to collaborate effectively. This blog post highlights 10 free collaboration tools that cater to various needs, from project management and communication to retrospectives and file sharing. Discover how tools like Trello, Slack, and Miro can empower your team without breaking the bank, and learn tips for maximizing their features to foster seamless remote collaboration. **Blog Title: How to Choose the Right Project Management Tool for Your Remote Team** **Abstract:** With countless project management tools available, selecting the perfect one for your remote team can be overwhelming. This post breaks down the key factors to consider, such as team size, project complexity, and integration capabilities, to help you make an informed decision. From Asana and Monday.com to ClickUp and Notion, explore the pros and cons of popular tools and find the one that best aligns with your team's unique needs and workflows.--------------------------------------------------Topic Name: **Remote Team Engagement and Team Building Activities**Top Keywords: team building activities for remote teams, team building for remote teams, retrospective ideas for remote teams, team building ideas for remote teams, fun retrospective ideas for remote teams, retro ideas for remote teams, team building exercises for remote teams, trust building exercises for remote teams, activities for remote teams, communication exercises for remote teamsBlog Post Ideas: **Blog Title: 10 Fun and Engaging Team Building Activities for Remote Teams** Abstract: Remote work can sometimes make team bonding challenging, but with the right activities, you can foster connection and collaboration. This blog post explores 10 creative and interactive team building ideas tailored for remote teams, from virtual escape rooms to online trivia nights. Each activity is designed to strengthen relationships, improve communication, and boost morale, ensuring your team stays connected and motivated, no matter the distance. **Blog Title: Retro Reimagined: 7 Fun Retrospective Ideas for Remote Teams** Abstract: Retrospectives are essential for remote teams to reflect, learn, and grow, but they don't have to be boring. This post introduces 7 fun and effective retrospective ideas that go beyond the standard meeting format. From \"Virtual Appreciation Boards\" to \"Future Visioning\" exercises, these ideas encourage open communication, celebrate successes, and identify areas for improvement in an engaging and interactive way. **Blog Title: Building Trust and Communication: 5 Essential Exercises for Remote Teams** Abstract: Trust and communication are the cornerstones of successful remote teams, but they require intentional effort to develop. This blog post highlights 5 powerful exercises designed to strengthen trust and enhance communication among remote team members. From \"Virtual Coffee Chats\" to \"Pair Programming\" and \"Empathy Mapping,\" these activities create safe spaces for vulnerability, foster understanding, and ensure everyone feels heard and valued in the remote workspace.--------------------------------------------------Topic Name: **Remote Team Engagement Games**Top Keywords: online games for remote teams, games for remote teams, retrospective games for remote teams, virtual games for remote teams, agile games for remote teams, fun games for remote teams, icebreaker games for remote teams, best games for remote teams, best online games for remote teams, bingo for remote teamsBlog Post Ideas: **Blog Title: 10 Best Online Games to Boost Team Bonding and Productivity for Remote Teams** Abstract: Remote work can sometimes feel isolating, but the right online games can transform virtual interactions into engaging, team-building experiences. This blog post explores the top 10 online games specifically designed to enhance collaboration, communication, and morale among remote teams. From icebreakers like virtual escape rooms to strategic games like Codenames, discover how these activities can foster a sense of unity while keeping your team productive and entertained. **Blog Title: Agile Retrospective Games for Remote Teams: Making Remote Sprints More Effective and Fun** Abstract: Retrospectives are a cornerstone of Agile methodology, but conducting them remotely can be challenging. This post introduces a collection of virtual retrospective games tailored for remote teams, such as \"Start, Stop, Continue\" and \"Sailboat,\" to make these sessions more interactive and insightful. Learn how these games can help your team reflect on past sprints, identify areas for improvement, and plan for future success in a fun and engaging way. **Blog Title: Virtual Icebreaker Games for Remote Teams: Breaking the Awkward Silence in Minutes** Abstract: Starting meetings with remote teams can often feel awkward, but icebreaker games are the perfect solution to ease tension and encourage participation. This blog post highlights creative and easy-to-implement virtual icebreakers, such as \"Two Truths and a Lie\" and \"Virtual Scavenger Hunt,\" that can quickly warm up your team. Whether you're onboarding new members or just looking to spice up daily stand-ups, these games will help your remote team connect on a personal level and build stronger relationships.--------------------------------------------------\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.354Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":19,"totalLines":98,"estimatedTokens":3613}}190{"id":"doc-end_to_end_rag_using_elasticsearch_and_cohere_co-1f1a9f44","source":"documentation","title":"End-to-end RAG using Elasticsearch and Cohere | Cohere","url":"https://docs.cohere.com/page/elasticsearch-and-cohere","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1pip install elasticsearch_serverless cohere\n```\n\nExample:\n```text\n1from elasticsearch_serverless import Elasticsearch, helpers2from getpass import getpass3import cohere4import json5import requests\n```\n\nExample:\n```text\n1ELASTICSEARCH_ENDPOINT = getpass(\"Elastic Endpoint: \")2ELASTIC_API_KEY = getpass(\"Elastic encoded API key: \") # Use the encoded API key34client = Elasticsearch(5 ELASTICSEARCH_ENDPOINT,6 api_key=ELASTIC_API_KEY7)\n```\n\nExample:\n```text\n1print(client.info())\n```\n\nExample:\n```text\n1COHERE_API_KEY = getpass(\"Enter Cohere API key: \")23client.options(ignore_status=[404]).inference.delete_model(inference_id=\"cohere_embeddings\")45client.inference.put_model(6 task_type=\"text_embedding\",7 inference_id=\"cohere_embeddings\",8 body={9 \"service\": \"cohere\",10 \"service_settings\": {11 \"api_key\": COHERE_API_KEY,12 \"model_id\": \"embed-v4.0\",13 \"embedding_type\": \"int8\",14 \"similarity\": \"cosine\"15 },16 \"task_settings\": {},17 },18)\n```\n\nExample:\n```text\n1client.options(ignore_status=[404]).ingest.delete_pipeline(id=\"cohere_embeddings\")23client.ingest.put_pipeline(4 id=\"cohere_embeddings\",5 description=\"Ingest pipeline for Cohere inference.\",6 processors=[7 {8 \"inference\": {9 \"model_id\": \"cohere_embeddings\",10 \"input_output\": {11 \"input_field\": \"text\",12 \"output_field\": \"text_embedding\",13 },14 }15 }16 ],17)\n```\n\nExample:\n```text\n1client.indices.delete(index=\"cohere-wiki-embeddings\", ignore_unavailable=True)2client.indices.create(3 index=\"cohere-wiki-embeddings\",4 settings={\"index\": {\"default_pipeline\": \"cohere_embeddings\"}},5 mappings={6 \"properties\": {7 \"text_embedding\": {8 \"type\": \"dense_vector\",9 \"dims\": 1024,10 \"element_type\": \"byte\"11 },12 \"text\": {\"type\": \"text\"},13 \"wiki_id\": {\"type\": \"integer\"},14 \"url\": {\"type\": \"text\"},15 \"views\": {\"type\": \"float\"},16 \"langs\": {\"type\": \"integer\"},17 \"title\": {\"type\": \"text\"},18 \"paragraph_id\": {\"type\": \"integer\"},19 \"id\": {\"type\": \"integer\"}20 }21 },22)\n```\n\nExample:\n```text\n1url = \"https://raw.githubusercontent.com/cohere-ai/cohere-developer-experience/main/notebooks/data/embed_jobs_sample_data.jsonl\"2response = requests.get(url)34jsonl_data = response.content.decode('utf-8').splitlines()56documents = []7for line in jsonl_data:8 data_dict = json.loads(line)9 documents.append({10 \"_index\": \"cohere-wiki-embeddings\",11 \"_source\": data_dict,12 }13 )1415helpers.bulk(client, documents)1617print(\"Done indexing documents into `cohere-wiki-embeddings` index!\")\n```\n\nExample:\n```text\n1query = \"When were the semi-finals of the 2022 FIFA world cup played?\"23response = client.search(4 index=\"cohere-wiki-embeddings\",5 size=100,6 knn={7 \"field\": \"text_embedding\",8 \"query_vector_builder\": {9 \"text_embedding\": {10 \"model_id\": \"cohere_embeddings\",11 \"model_text\": query,12 }13 },14 \"k\": 10,15 \"num_candidates\": 50,16 },17 query={18 \"multi_match\": {19 \"query\": query,20 \"fields\": [\"text\", \"title\"]21 }22 }23)2425raw_documents = response[\"hits\"][\"hits\"]2627for document in raw_documents[0:10]:28 print(f'Title: {document[\"_source\"][\"title\"]}\\nText: {document[\"_source\"][\"text\"]}\\n')2930documents = []31for hit in response[\"hits\"][\"hits\"]:32 documents.append(hit[\"_source\"][\"text\"])\n```\n\nExample:\n```text\n1client.options(ignore_status=[404]).inference.delete_model(inference_id=\"cohere_rerank\")23client.inference.put_model(4 task_type=\"rerank\",5 inference_id=\"cohere_rerank\",6 body={7 \"service\": \"cohere\",8 \"service_settings\":{9 \"api_key\": COHERE_API_KEY,10 \"model_id\": \"rerank-english-v3.0\"11 },12 \"task_settings\": {13 \"top_n\": 10,14 },15 }16)\n```\n\nExample:\n```text\n1response = client.inference.inference(2 inference_id=\"cohere_rerank\",3 body={4 \"query\": query,5 \"input\": documents,6 \"task_settings\": {7 \"return_documents\": False8 }9 }10)1112ranked_documents = []13for document in response.body[\"rerank\"]:14 ranked_documents.append({15 \"title\": raw_documents[int(document[\"index\"])][\"_source\"][\"title\"],16 \"text\": raw_documents[int(document[\"index\"])][\"_source\"][\"text\"]17 })1819for document in ranked_documents[0:10]:20 print(f\"Title: {document['title']}\\nText: {document['text']}\\n\")\n```\n\nExample:\n```text\n1co = cohere.Client(COHERE_API_KEY)\n```\n\nExample:\n```text\n1response = co.chat(2 message=query,3 documents=ranked_documents,4 model='command-a-03-2025'5)67source_documents = []8for citation in response.citations:9 for document_id in citation.document_ids:10 if document_id not in source_documents:11 source_documents.append(document_id)1213print(f\"Query: {query}\")14print(f\"Response: {response.text}\")15print(\"Sources:\")16for document in response.documents:17 if document['id'] in source_documents:18 print(f\"{document['title']}: {document['text']}\")\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.362Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":13,"totalLines":68,"estimatedTokens":1394}}191{"id":"doc-usage_patterns_for_tool_use_function_calling_coh-09b1c805","source":"documentation","title":"Usage patterns for tool use (function calling) | Cohere","url":"https://docs.cohere.com/docs/multi-step-tool-use","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere34co = cohere.ClientV2(5 \"COHERE_API_KEY\"6) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1def search_docs(query, top_k=3):2 # Implement any retrieval logic here (vector DB, keyword search, etc.)3 return [4 {5 \"title\": \"Tool use (function calling) overview\",6 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",7 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",8 },9 {10 \"title\": \"Structured outputs\",11 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",12 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",13 },14 {15 \"title\": \"Chat API reference (v2)\",16 \"url\": \"https://docs.cohere.com/reference/chat\",17 \"text\": \"Use the Chat endpoint to generate responses and optionally call tools.\",18 },19 ][:top_k]20 # Return a string or a list of objects. In Step 3, we'll wrap each object into a `document` content block.212223functions_map = {\"search_docs\": search_docs}2425tools = [26 {27 \"type\": \"function\",28 \"function\": {29 \"name\": \"search_docs\",30 \"description\": \"Search documentation and return relevant snippets as documents.\",31 \"parameters\": {32 \"type\": \"object\",33 \"properties\": {34 \"query\": {35 \"type\": \"string\",36 \"description\": \"The search query to look up in the docs.\",37 },38 \"top_k\": {39 \"type\": \"integer\",40 \"description\": \"How many documents to return.\",41 },42 },43 \"required\": [\"query\"],44 },45 },46 },47]\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"user\",4 \"content\": \"Find docs about tool use and structured outputs.\",5 }6]78response = co.chat(9 model=\"command-a-plus-05-2026\", messages=messages, tools=tools10)1112if response.message.tool_calls:13 messages.append(response.message)14 print(response.message.tool_plan, \"\\n\")15 print(response.message.tool_calls)\n```\n\nExample:\n```text\n1I will search the docs for tool use and structured outputs.23[4 ToolCallV2(5 id=\"search_docs_9b0nr4kg58a8\",6 type=\"function\",7 function=ToolCallV2Function(8 name=\"search_docs\", arguments='{\"query\":\"tool use\",\"top_k\":3}'9 ),10 ),11 ToolCallV2(12 id=\"search_docs_0qq0mz9gwnqr\",13 type=\"function\",14 function=ToolCallV2Function(15 name=\"search_docs\", arguments='{\"query\":\"structured outputs\",\"top_k\":3}'16 ),17 ),18]\n```\n\nExample:\n```text\n1import json23if response.message.tool_calls:4 for tc in response.message.tool_calls:5 tool_result = functions_map[tc.function.name](6 **json.loads(tc.function.arguments)7 )8 tool_content = []9 for data in tool_result:10 # Optional: the \"document\" object can take an \"id\" field for use in citations, otherwise auto-generated11 tool_content.append(12 {13 \"type\": \"document\",14 \"document\": {\"data\": json.dumps(data)},15 }16 )17 messages.append(18 {19 \"role\": \"tool\",20 \"tool_call_id\": tc.id,21 \"content\": tool_content,22 }23 )\n```\n\nExample:\n```text\n1messages = [{\"role\": \"user\", \"content\": \"What's 2+2?\"}]23response = co.chat(4 model=\"command-a-plus-05-2026\", messages=messages, tools=tools5)67if response.message.tool_calls:8 print(response.message.tool_plan, \"\\n\")9 print(response.message.tool_calls)1011else:12 print(response.message.content[0].text)\n```\n\nExample:\n```text\n1The answer to 2+2 is 4.\n```\n\nExample:\n```text\n1def search_docs(query, top_k=3):2 # Implement any retrieval logic here (vector DB, keyword search, etc.)3 return [4 {5 \"title\": \"Tool use (function calling) overview\",6 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",7 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",8 },9 {10 \"title\": \"Usage patterns for tool use\",11 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-usage-patterns\",12 \"text\": \"Common patterns include parallel tool calling, multi-step tool use, and more.\",13 },14 {15 \"title\": \"Structured outputs\",16 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",17 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",18 },19 ][:top_k]202122functions_map = {\"search_docs\": search_docs}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"search_docs\",6 \"description\": \"Search documentation and return relevant snippets as documents.\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"query\": {11 \"type\": \"string\",12 \"description\": \"The search query to look up in the docs.\",13 },14 \"top_k\": {15 \"type\": \"integer\",16 \"description\": \"How many documents to return.\",17 },18 },19 \"required\": [\"query\"],20 },21 },22 },23]\n```\n\nExample:\n```text\n1import json23# Step 1: Get the user message4messages = [5 {6 \"role\": \"user\",7 \"content\": \"Explain how tool use works and how to force tool usage. Please cite your sources.\",8 }9]1011# Step 2: Generate tool calls (if any)12model = \"command-a-plus-05-2026\"13response = co.chat(14 model=model, messages=messages, tools=tools, temperature=0.315)1617while response.message.tool_calls:18 print(\"TOOL PLAN:\")19 print(response.message.tool_plan, \"\\n\")20 print(\"TOOL CALLS:\")21 for tc in response.message.tool_calls:22 print(23 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"24 )25 print(\"=\" * 50)2627 messages.append(response.message)2829 # Step 3: Get tool results30 print(\"TOOL RESULT:\")31 for tc in response.message.tool_calls:32 tool_result = functions_map[tc.function.name](33 **json.loads(tc.function.arguments)34 )35 tool_content = []36 print(tool_result)37 for data in tool_result:38 # Optional: the \"document\" object can take an \"id\" field for use in citations, otherwise auto-generated39 tool_content.append(40 {41 \"type\": \"document\",42 \"document\": {\"data\": json.dumps(data)},43 }44 )45 messages.append(46 {47 \"role\": \"tool\",48 \"tool_call_id\": tc.id,49 \"content\": tool_content,50 }51 )5253 # Step 4: Generate response and citations54 response = co.chat(55 model=model,56 messages=messages,57 tools=tools,58 temperature=0.1,59 )6061messages.append(62 {63 \"role\": \"assistant\",64 \"content\": response.message.content[0].text,65 }66)6768# Print final response69print(\"RESPONSE:\")70print(response.message.content[0].text)71print(\"=\" * 50)7273# Print citations (if any)74verbose_source = (75 True # Change to True to display the contents of a source76)77if response.message.citations:78 print(\"CITATIONS:\\n\")79 for citation in response.message.citations:80 print(81 f\"Start: {citation.start}| End:{citation.end}| Text:'{citation.text}' \"82 )83 print(\"Sources:\")84 for idx, source in enumerate(citation.sources):85 print(f\"{idx+1}. {source.id}\")86 if verbose_source:87 print(f\"{source.tool_output}\")88 print(\"\\n\")\n```\n\nExample:\n```text\n1TOOL PLAN:2First, I will search the docs for how tool use works. Then, I will search for how to force tool usage (tool_choice).34TOOL CALLS:5Tool name: search_docs | Parameters: {\"query\":\"tool use\",\"top_k\":3}6==================================================7TOOL RESULT:8[{'title': 'Tool use (function calling) overview', 'url': 'https://docs.cohere.com/v2/docs/tool-use-overview', 'text': 'Tool use connects models to external tools like search engines and APIs.'}]9TOOL PLAN:10Now I'll search for how to force tool usage via the tool_choice parameter.1112TOOL CALLS:13Tool name: search_docs | Parameters: {\"query\":\"tool_choice REQUIRED NONE\",\"top_k\":3}14==================================================15TOOL RESULT:16[{'title': 'Usage patterns for tool use', 'url': 'https://docs.cohere.com/v2/docs/tool-use-usage-patterns', 'text': 'Common patterns include parallel tool calling, multi-step tool use, and more.'}]17RESPONSE:18Tool use lets models call external tools (like doc search) and then answer using tool results with citations. You can force tool usage with tool_choice=\"REQUIRED\" or force a direct response with tool_choice=\"NONE\".19==================================================20CITATIONS:2122Start: 126| End:135| Text:'tool_choice'23Sources:241. search_docs_p0dage9q1nv4:025{'title': 'Usage patterns for tool use', 'url': 'https://docs.cohere.com/v2/docs/tool-use-usage-patterns', 'text': 'Common patterns include parallel tool calling, multi-step tool use, and more.'}\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\",3 messages=messages,4 tools=tools,5 tool_choice=\"REQUIRED\" # optional, to force tool calls6 # tool_choice=\"NONE\" # optional, to force a direct response7)\n```\n\nExample:\n```text\n1from cohere import ToolCallV2, ToolCallV2Function23messages = [4 {5 \"role\": \"user\",6 \"content\": \"How does tool use work in Cohere? Please cite your sources.\",7 },8 {9 \"role\": \"assistant\",10 \"tool_plan\": \"I will search the docs for how tool use works in Cohere.\",11 \"tool_calls\": [12 ToolCallV2(13 id=\"search_docs_1byjy32y4hvq\",14 type=\"function\",15 function=ToolCallV2Function(16 name=\"search_docs\",17 arguments='{\"query\":\"tool use Cohere\",\"top_k\":3}',18 ),19 )20 ],21 },22 {23 \"role\": \"tool\",24 \"tool_call_id\": \"search_docs_1byjy32y4hvq\",25 \"content\": [26 {27 \"type\": \"document\",28 \"document\": {29 \"data\": '{\"title\":\"Tool use (function calling) overview\",\"url\":\"https://docs.cohere.com/v2/docs/tool-use-overview\",\"text\":\"Tool use connects models to external tools like search engines and APIs.\"}'30 },31 }32 ],33 },34 {35 \"role\": \"assistant\",36 \"content\": \"Tool use lets models call external tools (like doc search) and then answer using tool results with citations.\",37 },38]\n```\n\nExample:\n```text\n1messages.append(2 {\"role\": \"user\", \"content\": \"How do I force tool usage?\"}3)45response = co.chat(6 model=\"command-a-plus-05-2026\", messages=messages, tools=tools7)89if response.message.tool_calls:10 messages.append(response.message)11 print(response.message.tool_plan, \"\\n\")12 print(response.message.tool_calls)\n```\n\nExample:\n```text\n1I will search the docs for how to force tool usage using tool_choice.23[ToolCallV2(id='search_docs_8hwpm7d4wr14', type='function', function=ToolCallV2Function(name='search_docs', arguments='{\"query\":\"tool_choice REQUIRED NONE\",\"top_k\":3}'))]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.362Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":3038}}192{"id":"doc-grounded_summarization_using_command_r_cohere-37c63183","source":"documentation","title":"Grounded Summarization Using Command R | Cohere","url":"https://docs.cohere.com/page/grounded-summarization","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1%%capture23import cohere4import networkx as nx5import nltk6nltk.download(\"punkt\")7from nltk.tokenize import sent_tokenize8import numpy as np9import spacy1011from collections import deque12from getpass import getpass13import re14from typing import List, Tuple1516co_api_key = getpass(\"Enter your Cohere API key: \")17co_model = \"command-a-03-2025\"18co = cohere.Client(co_api_key)\n```\n\nExample:\n```text\n1from google.colab import drive2drive.mount(\"/content/drive\", force_remount=True)34fpath = \"drive/Shareddrives/FDE/Cookbooks/Long-form summarisation/ai_and_future_of_work.txt\"5with open(fpath, \"r\") as f:6 text = f.read()78num_tokens = co.tokenize(text).length9print(f\"Loaded IMF report with {num_tokens} tokens\")\n```\n\nExample:\n```text\n1def split_text_into_sentences(text: str) -> List[str]:2 sentences = sent_tokenize(text)3 return sentences45def group_sentences_into_passages(sentence_list: List[str], n_sentences_per_passage: int = 10):6 \"\"\"7 Group sentences into passages of n_sentences sentences.8 \"\"\"9 passages = []10 passage = \"\"11 for i, sentence in enumerate(sentence_list):12 passage += sentence + \" \"13 if (i + 1) % n_sentences_per_passage == 0:14 passages.append(passage)15 passage = \"\"16 return passages1718def build_simple_chunks(text, n_sentences: int = 10):19 \"\"\"20 Build chunks of text from the input text.21 \"\"\"22 sentences = split_text_into_sentences(text)23 chunks = group_sentences_into_passages(sentences, n_sentences_per_passage=n_sentences)24 return chunks25262728def insert_citations(text: str, citations: List[dict]):29 \"\"\"30 A helper function to pretty print citations.31 \"\"\"32 offset = 033 # Process citations in the order they were provided34 for citation in citations:35 # Adjust start/end with offset36 start, end = citation['start'] + offset, citation['end'] + offset37 placeholder = \"[\" + \", \".join(doc[4:] for doc in citation[\"document_ids\"]) + \"]\"38 # ^ doc[4:] removes the 'doc_' prefix, and leaves the quoted document39 modification = f'{text[start:end]} {placeholder}'40 # Replace the cited text with its bolded version + placeholder41 text = text[:start] + modification + text[end:]42 # Update the offset for subsequent replacements43 offset += len(modification) - (end - start)4445 return text46474849def textrank(text: str, co, max_tokens: int, n_sentences_per_passage: int) -> str:50 \"\"\"51 Shortens `text` by extracting key units of text from `text` based on their centrality and concatenating them.52 The output is the concatenation of those key units, in their original order. Centrality is graph-theoretic53 measure of connectedness of a node; the more connected a node is to surrounding nodes (and the more sparsely54 those neighbours are connected), the higher centrality.5556 Key passages are identified via clustering in a three-step process:57 1. Break up `long` into chunks (either sentences or passages, based on `unit`)58 2. Embed each chunk using Cohere's embedding model and construct a similarity matrix59 3. Compute the centrality of each chunk60 4. Keep the highest-centrality chunks until `max_tokens` is reached61 5. Put together shorterned text by reordering chunks in their original order6263 This approach is based on summarise.long_doc_summarization.extraction::extract_single_doc with sorting by64 centrality. Adapted here because installing the `summarise` repo would have added a lot of unused functionalities65 and dependencies.66 \"\"\"6768 # 1. Chunk text into units69 chunks = build_simple_chunks(text, n_sentences_per_passage)7071 # 2. Embed and construct similarity matrix72 embeddings = np.array(73 co.embed(74 texts=chunks,75 model=\"embed-v4.0\",76 input_type=\"clustering\",77 ).embeddings78 )79 similarities = np.dot(embeddings, embeddings.T)8081 # 3. Compute centrality and sort sentences by centrality82 # Easiest to use networkx's `degree` function with similarity as weight83 g = nx.from_numpy_array(similarities, edge_attr=\"weight\")84 centralities = g.degree(weight=\"weight\")85 idcs_sorted_by_centrality = [node for node, degree in sorted(centralities, key=lambda item: item[1], reverse=True)]8687 # 4. Add chunks back in order of centrality88 selected = _add_chunks_by_priority(co, chunks, idcs_sorted_by_centrality, max_tokens)8990 # 5. Put condensed text back in original order91 separator = \"\\n\"92 short = separator.join([chunk for index, chunk in sorted(selected, key=lambda item: item[0], reverse=False)])9394 return short959697def _add_chunks_by_priority(98 co, chunks: List[str], idcs_sorted_by_priority: List[int], max_tokens: int99) -> List[Tuple[int, str]]:100 \"\"\"101 Given chunks of text and their indices sorted by priority (highest priority first), this function102 fills the model context window with as many highest-priority chunks as possible.103104 The output is a list of (index, chunk) pairs, ordered by priority. To stitch back the chunks into105 a cohesive text that preserves chronological order, sort the output on its index.106 \"\"\"107108 selected = []109 num_tokens = 0110 idcs_queue = deque(idcs_sorted_by_priority)111112 while num_tokens < max_tokens and len(idcs_queue) > 0:113 next_idx = idcs_queue.popleft()114 num_tokens += co.tokenize(chunks[next_idx]).length - 2115 # num_tokens += len(tokenizer.encode(chunks[next_idx]).ids) - 2116 # ^ removing BOS and EOS tokens from count117 selected.append((next_idx, chunks[next_idx]))118 # ^ keep index and chunk, to reorder chronologically119 if num_tokens > max_tokens:120 selected.pop()121122 return selected\n```\n\nExample:\n```text\n1prompt_template = \"\"\"\\2## text3{text}45## instructions6Step 1. Read the entire text from the first to the last page.7Step 2. Create a summary of every chapter from the first to the last page.89## summary10\"\"\"1112prompt = prompt_template.format(text=text)13resp = co.chat(14 message=prompt,15 model=co_model,16 temperature=0.3,17 return_prompt=True18)1920num_tokens_in = co.tokenize(resp.prompt).length21num_tokens_out = resp.meta[\"billed_units\"][\"output_tokens\"]22print(f\"Generated summary with {num_tokens_in} tokens in, {num_tokens_out} tokens out\")23print()24print(\"--- Out-of-the-box summary with Command-R ---\")25print()26print(resp.text)\n```\n\nExample:\n```text\n1summarize_preamble = \"\"\"\\2You will receive a series of text fragments from an article that are presented in chronological order. \\3As the assistant, you must generate responses to user's requests based on the information given in the fragments. \\4Ensure that your responses are accurate and truthful, and that you reference your sources where appropriate to answer \\5the queries, regardless of their complexity.\\6\"\"\"78instructions = \"\"\"\\9## instructions10Step 1. Read the entire text from the first to the last page.11Step 2. Create a summary of every chapter from the first to the last page.12\"\"\"1314chunked = build_simple_chunks(text, n_sentences=30)15resp = co.chat(16 preamble=summarize_preamble,17 message=instructions,18 documents=[{\"text\": chunk} for chunk in chunked],19 model=co_model,20 temperature=0.3,21 return_prompt=True22)2324num_tokens_in = co.tokenize(resp.prompt).length25num_tokens_out = resp.meta[\"billed_units\"][\"output_tokens\"]26print(f\"Generated summary with {num_tokens_in} tokens in, {num_tokens_out} tokens out\")27print()28print(\"--- Summary with citations using grounded generation in Command-R ---\")29print()30print(resp.text)\n```\n\nExample:\n```text\n1print(insert_citations(resp.text, resp.citations))\n```\n\nExample:\n```text\nAround 40% of employment worldwide is exposed to AI [1, 6]\n```\n\nExample:\n```text\n1print(chunked[6])\n```\n\nExample:\n```text\n1num_tokens = 81922shortened = textrank(text, co, num_tokens, n_sentences_per_passage=30)34chunked = build_simple_chunks(shortened)5resp = co.chat(6 message=instructions,7 documents=[{\"text\": chunk} for chunk in chunked],8 model=co_model,9 temperature=0.3,10 return_prompt=True11)1213num_tokens_in = co.tokenize(resp.prompt).length14num_tokens_out = resp.meta[\"billed_units\"][\"output_tokens\"]15print(f\"Generated summary with {num_tokens_in} tokens in, {num_tokens_out} tokens out\")16print()17print(\"--- Summary with citations using text-rank + grounding in Command-R ---\")18print()19print(resp.text)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.364Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":9,"totalLines":48,"estimatedTokens":2186}}193{"id":"doc-serverless_semantic_search_with_cohere_and_pinec-6406d122","source":"documentation","title":"Serverless Semantic Search with Cohere and Pinecone | Cohere","url":"https://docs.cohere.com/page/embed-jobs-serverless-pinecone","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import os2import json3import time4import numpy as np5import cohere6from pinecone import Pinecone78co = cohere.Client('COHERE_API_KEY')9pc = Pinecone(10 api_key=\"PINECONE_API_KEY\",11 source_tag=\"cohere\"12)\n```\n\nExample:\n```text\n/usr/local/lib/python3.10/dist-packages/pinecone/data/index.py:1: TqdmExperimentalWarning: Using `tqdm.autonotebook.tqdm` in notebook mode. Use `tqdm.tqdm` instead to force console mode (e.g. in jupyter console) from tqdm.autonotebook import tqdm\n```\n\nExample:\n```text\n1dataset_file_path = \"data/embed_jobs_sample_data.jsonl\" # Full path - https://raw.githubusercontent.com/cohere-ai/cohere-developer-experience/main/notebooks/data/embed_jobs_sample_data.jsonl23ds=co.create_dataset(4\tname='sample_file',5\t# insert your file path here - you can upload it on the right - we accept .csv and jsonl files6\tdata=open(dataset_file_path, 'rb'),7\tdataset_type=\"embed-input\"8\t)910print(ds.await_validation())\n```\n\nExample:\n```text\nuploading file, starting validation...sample-file-2gwgxq was uploaded...cohere.Dataset { id: sample-file-2gwgxq name: sample_file dataset_type: embed-input validation_status: validated created_at: 2024-01-13 02:47:32.563080 updated_at: 2024-01-13 02:47:32.563081 download_urls: [''] validation_error: None validation_warnings: []}\n```\n\nExample:\n```text\n1job = co.create_embed_job(dataset_id=ds.id,2 input_type='search_document',3 model='embed-english-v3.0',4 embeddings_types=['float'])56job.wait() # poll the server until the job is completed\n```\n\nExample:\n```text\n......\n```\n\nExample:\n```text\n1print(job)\n```\n\nExample:\n```text\n$cohere.EmbedJob {> job_id: 6d691fbe-e026-436a-826a-16e70b293e51> status: complete> created_at: 2024-01-13T02:47:46.385016Z> input_dataset_id: sample-file-2gwgxq> output_urls: None> model: embed-english-v3.0> truncate: RIGHT> percent_complete: 100> output: cohere.Dataset {> id: embeded-sample-file-mdse2h> name: embeded-sample-file> dataset_type: embed-result> validation_status: validated> created_at: 2024-01-13 02:47:47.850097> updated_at: 2024-01-13 02:47:47.850097> download_urls: ['']> validation_error: None> validation_warnings: []> }>}\n```\n\nExample:\n```text\n1output_dataset=co.get_dataset(job.output.id)2data_array = []3for record in output_dataset:4 data_array.append(record)56ids = [str(i) for i in range(len(data_array))]7meta = [{'text':str(data_array[i]['text'])} for i in range(len(data_array))]8embeds=[np.float32(data_array[i]['embeddings']['float']) for i in range(len(data_array))]910to_upsert = list(zip(ids, embeds, meta))\n```\n\nExample:\n```text\n1from pinecone import ServerlessSpec23index_name = \"embed-jobs-serverless-test-example\"45pc.create_index(6name=index_name,7dimension=1024,8metric=\"cosine\",9spec=ServerlessSpec(cloud='aws', region='us-west-2')10)1112idx = pc.Index(index_name)\n```\n\nExample:\n```text\n1batch_size = 12823for i in range(0, len(data_array), batch_size):4 i_end = min(i+batch_size, len(data_array))5 idx.upsert(vectors=to_upsert[i:i_end])67print(idx.describe_index_stats())\n```\n\nExample:\n```text\n{'dimension': 1024,'index_fullness': 0.0,'namespaces': {'': {'vector_count': 3664}},'total_vector_count': 3664}\n```\n\nExample:\n```text\n1query = \"What did Microsoft announce in Las Vegas?\"23xq = co.embed(4 texts=[query],5 model='embed-english-v3.0',6 input_type='search_query',7 truncate='END'8).embeddings910print(np.array(xq).shape)1112res = idx.query(xq, top_k=20, include_metadata=True)\n```\n\nExample:\n```text\n(1, 1024)\n```\n\nExample:\n```text\n1for match in res['matches']:2 print(f\"{match['score']:.2f}: {match['metadata']['text']}\")\n```\n\nExample:\n```text\n0.48: On October 22, 2012, Microsoft announced the release of new features including co-authoring, performance improvements and touch support.0.45: On May 2, 2019, at F8, the company announced its new vision with the tagline \"the future is private\". A redesign of the website and mobile app was introduced, dubbed as \"FB5\". The event also featured plans for improving groups, a dating platform, end-to-end encryption on its platforms, and allowing users on Messenger to communicate directly with WhatsApp and Instagram users.0.42: On July 13, 2009, Microsoft announced at its Worldwide Partners Conference 2009 in New Orleans that Microsoft Office 2010 reached its \"Technical Preview\" development milestone and features of Office Web Apps were demonstrated to the public for the first time. Additionally, Microsoft announced that Office Web Apps would be made available to consumers online and free of charge, while Microsoft Software Assurance customers will have the option of running them on premises. Office 2010 beta testers were not given access to Office Web Apps at this date, and it was announced that it would be available for testers during August 2009. However, in August 2009, a Microsoft spokesperson stated that there had been a delay in the release of Office Web Apps Technical Preview and it would not be available by the end of August.0.42: On January 17, 2017, Facebook COO Sheryl Sandberg planned to open Station F, a startup incubator campus in Paris, France. On a six-month cycle, Facebook committed to work with ten to 15 data-driven startups there. On April 18, Facebook announced the beta launch of at its annual F8 developer conference. Facebook Spaces is a virtual reality version of Facebook for Oculus VR goggles. In a virtual and shared space, users can access a curated selection of 360-degree photos and videos using their avatar, with the support of the controller. Users can access their own photos and videos, along with media shared on their newsfeed. In September, Facebook announced it would spend up to US$1 billion on original shows for its Facebook Watch platform. On October 16, it acquired the anonymous compliment app tbh, announcing its intention to leave the app independent.0.41: On September 26, 2017, Microsoft announced that the next version of the suite for Windows desktop, Office 2019, was in development. On April 27, 2018, Microsoft released Office 2019 Commercial Preview for Windows 10. It was released to general availability for Windows 10 and for macOS on September 24, 2018.0.41: Microsoft Office, or simply Office, is the former name of a family of client software, server software, and services developed by Microsoft. It was first announced by Bill Gates on August 1, 1988, at COMDEX in Las Vegas. Initially a marketing term for an office suite (bundled set of productivity applications), the first version of Office contained Microsoft Word, Microsoft Excel, and Microsoft PowerPoint. Over the years, Office applications have grown substantially closer with shared features such as a common spell checker, Object Linking and Embedding data integration and Visual Basic for Applications scripting language. Microsoft also positions Office as a development platform for line-of-business software under the Office Business Applications brand.0.40: On August 12, 2009, it was announced that Office Mobile would also be released for the Symbian platform as a joint agreement between Microsoft and Nokia. It was the first time Microsoft would develop Office mobile applications for another smartphone platform. The first application to appear on Nokia Eseries smartphones was Microsoft Office Communicator. In February 2012, Microsoft released OneNote, Lync 2010, Document Connection and PowerPoint Broadcast for Symbian. In April, Word Mobile, PowerPoint Mobile and Excel Mobile joined the Office Suite.0.40: In 2010, Microsoft introduced a software as a service platform known as Office 365, to provide cloud-hosted versions of Office's server software, including Exchange e-mail and SharePoint, on a subscription basis (competing in particular with Google Apps). Following the release of Office 2013, Microsoft began to offer Office 365 plans for the consumer market, with access to Microsoft Office software on multiple devices with free feature updates over the life of the subscription, as well as other services such as OneDrive storage.0.40: On April 12, 2016, Zuckerberg outlined his 10-year vision, which rested on three main pillars: artificial intelligence, increased global connectivity, and virtual and augmented reality. In July, a suit was filed against the company alleging that it permitted Hamas to use it to perform assaults that cost the lives of four people. Facebook released its blueprints of Surround 360 camera on GitHub under an open-source license. In September, it won an Emmy for its animated short \"Henry\". In October, Facebook announced a fee-based communications tool called Workplace that aims to \"connect everyone\" at work. Users can create profiles, see updates from co-workers on their news feed, stream live videos and participate in secure group chats.0.40: On January 22, 2015, the Microsoft Office blog announced that the next version of the suite for Windows desktop, Office 2016, was in development. On May 4, 2015, a public preview of Microsoft Office 2016 was released. Office 2016 was released for Mac OS X on July 9, 2015 and for Windows on September 22, 2015.0.39: On November 6, 2013, Microsoft announced further new features including \"real-time\" co-authoring and an Auto-Save feature in Word (replacing the save button).0.39: In February 2014, Office Web Apps were re-branded Office Online and incorporated into other Microsoft web services, including Calendar, OneDrive, Outlook.com, and People. Microsoft had previously attempted to unify its online services suite (including Microsoft Passport, Hotmail, MSN Messenger, and later SkyDrive) under a brand known as Windows Live, first launched in 2005. However, with the impending launch of Windows 8 and its increased use of cloud services, Microsoft dropped the Windows Live brand to emphasize that these services would now be built directly into Windows and not merely be a \"bolted on\" add-on. Critics had criticized the Windows Live brand for having no clear vision, as it was being applied to an increasingly broad array of unrelated services. At the same time, Windows Live Hotmail was re-launched as Outlook.com (sharing its name with the Microsoft Outlook personal information manager).0.39: On February 18, 2021, Microsoft announced that the next version of the suite for Windows desktop, Office 2021, was in development. This new version will be supported for five years and was released on October 5, 2021.0.38: Since Office 2013, Microsoft has promoted Office 365 as the primary means of obtaining Microsoft Office: it allows the use of the software and other services on a subscription business model, and users receive feature updates to the software for the lifetime of the subscription, including new features and cloud computing integration that are not necessarily included in the \"on-premises\" releases of Office sold under conventional license terms. In 2017, revenue from Office 365 overtook conventional license sales. Microsoft also rebranded most of their standard Office 365 editions as \"Microsoft 365\" to reflect their inclusion of features and services beyond the core Microsoft Office suite.0.38: Microsoft has since promoted Office 365 as the primary means of purchasing Microsoft Office. Although there are still \"on-premises\" releases roughly every three years, Microsoft marketing emphasizes that they do not receive new features or access to new cloud-based services as they are released unlike Office 365, as well as other benefits for consumer and business markets. Office 365 revenue overtook traditional license sales for Office in 2017.0.38: A technical preview of Microsoft Office 2013 (Build 15.0.3612.1010) was released on January 30, 2012, and a Customer Preview version was made available to consumers on July 16, 2012. It sports a revamped application interface; the interface is based on Metro, the interface of Windows Phone and Windows 8. Microsoft Outlook has received the most pronounced changes so far; for example, the Metro interface provides a new visualization for scheduled tasks. PowerPoint includes more templates and transition effects, and OneNote includes a new splash screen.0.38: On January 21, 2015, during the \"Windows 10: The Next Chapter\" press event, Microsoft unveiled Office for Windows 10, Windows Runtime ports of the Android and iOS versions of the Office Mobile suite. Optimized for smartphones and tablets, they are universal apps that can run on both Windows and Windows for phones, and share similar underlying code. A simplified version of Outlook was also added to the suite. They will be bundled with Windows 10 mobile devices, and available from the Windows Store for the PC version of Windows 10. Although the preview versions were free for most editing, the release versions will require an Office 365 subscription on larger tablets (screen size larger than 10.1 inches) and desktops for editing, as with large Android tablets. Smaller tablets and phones will have most editing features for free.0.38: In May 2018 at F8, the company announced it would offer its own dating service. Shares in competitor Match Group fell by 22%. Facebook Dating includes privacy features and friends are unable to view their friends' dating profile. In July, Facebook was charged £500,000 by UK watchdogs for failing to respond to data erasure requests. On July 18, Facebook established a subsidiary named Lianshu Science & Technology in Hangzhou City, China, with $30 million ($ in dollars) of capital. All its shares are held by Facebook Hong. Approval of the registration of the subsidiary was then withdrawn, due to a disagreement between officials in Zhejiang province and the Cyberspace Administration of China. On July 26, Facebook became the first company to lose over $100 billion ($ in dollars) worth of market capitalization in one day, dropping from nearly $630 billion to $510 billion after disappointing sales reports. On July 31, Facebook said that the company had deleted 17 accounts related to the 2018 U.S. midterm elections. On September 19, Facebook announced that, for news distribution outside the United States, it would work with U.S. funded democracy promotion organizations, International Republican Institute and the National Democratic Institute, which are loosely affiliated with the Republican and Democratic parties. Through the Digital Forensic Research Lab Facebook partners with the Atlantic Council, a NATO-affiliated think tank. In November, Facebook launched smart displays branded Portal and Portal Plus (Portal+). They support Amazon's Alexa (intelligent personal assistant service). The devices include video chat function with Facebook Messenger.0.37: The first Preview version of Microsoft Office 2016 for Mac was released on March 5, 2015. On July 9, 2015, Microsoft released the final version of Microsoft Office 2016 for Mac which includes Word, Excel, PowerPoint, Outlook and OneNote. It was immediately made available for Office 365 subscribers with either a Home, Personal, Business, Business Premium, E3 or ProPlus subscription. A non–Office 365 edition of Office 2016 was made available as a one-time purchase option on September 22, 2015.0.37: In October 2022, Microsoft announced that it will phase out the Microsoft Office brand in favor of \"Microsoft 365\" by January 2023. The name will continue to be used for legacy product offerings.\n```\n\nExample:\n```text\n1docs =[match['metadata']['text'] for match in res['matches']]23rerank_response = co.rerank(4 model = 'rerank-english-v2.0',5 query = query,6 documents = docs,7 top_n = 3,8)9for response in rerank_response:10 print(f\"{response.relevance_score:.2f}: {response.document['text']}\")\n```\n\nExample:\n```text\n0.99: Microsoft Office, or simply Office, is the former name of a family of client software, server software, and services developed by Microsoft. It was first announced by Bill Gates on August 1, 1988, at COMDEX in Las Vegas. Initially a marketing term for an office suite (bundled set of productivity applications), the first version of Office contained Microsoft Word, Microsoft Excel, and Microsoft PowerPoint. Over the years, Office applications have grown substantially closer with shared features such as a common spell checker, Object Linking and Embedding data integration and Visual Basic for Applications scripting language. Microsoft also positions Office as a development platform for line-of-business software under the Office Business Applications brand.0.93: On January 21, 2015, during the \"Windows 10: The Next Chapter\" press event, Microsoft unveiled Office for Windows 10, Windows Runtime ports of the Android and iOS versions of the Office Mobile suite. Optimized for smartphones and tablets, they are universal apps that can run on both Windows and Windows for phones, and share similar underlying code. A simplified version of Outlook was also added to the suite. They will be bundled with Windows 10 mobile devices, and available from the Windows Store for the PC version of Windows 10. Although the preview versions were free for most editing, the release versions will require an Office 365 subscription on larger tablets (screen size larger than 10.1 inches) and desktops for editing, as with large Android tablets. Smaller tablets and phones will have most editing features for free.0.87: In October 2022, Microsoft announced that it will phase out the Microsoft Office brand in favor of \"Microsoft 365\" by January 2023. The name will continue to be used for legacy product offerings.\n```\n\nExample:\n```text\n1query = \"What was the first youtube video about?\"23xq = co.embed(4 texts=[query],5 model='embed-english-v3.0',6 input_type='search_query',7 truncate='END'8).embeddings910print(np.array(xq).shape)1112res = idx.query(xq, top_k=20, include_metadata=True)1314for match in res['matches']:15 print(f\"{match['score']:.2f}: {match['metadata']['text']}\")\n```\n\nExample:\n```text\n(1, 1024)0.66: YouTube began as a venture capital–funded technology startup. Between November 2005 and April 2006, the company raised money from various investors, with Sequoia Capital, $11.5 million, and Artis Capital Management, $8 million, being the largest two. YouTube's early headquarters were situated above a pizzeria and a Japanese restaurant in San Mateo, California. In February 2005, the company activated codice_1. The first video was uploaded April 23, 2005. Titled \"Me at the zoo\", it shows co-founder Jawed Karim at the San Diego Zoo and can still be viewed on the site. In May, the company launched a public beta and by November, a Nike ad featuring Ronaldinho became the first video to reach one million total views. The site launched officially on December 15, 2005, by which time the site was receiving 8 million views a day. Clips at the time were limited to 100 megabytes, as little as 30 seconds of footage.0.58: Karim said the inspiration for YouTube first came from the Super Bowl XXXVIII halftime show controversy when Janet Jackson's breast was briefly exposed by Justin Timberlake during the halftime show. Karim could not easily find video clips of the incident and the 2004 Indian Ocean Tsunami online, which led to the idea of a video-sharing site. Hurley and Chen said that the original idea for YouTube was a video version of an online dating service, and had been influenced by the website Hot or Not. They created posts on Craigslist asking attractive women to upload videos of themselves to YouTube in exchange for a $100 reward. Difficulty in finding enough dating videos led to a change of plans, with the site's founders deciding to accept uploads of any video.0.55: YouTube was not the first video-sharing site on the Internet; Vimeo was launched in November 2004, though that site remained a side project of its developers from CollegeHumor at the time and did not grow much, either. The week of YouTube's launch, NBC-Universal's \"Saturday Night Live\" ran a skit \"Lazy Sunday\" by The Lonely Island. Besides helping to bolster ratings and long-term viewership for \"Saturday Night Live\", \"Lazy Sunday\"'s status as an early viral video helped establish YouTube as an important website. Unofficial uploads of the skit to YouTube drew in more than five million collective views by February 2006 before they were removed when NBCUniversal requested it two months later based on copyright concerns. Despite eventually being taken down, these duplicate uploads of the skit helped popularize YouTube's reach and led to the upload of more third-party content. The site grew rapidly; in July 2006, the company announced that more than 65,000 new videos were being uploaded every day and that the site was receiving 100 million video views per day.0.55: According to a story that has often been repeated in the media, Hurley and Chen developed the idea for YouTube during the early months of 2005, after they had experienced difficulty sharing videos that had been shot at a dinner party at Chen's apartment in San Francisco. Karim did not attend the party and denied that it had occurred, but Chen remarked that the idea that YouTube was founded after a dinner party \"was probably very strengthened by marketing ideas around creating a story that was very digestible\".0.53: In December 2009, YouTube partnered with Vevo. In April 2010, Lady Gaga's \"Bad Romance\" became the most viewed video, becoming the first video to reach 200 million views on May 9, 2010.0.53: YouTube is a global online video sharing and social media platform headquartered in San Bruno, California. It was launched on February 14, 2005, by Steve Chen, Chad Hurley, and Jawed Karim. It is owned by Google, and is the second most visited website, after Google Search. YouTube has more than 2.5 billion monthly users who collectively watch more than one billion hours of videos each day. , videos were being uploaded at a rate of more than 500 hours of content per minute.0.53: YouTube has faced numerous challenges and criticisms in its attempts to deal with copyright, including the site's first viral video, Lazy Sunday, which had to be taken down, due to copyright concerns. At the time of uploading a video, YouTube users are shown a message asking them not to violate copyright laws. Despite this advice, many unauthorized clips of copyrighted material remain on YouTube. YouTube does not view videos before they are posted online, and it is left to copyright holders to issue a DMCA takedown notice pursuant to the terms of the Online Copyright Infringement Liability Limitation Act. Any successful complaint about copyright infringement results in a YouTube copyright strike. Three successful complaints for copyright infringement against a user account will result in the account and all of its uploaded videos being deleted. From 2007 to 2009 organizations including Viacom, Mediaset, and the English Premier League have filed lawsuits against YouTube, claiming that it has done too little to prevent the uploading of copyrighted material.0.51: Some YouTube videos have themselves had a direct effect on world events, such as \"Innocence of Muslims\" (2012) which spurred protests and related anti-American violence internationally. TED curator Chris Anderson described a phenomenon by which geographically distributed individuals in a certain field share their independently developed skills in YouTube videos, thus challenging others to improve their own skills, and spurring invention and evolution in that field. Journalist Virginia Heffernan stated in \"The New York Times\" that such videos have \"surprising implications\" for the dissemination of culture and even the future of classical music.0.50: Observing that face-to-face communication of the type that online videos convey has been \"fine-tuned by millions of years of evolution,\" TED curator Chris Anderson referred to several YouTube contributors and asserted that \"what Gutenberg did for writing, online video can now do for face-to-face communication.\" Anderson asserted that it is not far-fetched to say that online video will dramatically accelerate scientific advance, and that video contributors may be about to launch \"the biggest learning cycle in human history.\" In education, for example, the Khan Academy grew from YouTube video tutoring sessions for founder Salman Khan's cousin into what \"Forbes\" Michael Noer called \"the largest school in the world,\" with technology poised to disrupt how people learn. YouTube was awarded a 2008 George Foster Peabody Award, the website being described as a Speakers' Corner that \"both embodies and promotes democracy.\" \"The Washington Post\" reported that a disproportionate share of YouTube's most subscribed channels feature minorities, contrasting with mainstream television in which the stars are largely white. A Pew Research Center study reported the development of \"visual journalism,\" in which citizen eyewitnesses and established news organizations share in content creation. The study also concluded that YouTube was becoming an important platform by which people acquire news.0.50: YouTube was founded by Steve Chen, Chad Hurley, and Jawed Karim. The trio were early employees of PayPal, which left them enriched after the company was bought by eBay. Hurley had studied design at the Indiana University of Pennsylvania, and Chen and Karim studied computer science together at the University of Illinois Urbana-Champaign.0.49: In 2013, YouTube teamed up with satirical newspaper company \"The Onion\" to claim in an uploaded video that the video-sharing website was launched as a contest which had finally come to an end, and would shut down for ten years before being re-launched in 2023, featuring only the winning video. The video starred several YouTube celebrities, including Antoine Dodson. A video of two presenters announcing the nominated videos streamed live for 12 hours.0.48: Since its purchase by Google, YouTube has expanded beyond the core website into mobile apps, network television, and the ability to link with other platforms. Video categories on YouTube include music videos, video clips, news, short films, feature films, documentaries, audio recordings, movie trailers, teasers, live streams, vlogs, and more. Most content is generated by individuals, including collaborations between YouTubers and corporate sponsors. Established media corporations such as Disney, Paramount, and Warner Bros. Discovery have also created and expanded their corporate YouTube channels to advertise to a larger audience.0.47: YouTube has enabled people to more directly engage with government, such as in the CNN/YouTube presidential debates (2007) in which ordinary people submitted questions to U.S. presidential candidates via YouTube video, with a techPresident co-founder saying that Internet video was changing the political landscape. Describing the Arab Spring (2010–2012), sociologist Philip N. Howard quoted an activist's succinct description that organizing the political unrest involved using \"Facebook to schedule the protests, Twitter to coordinate, and YouTube to tell the world.\" In 2012, more than a third of the U.S. Senate introduced a resolution condemning Joseph Kony 16 days after the \"Kony 2012\" video was posted to YouTube, with resolution co-sponsor Senator Lindsey Graham remarking that the video \"will do more to lead to (Kony's) demise than all other action combined.\"0.47: YouTube carried out early experiments with live streaming, including a concert by U2 in 2009, and a question-and-answer session with US President Barack Obama in February 2010. These tests had relied on technology from 3rd-party partners, but in September 2010, YouTube began testing its own live streaming infrastructure. In April 2011, YouTube announced the rollout of \"YouTube Live\". The creation of live streams was initially limited to select partners. It was used for real-time broadcasting of events such as the 2012 Olympics in London. In October 2012, more than 8 million people watched Felix Baumgartner's jump from the edge of space as a live stream on YouTube.0.46: In June 2007, YouTube began trials of a system for automatic detection of uploaded videos that infringe copyright. Google CEO Eric Schmidt regarded this system as necessary for resolving lawsuits such as the one from Viacom, which alleged that YouTube profited from content that it did not have the right to distribute. The system, which was initially called \"Video Identification\" and later became known as Content ID, creates an ID File for copyrighted audio and video material, and stores it in a database. When a video is uploaded, it is checked against the database, and flags the video as a copyright violation if a match is found. When this occurs, the content owner has the choice of blocking the video to make it unviewable, tracking the viewing statistics of the video, or adding advertisements to the video.0.46: In January 2009, YouTube launched \"YouTube for TV\", a version of the website tailored for set-top boxes and other TV-based media devices with web browsers, initially allowing its videos to be viewed on the PlayStation 3 and Wii video game consoles.0.46: In September 2012, YouTube launched its first app for the iPhone, following the decision to drop YouTube as one of the preloaded apps in the iPhone 5 and iOS 6 operating system. According to GlobalWebIndex, YouTube was used by 35% of smartphone users between April and June 2013, making it the third-most used app.0.46: Conversely, YouTube has also allowed government to more easily engage with citizens, the White House's official YouTube channel being the seventh top news organization producer on YouTube in 2012 and in 2013 a healthcare exchange commissioned Obama impersonator Iman Crosson's YouTube music video spoof to encourage young Americans to enroll in the Affordable Care Act (Obamacare)-compliant health insurance. In February 2014, U.S. President Obama held a meeting at the White House with leading YouTube content creators to not only promote awareness of Obamacare but more generally to develop ways for government to better connect with the \"YouTube Generation.\" Whereas YouTube's inherent ability to allow presidents to directly connect with average citizens was noted, the YouTube content creators' new media savvy was perceived necessary to better cope with the website's distracting content and fickle audience.0.46: Later that year, YouTube came under criticism for showing inappropriate videos targeted at children and often featuring popular characters in violent, sexual or otherwise disturbing situations, many of which appeared on YouTube Kids and attracted millions of views. The term \"Elsagate\" was coined on the Internet and then used by various news outlets to refer to this controversy. On November 11, 2017, YouTube announced it was strengthening site security to protect children from unsuitable content. Later that month, the company started to mass delete videos and channels that made improper use of family-friendly characters. As part of a broader concern regarding child safety on YouTube, the wave of deletions also targeted channels that showed children taking part in inappropriate or dangerous activities under the guidance of adults. Most notably, the company removed \"Toy Freaks\", a channel with over 8.5 million subscribers, that featured a father and his two daughters in odd and upsetting situations. According to analytics specialist SocialBlade, it earned up to £8.7 million annually prior to its deletion.0.45: In September 2020, YouTube announced that it would be launching a beta version of a new platform of 15-second videos, similar to TikTok, called YouTube Shorts. The platform was first tested in India but as of March 2021 has expanded to other countries including the United States with videos now able to be up to 1 minute long. The platform is not a standalone app, but is integrated into the main YouTube app. Like TikTok, it gives users access to built-in creative tools, including the possibility of adding licensed music to their videos. The platform had its global beta launch in July 2021.\n```\n\nExample:\n```text\n0.95: YouTube began as a venture capital–funded technology startup. Between November 2005 and April 2006, the company raised money from various investors, with Sequoia Capital, $11.5 million, and Artis Capital Management, $8 million, being the largest two. YouTube's early headquarters were situated above a pizzeria and a Japanese restaurant in San Mateo, California. In February 2005, the company activated codice_1. The first video was uploaded April 23, 2005. Titled \"Me at the zoo\", it shows co-founder Jawed Karim at the San Diego Zoo and can still be viewed on the site. In May, the company launched a public beta and by November, a Nike ad featuring Ronaldinho became the first video to reach one million total views. The site launched officially on December 15, 2005, by which time the site was receiving 8 million views a day. Clips at the time were limited to 100 megabytes, as little as 30 seconds of footage.0.92: Karim said the inspiration for YouTube first came from the Super Bowl XXXVIII halftime show controversy when Janet Jackson's breast was briefly exposed by Justin Timberlake during the halftime show. Karim could not easily find video clips of the incident and the 2004 Indian Ocean Tsunami online, which led to the idea of a video-sharing site. Hurley and Chen said that the original idea for YouTube was a video version of an online dating service, and had been influenced by the website Hot or Not. They created posts on Craigslist asking attractive women to upload videos of themselves to YouTube in exchange for a $100 reward. Difficulty in finding enough dating videos led to a change of plans, with the site's founders deciding to accept uploads of any video.0.91: YouTube was not the first video-sharing site on the Internet; Vimeo was launched in November 2004, though that site remained a side project of its developers from CollegeHumor at the time and did not grow much, either. The week of YouTube's launch, NBC-Universal's \"Saturday Night Live\" ran a skit \"Lazy Sunday\" by The Lonely Island. Besides helping to bolster ratings and long-term viewership for \"Saturday Night Live\", \"Lazy Sunday\"'s status as an early viral video helped establish YouTube as an important website. Unofficial uploads of the skit to YouTube drew in more than five million collective views by February 2006 before they were removed when NBCUniversal requested it two months later based on copyright concerns. Despite eventually being taken down, these duplicate uploads of the skit helped popularize YouTube's reach and led to the upload of more third-party content. The site grew rapidly; in July 2006, the company announced that more than 65,000 new videos were being uploaded every day and that the site was receiving 100 million video views per day.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.368Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":21,"totalLines":108,"estimatedTokens":8798}}194{"id":"doc-hello_world_explore_language_ai_with_cohere_cohe-3e0f58ff","source":"documentation","title":"Hello World! Explore Language AI with Cohere | Cohere","url":"https://docs.cohere.com/page/hello-world-meet-ai","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1! pip install cohere altair umap-learn -q\n```\n\nExample:\n```text\n1import cohere2import pandas as pd3import numpy as np4import altair as alt56co = cohere.Client(\"COHERE_API_KEY\") # Get your API key: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1prompt = \"What is a Hello World program.\"23response = co.chat(4 message=prompt,5 model='command-r')67print(response.text)\n```\n\nExample:\n```text\nA \"Hello World\" program is a traditional and simple program that is often used as an introduction to a new programming language. The program typically displays the message \"Hello World\" as its output. The concept of a \"Hello World\" program originated from the book *The C Programming Language* written by Kernighan and Ritchie, where the example program in the book displayed the message using the C programming language.The \"Hello World\" program serves as a basic and straightforward way to verify that your development environment is set up correctly and to familiarize yourself with the syntax and fundamentals of the programming language. It's a starting point for learning how to write and run programs in a new language.The program's simplicity makes it accessible to programmers of all skill levels, and it's often one of the first programs beginners write when learning to code. The exact implementation of a \"Hello World\" program varies depending on the programming language being used, but the core idea remains the same—to display the \"Hello World\" message.Here's how a \"Hello World\" program can be written in a few select languages:1. **C**:```c#include <stdio.h>int main() { printf(\"Hello World\\n\"); return 0;}```2. **Python**:```python PYTHONprint(\"Hello World\")```3. **Java**:```java JAVAclass HelloWorld { public static void main(String[] args) { System.out.println(\"Hello World\"); }}```4. **JavaScript**:```javascriptconsole.log(\"Hello World\");```5. **C#**:```csharpusing System;class Program { static void Main() { Console.WriteLine(\"Hello World\"); }}```The \"Hello World\" program is a testament to the power of programming, as a simple and concise message can be displayed in numerous languages with just a few lines of code. It's an exciting first step into the world of software development!\n```\n\nExample:\n```text\n1prompt = \"\"\"2Write the first paragraph of a blog post given a blog title.3--4Blog Title: Best Activities in Toronto5First Paragraph: Looking for fun things to do in Toronto? When it comes to exploring Canada's6largest city, there's an ever-evolving set of activities to choose from. Whether you're looking to7visit a local museum or sample the city's varied cuisine, there is plenty to fill any itinerary. In8this blog post, I'll share some of my favorite recommendations9--10Blog Title: Mastering Dynamic Programming11First Paragraph: In this piece, we'll help you understand the fundamentals of dynamic programming,12and when to apply this optimization technique. We'll break down bottom-up and top-down approaches to13solve dynamic programming problems.14--15Blog Title: Learning to Code with Hello, World!16First Paragraph:\"\"\"1718response = co.chat(19 message=prompt,20 model='command-r')2122print(response.text)\n```\n\nExample:\n```text\nStarting to code can be daunting, but it's actually simpler than you think! The famous first program, \"Hello, World!\" is a rite of passage for all coders, and an excellent starting point to begin your coding journey. This blog will guide you through the process of writing your very first line of code, and help you understand why learning to code is an exciting and valuable skill to have, covering the fundamentals and the broader implications of this seemingly simple phrase.\n```\n\nExample:\n```text\n1def generate_text(topic):2 prompt = f\"\"\"3Write the first paragraph of a blog post given a blog title.4--5Blog Title: Best Activities in Toronto6First Paragraph: Looking for fun things to do in Toronto? When it comes to exploring Canada's7largest city, there's an ever-evolving set of activities to choose from. Whether you're looking to8visit a local museum or sample the city's varied cuisine, there is plenty to fill any itinerary. In9this blog post, I'll share some of my favorite recommendations10--11Blog Title: Mastering Dynamic Programming12First Paragraph: In this piece, we'll help you understand the fundamentals of dynamic programming,13and when to apply this optimization technique. We'll break down bottom-up and top-down approaches to14solve dynamic programming problems.15--16Blog Title: {topic}17First Paragraph:\"\"\"18 # Generate text by calling the Chat endpoint19 response = co.chat(20 message=prompt,21 model='command-r')2223 return response.text\n```\n\nExample:\n```text\n1topics = [\"How to Grow in Your Career\",2 \"The Habits of Great Software Developers\",3 \"Ideas for a Relaxing Weekend\"]\n```\n\nExample:\n```text\n1paragraphs = []23for topic in topics:4 paragraphs.append(generate_text(topic))56for topic,para in zip(topics,paragraphs):7 print(f\"Topic: {topic}\")8 print(f\"First Paragraph: {para}\")9 print(\"-\"*10)\n```\n\nExample:\n```text\nTopic: How to Grow in Your CareerFirst Paragraph: Advancing in your career can seem like a daunting task, especially if you're unsure of the path ahead. In this ever-changing professional landscape, there are numerous factors to consider. This blog aims to shed light on the strategies and skills that can help you navigate the complexities of career progression and unlock your full potential. Whether you're looking to secure a promotion or explore new opportunities, these insights will help you chart a course for your future. Let's embark on this journey of self-improvement and professional growth, equipping you with the tools to succeed in your career aspirations.----------Topic: The Habits of Great Software DevelopersFirst Paragraph: Great software developers are renowned for their ability to write robust code and create innovative applications, but what sets them apart from their peers? In this blog, we'll delve into the daily habits that contribute to their success. From their approach to coding challenges to the ways they stay organized, we'll explore the routines and practices that help them excel in the fast-paced world of software development. Understanding these habits can help you elevate your own skills and join the ranks of these industry leaders.----------Topic: Ideas for a Relaxing WeekendFirst Paragraph: Life can be stressful, and sometimes we just need a relaxing weekend to unwind and recharge. In this fast-paced world, taking some time to slow down and rejuvenate is essential. This blog post is here to help you plan the perfect low-key weekend with some easy and accessible ideas. From cozy indoor activities to peaceful outdoor adventures, I'll share some ideas to help you renew your mind, body, and spirit. Whether you're a homebody or an adventure seeker, there's something special for everyone. So, grab a cup of tea, sit back, and get ready to dive into a calming weekend of self-care and relaxation!----------\n```\n\nExample:\n```text\n1from cohere import ClassifyExample23examples = [4 ClassifyExample(text=\"I’m so proud of you\", label=\"positive\"),5 ClassifyExample(text=\"What a great time to be alive\", label=\"positive\"),6 ClassifyExample(text=\"That’s awesome work\", label=\"positive\"),7 ClassifyExample(text=\"The service was amazing\", label=\"positive\"),8 ClassifyExample(text=\"I love my family\", label=\"positive\"),9 ClassifyExample(text=\"They don't care about me\", label=\"negative\"),10 ClassifyExample(text=\"I hate this place\", label=\"negative\"),11 ClassifyExample(text=\"The most ridiculous thing I've ever heard\", label=\"negative\"),12 ClassifyExample(text=\"I am really frustrated\", label=\"negative\"),13 ClassifyExample(text=\"This is so unfair\", label=\"negative\"),14 ClassifyExample(text=\"This made me think\", label=\"neutral\"),15 ClassifyExample(text=\"The good old days\", label=\"neutral\"),16 ClassifyExample(text=\"What's the difference\", label=\"neutral\"),17 ClassifyExample(text=\"You can't ignore this\", label=\"neutral\"),18 ClassifyExample(text=\"That's how I see it\", label=\"neutral\")19]\n```\n\nExample:\n```text\n1inputs=[\"Hello, world! What a beautiful day\",2 \"It was a great time with great people\",3 \"Great place to work\",4 \"That was a wonderful evening\",5 \"Maybe this is why\",6 \"Let's start again\",7 \"That's how I see it\",8 \"These are all facts\",9 \"This is the worst thing\",10 \"I cannot stand this any longer\",11 \"This is really annoying\",12 \"I am just plain fed up\"13 ]\n```\n\nExample:\n```text\n1def classify_text(inputs, examples):2 \"\"\"3 Classify a list of input texts4 Arguments:5 inputs(list[str]): a list of input texts to be classified6 examples(list[Example]): a list of example texts and class labels7 Returns:8 classifications(list): each result contains the text, labels, and conf values9 \"\"\"10 # Classify text by calling the Classify endpoint11 response = co.classify(12 model='embed-v4.0',13 inputs=inputs,14 examples=examples)1516 classifications = response.classifications1718 return classifications\n```\n\nExample:\n```text\n1predictions = classify_text(inputs,examples)23classes = [\"positive\",\"negative\",\"neutral\"]4for inp,pred in zip(inputs,predictions):5 class_pred = pred.predictions[0]6 class_idx = classes.index(class_pred)7 class_conf = pred.confidences[0]89 print(f\"Input: {inp}\")10 print(f\"Prediction: {class_pred}\")11 print(f\"Confidence: {class_conf:.2f}\")12 print(\"-\"*10)\n```\n\nExample:\n```text\nInput: Hello, world! What a beautiful dayPrediction: positiveConfidence: 0.84----------Input: It was a great time with great peoplePrediction: positiveConfidence: 0.99----------Input: Great place to workPrediction: positiveConfidence: 0.91----------Input: That was a wonderful eveningPrediction: positiveConfidence: 0.96----------Input: Maybe this is whyPrediction: neutralConfidence: 0.70----------Input: Let's start againPrediction: neutralConfidence: 0.83----------Input: That's how I see itPrediction: neutralConfidence: 1.00----------Input: These are all factsPrediction: neutralConfidence: 0.78----------Input: This is the worst thingPrediction: negativeConfidence: 0.93----------Input: I cannot stand this any longerPrediction: negativeConfidence: 0.93----------Input: This is really annoyingPrediction: negativeConfidence: 0.99----------Input: I am just plain fed upPrediction: negativeConfidence: 1.00----------\n```\n\nExample:\n```text\n1df = pd.read_csv(\"https://github.com/cohere-ai/cohere-developer-experience/raw/main/notebooks/data/hello-world-kw.csv\", names=[\"search_term\"])2df.head()\n```\n\nExample:\n```text\n1def embed_text(texts, input_type):2 \"\"\"3 Turns a piece of text into embeddings4 Arguments:5 text(str): the text to be turned into embeddings6 Returns:7 embedding(list): the embeddings8 \"\"\"9 # Embed text by calling the Embed endpoint10 response = co.embed(11 model=\"embed-v4.0\",12 input_type=input_type,13 texts=texts)1415 return response.embeddings\n```\n\nExample:\n```text\n1df[\"search_term_embeds\"] = embed_text(texts=df[\"search_term\"].tolist(),2 input_type=\"search_document\")3doc_embeds = np.array(df[\"search_term_embeds\"].tolist())\n```\n\nExample:\n```text\n1query = \"what is the history of hello world\"23query_embeds = embed_text(texts=[query],4 input_type=\"search_query\")[0]\n```\n\nExample:\n```text\n1from sklearn.metrics.pairwise import cosine_similarity23def get_similarity(target, candidates):4 \"\"\"5 Computes the similarity between a target text and a list of other texts6 Arguments:7 target(list[float]): the target text8 candidates(list[list[float]]): a list of other texts, or candidates9 Returns:10 sim(list[tuple]): candidate IDs and the similarity scores11 \"\"\"12 # Turn list into array13 candidates = np.array(candidates)14 target = np.expand_dims(np.array(target),axis=0)1516 # Calculate cosine similarity17 sim = cosine_similarity(target,candidates)18 sim = np.squeeze(sim).tolist()1920 # Sort by descending order in similarity21 sim = list(enumerate(sim))22 sim = sorted(sim, key=lambda x:x[1], reverse=True)2324 # Return similarity scores25 return sim\n```\n\nExample:\n```text\n1similarity = get_similarity(query_embeds,doc_embeds)23print(\"New query:\")4print(new_query,'\\n')56print(\"Similar queries:\")7for idx,score in similarity[:5]:8 print(f\"Similarity: {score:.2f};\", df.iloc[idx][\"search_term\"])\n```\n\nExample:\n```text\nNew query:what is the history of hello worldSimilar queries:Similarity: 0.58; how did hello world originateSimilarity: 0.56; where did hello world come fromSimilarity: 0.54; why hello worldSimilarity: 0.53; why is hello world so famousSimilarity: 0.53; what is hello world\n```\n\nExample:\n```text\n1import umap2reducer = umap.UMAP(n_neighbors=49)3umap_embeds = reducer.fit_transform(doc_embeds)45df['x'] = umap_embeds[:,0]6df['y'] = umap_embeds[:,1]\n```\n\nExample:\n```text\n1chart = alt.Chart(df).mark_circle(size=500).encode(2 x=3 alt.X('x',4 scale=alt.Scale(zero=False),5 axis=alt.Axis(labels=False, ticks=False, domain=False)6 ),78 y=9 alt.Y('y',10 scale=alt.Scale(zero=False),11 axis=alt.Axis(labels=False, ticks=False, domain=False)12 ),1314 tooltip=['search_term']15 )1617text = chart.mark_text(align='left', dx=15, size=12, color='black'18 ).encode(text='search_term', color= alt.value('black'))1920result = (chart + text).configure(background=\"#FDF7F0\"21 ).properties(22 width=1000,23 height=700,24 title=\"2D Embeddings\"25 )2627result\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.369Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":24,"totalLines":123,"estimatedTokens":3479}}195{"id":"doc-migrating_monolithic_prompts_to_command_a_with_r-31890e9a","source":"documentation","title":"Migrating Monolithic Prompts to Command A with RAG | Cohere","url":"https://docs.cohere.com/page/migrating-prompts","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1#!pip install cohere\n```\n\nExample:\n```text\n1import json2import os3import re45import cohere6import getpass\n```\n\nExample:\n```text\n1CO_API_KEY = getpass.getpass('cohere API key:')\n```\n\nExample:\n```text\ncohere API key:··········\n```\n\nExample:\n```text\n1co = cohere.Client(CO_API_KEY)\n```\n\nExample:\n```text\n1original_prompt = '''## information2Current Job Title: Senior Software Engineer3Current Company Name: GlobalSolTech4Work Experience: Over 15 years of experience in software engineering, specializing in AI and machine learning. Proficient in Python, C++, and Java, with expertise in developing algorithms for natural language processing, computer vision, and recommendation systems.5Current Department Name: AI Research and Development6Education: B.Sc. in Physics from Trent University (2004), Ph.D. in Statistics from HEC in Paris (2010)7Hobbies: I love hiking in the mountains, free diving, and collecting and restoring vintage world war one mechanical watches.8Family: Married with 4 children and 3 grandchildren.910## instructions11Your task is to assist a user in writing a short biography for social media.12The length of the text should be no more than 100 words.13Write the summary in first person.'''\n```\n\nExample:\n```text\n1response = co.chat(2 message=original_prompt,3 model='command-a-03-2025',4)\n```\n\nExample:\n```text\n1print(response.text)\n```\n\nExample:\n```text\nI'm a Senior Software Engineer at GlobalSolTech, with over 15 years of experience in AI and machine learning. My expertise lies in developing innovative algorithms for natural language processing, computer vision, and recommendation systems. I hold a B.Sc. in Physics and a Ph.D. in Statistics and enjoy hiking, free diving, and collecting vintage watches in my spare time. I'm passionate about using my skills to contribute to cutting-edge AI research and development. At GlobalSolTech, I'm proud to be part of a dynamic team driving technological advancement.\n```\n\nExample:\n```text\n1meta_prompt = f'''Below is a task for an LLM delimited with ## Original Task. Your task is to split that task into two parts: (1) the context; and (2) the instructions.2The context should be split into several separate parts and returned as a JSON object where each part has a name describing its contents and the value is the contents itself.3Make sure to include all of the context contained in the original task description and do not change its meaning.4The instructions should be re-written so that they are very clear and concise. Do not change the meaning of the instructions or task, just make sure they are very direct and clear.5Return everything in a JSON object with the following structure:67{{8 \"context\": [{{\"<description 1=\"\" of=\"\" part=\"\">\": \"<content 1=\"\" of=\"\" part=\"\">\"}}, ...],9 \"instructions\": \"<the instructions=\"\" re-written=\"\">\"10}}1112## Original Task13{original_prompt}14'''\n```\n\nExample:\n```text\n1print(meta_prompt)\n```\n\nExample:\n```text\nBelow is a task for an LLM delimited with ## Original Task. Your task is to split that task into two parts: (1) the context; and (2) the instructions.The context should be split into several separate parts and returned as a JSON object where each part has a name describing its contents and the value is the contents itself.Make sure to include all of the context contained in the original task description and do not change its meaning.The instructions should be re-written so that they are very clear and concise. Do not change the meaning of the instructions or task, just make sure they are very direct and clear.Return everything in a JSON object with the following structure:{ \"context\": [{\"<description 1=\"\" of=\"\" part=\"\">\": \"<content 1=\"\" of=\"\" part=\"\">\"}, ...], \"instructions\": \"<the instructions=\"\" re-written=\"\">\"}## Original Task## informationCurrent Job Title: Senior Software EngineerCurrent Company Name: GlobalSolTechWork Experience: Over 15 years of experience in software engineering, specializing in AI and machine learning. Proficient in Python, C++, and Java, with expertise in developing algorithms for natural language processing, computer vision, and recommendation systems.Current Department Name: AI Research and DevelopmentEducation: B.Sc. in Physics from Trent University (2004), Ph.D. in Statistics from HEC in Paris (2010)Hobbies: I love hiking in the mountains, free diving, and collecting and restoring vintage world war one mechanical watches.Family: Married with 4 children and 3 grandchildren.## instructionsYour task is to assist a user in writing a short biography for social media.The length of the text should be no more than 100 words.Write the summary in first person.\n```\n\nExample:\n```text\n1upgraded_prompt = co.chat(2 message=meta_prompt,3 model='command-a-03-2025',4)\n```\n\nExample:\n```text\n1print(upgraded_prompt.text)\n```\n\nExample:\n```text\nHere is the task delved into a JSON object as requested:```json JSON{ \"context\": [ { \"Work Experience\": \"Over 15 years of AI and machine learning engineering experience. Proficient in Python, C++, and Java, with expertise in developing algorithms for natural language processing, computer vision, and recommendation systems.\" }, { \"Education\": \"B.Sc. in Physics (Trent University, 2004) and Ph.D. in Statistics (HEC Paris, 2010).\" }, { \"Personal Life\": \"I’m a married senior software engineer with 4 children and 3 grandchildren. I enjoy hiking, free diving, and vintage watch restoration.\" }, { \"Current Position\": \"I work at GlobalSolTech in the AI Research and Development department as a senior software engineer.\" } ], \"instructions\": \"Using the provided information, write a concise, first-person social media biography of no more than 100 words.\"}```\n```\n\nExample:\n```text\n1def get_json(text: str) -> str:2 matches = [m.group(1) for m in re.finditer(\"```([\\w\\W]*?)```\", text)]3 if len(matches):4 postproced = matches[0]5 if postproced[:4] == 'json':6 return postproced[4:]7 return postproced8 return text\n```\n\nExample:\n```text\n1def get_prompt_and_docs(text: str) -> tuple:2 json_obj = json.loads(get_json(text))3 prompt = json_obj['instructions']4 docs = []5 for item in json_obj['context']:6 for k,v in item.items():7 docs.append({\"title\": k, \"snippet\": v})8 return prompt, docs\n```\n\nExample:\n```text\n1new_prompt, docs = get_prompt_and_docs(upgraded_prompt.text)\n```\n\nExample:\n```text\n1new_prompt, docs\n```\n\nExample:\n```text\n('Using the provided information, write a concise, first-person social media biography of no more than 100 words.', [{'title': 'Work Experience', 'snippet': 'Over 15 years of AI and machine learning engineering experience. Proficient in Python, C++, and Java, with expertise in developing algorithms for natural language processing, computer vision, and recommendation systems.'}, {'title': 'Education', 'snippet': 'B.Sc. in Physics (Trent University, 2004) and Ph.D. in Statistics (HEC Paris, 2010).'}, {'title': 'Personal Life', 'snippet': 'I’m a married senior software engineer with 4 children and 3 grandchildren. I enjoy hiking, free diving, and vintage watch restoration.'}, {'title': 'Current Position', 'snippet': 'I work at GlobalSolTech in the AI Research and Development department as a senior software engineer.'}])\n```\n\nExample:\n```text\n1response = co.chat(2 message=new_prompt,3 model='command-a-03-2025',4 documents=docs,5)\n```\n\nExample:\n```text\nI'm a senior software engineer with a Ph.D. in Statistics and over 15 years of AI and machine learning engineering experience. My current focus at GlobalSolTech's AI R&D department is developing algorithms for natural language processing, computer vision, and recommendation systems. In my free time, I enjoy hiking, freediving, and restoring vintage watches, and I'm a married father of four with three grandchildren.\n```\n\nExample:\n```text\n1def insert_citations(text: str, citations: list[dict], add_one: bool=False):2 \"\"\"3 A helper function to pretty print citations.4 \"\"\"5 offset = 06 # Process citations in the order they were provided7 for citation in citations:8 # Adjust start/end with offset9 start, end = citation.start + offset, citation.end + offset10 if add_one:11 cited_docs = [str(int(doc[4:]) + 1) for doc in citation.document_ids]12 else:13 cited_docs = [doc[4:] for doc in citation.document_ids]14 # Shorten citations if they're too long for convenience15 if len(cited_docs) > 3:16 placeholder = \"[\" + \", \".join(cited_docs[:3]) + \"...]\"17 else:18 placeholder = \"[\" + \", \".join(cited_docs) + \"]\"19 # ^ doc[4:] removes the 'doc_' prefix, and leaves the quoted document20 modification = f'{text[start:end]} {placeholder}'21 # Replace the cited text with its bolded version + placeholder22 text = text[:start] + modification + text[end:]23 # Update the offset for subsequent replacements24 offset += len(modification) - (end - start)2526 return text\n```\n\nExample:\n```text\n1print(insert_citations(response.text, response.citations, True))\n```\n\nExample:\n```text\nI'm a senior software engineer [3, 4] with a Ph.D. in Statistics [2] and over 15 years of AI and machine learning engineering experience. [1] My current focus at GlobalSolTech's AI R&D department [4] is developing algorithms for natural language processing, computer vision, and recommendation systems. [1] In my free time, I enjoy hiking, freediving, and restoring vintage watches [3], and I'm a married father of four with three grandchildren. [3]\n```\n\nExample:\n```text\n1apple = open('data/apple_mod.txt').read()\n```\n\nExample:\n```text\n1tokens = co.tokenize(text=apple, model='command-a-03-2025')2len(tokens.tokens)\n```\n\nExample:\n```text\n29697\n```\n\nExample:\n```text\n1prompt_template = '''2{legal_text}34{question}5'''\n```\n\nExample:\n```text\n1question = '''Please summarize the attached legal complaint succinctly. Focus on answering the question: what does the complaint allege?'''2rendered_prompt = prompt_template.format(legal_text=apple, question=question)\n```\n\nExample:\n```text\n1response = co.chat(2 message=rendered_prompt,3 model='command-a-03-2025',4 temperature=0.3,5)\n```\n\nExample:\n```text\nThe complaint alleges that Apple has violated antitrust laws by engaging in a pattern of anticompetitive conduct to maintain its monopoly power over the U.S. markets for smartphones and performance smartphones. Apple is accused of using its control over app distribution and access to its operating system to impede competition and innovation. Specifically, the company is said to have restricted developers' ability to create certain apps and limited the functionality of others, making it harder for consumers to switch away from iPhones to rival smartphones. This conduct is alleged to have harmed consumers and developers by reducing choice, increasing prices, and stifling innovation. The plaintiffs seek injunctive relief and potential monetary awards to remedy these illegal practices.\n```\n\nExample:\n```text\n1question = '''Does the DOJ allege that Apple could encrypt text messages sent to Android phones?'''2rendered_prompt = prompt_template.format(legal_text=apple, question=question)\n```\n\nExample:\n```text\n1response = co.chat(2 message=rendered_prompt,3 model='command-a-03-2025',4)\n```\n\nExample:\n```text\nYes, the DOJ alleges that Apple could allow iPhone users to send encrypted messages to Android users while still using iMessage on their iPhones but chooses not to do so. According to the DOJ, this would instantly improve the privacy and security of iPhones and other smartphones.\n```\n\nExample:\n```text\n1def chunk_doc(input_doc: str) -> list:2 chunks = []3 current_para = 'Preamble'4 current_chunk = ''5 # pattern to find an integer number followed by a dot (finding the explicitly numbered paragraph numbers)6 pattern = r'^\\d+\\.$'78 for line in input_doc.splitlines():9 if re.match(pattern, line):10 chunks.append((current_para.replace('.', ''), current_chunk))11 current_chunk = ''12 current_para = line13 else:14 current_chunk += line + '\\n'1516 docs = []17 for chunk in chunks:18 docs.append({\"title\": chunk[0], \"snippet\": chunk[1]})1920 return docs\n```\n\nExample:\n```text\n1chunks = chunk_doc(apple)\n```\n\nExample:\n```text\n1print(chunks[18])\n```\n\nExample:\n```text\n1 {'title': '18', 'snippet': '\\nProtecting competition and the innovation that competition inevitably ushers in\\nfor consumers, developers, publishers, content creators, and device manufacturers is why\\nPlaintiffs bring this lawsuit under Section 2 of the Sherman Act to challenge Apple’s\\nmaintenance of its monopoly over smartphone markets, which affect hundreds of millions of\\nAmericans every day. Plaintiffs bring this case to rid smartphone markets of Apple’s\\nmonopolization and exclusionary conduct and to ensure that the next generation of innovators\\ncan upend the technological world as we know it with new and transformative technologies.\\n\\n\\nII.\\n\\nDefendant Apple\\n\\n'}\n```\n\nExample:\n```text\n1response = co.chat(2 message='''Does the DOJ allege that Apple could encrypt text messages sent to Android phones?''',3 model='command-a-03-2025',4 documents=chunks,5)\n```\n\nExample:\n```text\nYes, according to the DOJ, Apple could encrypt text messages sent from iPhones to Android phones. The DOJ claims that Apple degrades the security and privacy of its users by impeding cross-platform encryption and preventing developers from fixing the broken cross-platform messaging experience. Apple's conduct makes it harder to switch from iPhone to Android, as messages sent from iPhones to Android phones are unencrypted.\n```\n\nExample:\n```text\n1print(insert_citations(response.text, response.citations))\n```\n\nExample:\n```text\nYes, according to the DOJ, Apple could encrypt text messages sent from iPhones to Android phones. [144] The DOJ claims that Apple degrades the security and privacy [144] of its users by impeding cross-platform encryption [144] and preventing developers from fixing the broken cross-platform messaging experience. [93] Apple's conduct makes it harder to switch from iPhone to Android [144], as messages sent from iPhones to Android phones are unencrypted. [144]\n```\n\nExample:\n```text\n1print(chunks[144]['snippet'])\n```\n\nExample:\n```text\nApple is also willing to make the iPhone less secure and less private if that helpsmaintain its monopoly power. For example, text messages sent from iPhones to Android phonesare unencrypted as a result of Apple’s conduct. If Apple wanted to, Apple could allow iPhoneusers to send encrypted messages to Android users while still using iMessage on their iPhone,which would instantly improve the privacy and security of iPhone and other smartphone users.\n```\n\nExample:\n```text\n1print(chunks[93]['snippet'])\n```\n\nExample:\n```text\nRecently, Apple blocked a third-party developer from fixing the broken cross-platform messaging experience in Apple Messages and providing end-to-end encryption formessages between Apple Messages and Android users. By rejecting solutions that would allowfor cross-platform encryption, Apple continues to make iPhone users’ less secure than they couldotherwise be.ii.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.372Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":47,"totalLines":238,"estimatedTokens":3899}}196{"id":"doc-financial_csv_agent_with_native_multi_step_coher-a5c8a75e","source":"documentation","title":"Financial CSV Agent with Native Multi-Step Cohere API | Cohere","url":"https://docs.cohere.com/page/csv-agent-native-api","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import os2from typing import List34import cohere5import langchain6import langchain_core7import langchain_experimental8import pandas as pd9from langchain.agents import Tool10from langchain_core.pydantic_v1 import BaseModel, Field11from langchain_experimental.utilities import PythonREPL\n```\n\nExample:\n```text\n1# Uncomment if you need to install the following packages2# !pip install --quiet langchain langchain_experimental cohere --upgrade\n```\n\nExample:\n```text\n1# versions2print('cohere version:', cohere.__version__)3print('langchain version:', langchain.__version__)4print('langchain_core version:', langchain_core.__version__)5print('langchain_experimental version:', langchain_experimental.__version__)\n```\n\nExample:\n```text\ncohere version: 5.5.1langchain version: 0.2.0langchain_core version: 0.2.0langchain_experimental version: 0.0.59\n```\n\nExample:\n```text\n1COHERE_API_KEY = os.environ[\"COHERE_API_KEY\"]2CHAT_URL= \"https://api.cohere.ai/v1/chat\"3COHERE_MODEL = 'command-a-03-2025'4co = cohere.Client(api_key=COHERE_API_KEY)\n```\n\nExample:\n```text\n1income_statement = pd.read_csv('income_statement.csv')2balance_sheet = pd.read_csv('balance_sheet.csv')\n```\n\nExample:\n```text\n1income_statement.head(2)\n```\n\nExample:\n```text\n1balance_sheet.head(2)\n```\n\nExample:\n```text\n1python_repl = PythonREPL()2python_tool = Tool(3 name=\"python_repl\",4 description=\"Executes python code and returns the result. The code runs in a static sandbox without interactive mode, so print output or save output to a file.\",5 func=python_repl.run,6)7python_tool.name = \"python_interpreter\"89class ToolInput(BaseModel):10 code: str = Field(description=\"Python code to execute.\")11python_tool.args_schema = ToolInput1213def run_python_code(code: str) -> dict:14 \"\"\"15 Function to run given python code16 \"\"\"17 input_code = ToolInput(code=code)18 return {'python_answer': python_tool.func(input_code.code)}1920functions_map = {21 \"run_python_code\": run_python_code,22}2324tools = [25 {26 \"name\": \"run_python_code\",27 \"description\": \"given a python code, runs it\",28 \"parameter_definitions\": {29 \"code\": {30 \"description\": \"executable python code\",31 \"type\": \"str\",32 \"required\": True33 }34 }35 },]\n```\n\nExample:\n```text\n1def cohere_agent(2 message: str,3 preamble: str,4 tools: List[dict],5 force_single_step=False,6 verbose: bool = False,7) -> str:8 \"\"\"9 Function to handle multi-step tool use api.1011 Args:12 message (str): The message to send to the Cohere AI model.13 preamble (str): The preamble or context for the conversation.14 tools (list of dict): List of tools to use in the conversation.15 verbose (bool, optional): Whether to print verbose output. Defaults to False.1617 Returns:18 str: The final response from the call.19 \"\"\"2021 counter = 12223 response = co.chat(24 model=COHERE_MODEL,25 message=message,26 preamble=preamble,27 tools=tools,28 force_single_step=force_single_step,29 )3031 if verbose:32 print(f\"\\nrunning 0th step.\")33 print(response.text)3435 while response.tool_calls:36 tool_results = []3738 if verbose:39 print(f\"\\nrunning {counter}th step.\")4041 for tool_call in response.tool_calls:42 output = functions_map[tool_call.name](**tool_call.parameters)43 outputs = [output]44 tool_results.append({\"call\": tool_call, \"outputs\": outputs})4546 if verbose:47 print(48 f\"= running tool {tool_call.name}, with parameters: {tool_call.parameters}\"49 )50 print(f\"== tool results: {outputs}\")5152 response = co.chat(53 model=COHERE_MODEL,54 message=\"\",55 chat_history=response.chat_history,56 preamble=preamble,57 tools=tools,58 force_single_step=force_single_step,59 tool_results=tool_results,60 )6162 if verbose:63 print(response.text)6465 counter += 16667 return response.text686970# test71output = cohere_agent(\"can you use python to answer 1 + 1\", None, tools, verbose=True)\n```\n\nExample:\n```text\nrunning 0th step.I will use Python to answer this question.running 1th step.= running tool run_python_code, with parameters: {'code': 'print(1 + 1)'}== tool results: [{'python_answer': '2\\n'}]The answer is **2**.\n```\n\nExample:\n```text\n1question_dict ={2 'q1': ['what is the highest value of cost of goods and service?',169559000000],3 'q2': ['what is the largest gross profit margin?',0.3836194330595236],4 'q3': ['what is the minimum ratio of operating income loss divided by non operating income expense?',35.360599]5}\n```\n\nExample:\n```text\n1preamble = \"\"\"2You are an expert who answers the user's question. You are working with a pandas dataframe in Python. The name of the dataframe is `income_statement.csv`.3Here is a preview of the dataframe:4{head_df}5\"\"\".format(head_df=income_statement.head(3).to_markdown())67print(preamble)\n```\n\nExample:\n```text\n1for qsn,val in question_dict.items():2 print(f'question:{qsn}')3 question = val[0]4 answer = val[1]5 output = cohere_agent(question, preamble, tools, verbose=True)6 print(f'GT Answer:{val[1]}')7 print('-'*50)\n```\n\nExample:\n```text\nquestion:q1running 0th step.I will use Python to find the highest value of 'CostOfGoodsAndServicesSold' in the 'income_statement.csv' dataframe.running 1th step.= running tool run_python_code, with parameters: {'code': 'import pandas as pd\\n\\ndf = pd.read_csv(\\'income_statement.csv\\')\\n\\n# Find the highest value of \\'CostOfGoodsAndServicesSold\\'\\nhighest_cost = df[\\'CostOfGoodsAndServicesSold\\'].max()\\n\\nprint(f\"The highest value of \\'CostOfGoodsAndServicesSold\\' is {highest_cost}\")'}== tool results: [{'python_answer': \"The highest value of 'CostOfGoodsAndServicesSold' is 169559000000.0\\n\"}]The highest value of 'CostOfGoodsAndServicesSold' is 169559000000.0.GT Answer:169559000000--------------------------------------------------question:q2running 0th step.I will write and execute Python code to find the largest gross profit margin.running 1th step.= running tool run_python_code, with parameters: {'code': 'import pandas as pd\\n\\ndf = pd.read_csv(\\'income_statement.csv\\')\\n\\n# Calculate gross profit margin\\ndf[\\'GrossProfitMargin\\'] = df[\\'GrossProfit\\'] / df[\\'RevenueFromContractWithCustomerExcludingAssessedTax\\'] * 100\\n\\n# Find the largest gross profit margin\\nlargest_gross_profit_margin = df[\\'GrossProfitMargin\\'].max()\\n\\nprint(f\"The largest gross profit margin is {largest_gross_profit_margin:.2f}%\")'}== tool results: [{'python_answer': 'The largest gross profit margin is 38.36%\\n'}]The largest gross profit margin is 38.36%.GT Answer:0.3836194330595236--------------------------------------------------question:q3running 0th step.I will use Python to find the minimum ratio of operating income loss divided by non-operating income expense.running 1th step.= running tool run_python_code, with parameters: {'code': 'import pandas as pd\\n\\ndf = pd.read_csv(\"income_statement.csv\")\\n\\n# Calculate the ratio of operating income loss to non-operating income expense\\ndf[\"OperatingIncomeLossRatio\"] = df[\"OperatingIncomeLoss\"] / df[\"NonoperatingIncomeExpense\"]\\n\\n# Find the minimum ratio\\nmin_ratio = df[\"OperatingIncomeLossRatio\"].min()\\n\\nprint(f\"The minimum ratio of operating income loss to non-operating income expense is: {min_ratio:.2f}\")'}== tool results: [{'python_answer': 'The minimum ratio of operating income loss to non-operating income expense is: 35.36\\n'}]The minimum ratio of operating income loss to non-operating income expense is 35.36.GT Answer:35.360599--------------------------------------------------\n```\n\nExample:\n```text\n1question_dict ={2 'q1': ['what is the ratio of the largest stockholders equity to the smallest revenue'],3}\n```\n\nExample:\n```text\n1# get the largest stockholders equity2x = balance_sheet['StockholdersEquity'].astype(float).max()3print(f\"The largest stockholders equity value is: {x}\")45# get the smallest revenue6y = income_statement['RevenueFromContractWithCustomerExcludingAssessedTax'].astype(float).min()7print(f\"The smallest revenue value is: {y}\")89# compute the ratio10ratio = x/y11print(f\"Their ratio is: {ratio}\")\n```\n\nExample:\n```text\nThe largest stockholders equity value is: 134047000000.0The smallest revenue value is: 53809000000.0Their ratio is: 2.4911631883142227\n```\n\nExample:\n```text\n1preamble = \"\"\"2You are an expert who answers the user's question in complete sentences. You are working with two pandas dataframe in Python. Ensure your output is a string.34Here is a preview of the `income_statement.csv` dataframe:5{table_1}67Here is a preview of the `balance_sheet.csv` dataframe:8{table_2}9\"\"\".format(table_1=income_statement.head(3).to_markdown(),table_2=balance_sheet.head(3).to_markdown())101112print(preamble)\n```\n\nExample:\n```text\nYou are an expert who answers the user's question in complete sentences. You are working with two pandas dataframe in Python. Ensure your output is a string.Here is a preview of the `income_statement.csv` dataframe:| | Unnamed: 0 | index | RevenueFromContractWithCustomerExcludingAssessedTax | CostOfGoodsAndServicesSold | GrossProfit | ResearchAndDevelopmentExpense | SellingGeneralAndAdministrativeExpense | OperatingExpenses | OperatingIncomeLoss | NonoperatingIncomeExpense | IncomeLossFromContinuingOperationsBeforeIncomeTaxesExtraordinaryItemsNoncontrollingInterest | IncomeTaxExpenseBenefit | NetIncomeLoss | EarningsPerShareBasic | EarningsPerShareDiluted | WeightedAverageNumberOfSharesOutstandingBasic | WeightedAverageNumberOfDilutedSharesOutstanding ||---:|-------------:|:----------------------|------------------------------------------------------:|-----------------------------:|--------------:|--------------------------------:|-----------------------------------------:|--------------------:|----------------------:|----------------------------:|----------------------------------------------------------------------------------------------:|--------------------------:|----------------:|------------------------:|--------------------------:|------------------------------------------------:|--------------------------------------------------:|| 0 | 0 | 2017-10-01-2018-09-29 | 265595000000 | 1.63756e+11 | 101839000000 | 1.4236e+10 | 1.6705e+10 | 3.0941e+10 | 7.0898e+10 | 2.005e+09 | 7.2903e+10 | 1.3372e+10 | 59531000000 | 3 | 2.98 | 1.98215e+10 | 2.00004e+10 || 1 | 1 | 2018-09-30-2018-12-29 | 84310000000 | nan | 32031000000 | nan | nan | nan | nan | nan | nan | nan | 19965000000 | 1.05 | 1.05 | nan | nan || 2 | 2 | 2018-09-30-2019-09-28 | 260174000000 | 1.61782e+11 | 98392000000 | 1.6217e+10 | 1.8245e+10 | 3.4462e+10 | 6.393e+10 | 1.807e+09 | 6.5737e+10 | 1.0481e+10 | 55256000000 | 2.99 | 2.97 | 1.84713e+10 | 1.85957e+10 |Here is a preview of the `balance_sheet.csv` dataframe:| | Unnamed: 0 | index | CashAndCashEquivalentsAtCarryingValue | MarketableSecuritiesCurrent | AccountsReceivableNetCurrent | InventoryNet | NontradeReceivablesCurrent | OtherAssetsCurrent | AssetsCurrent | MarketableSecuritiesNoncurrent | PropertyPlantAndEquipmentNet | OtherAssetsNoncurrent | AssetsNoncurrent | Assets | AccountsPayableCurrent | OtherLiabilitiesCurrent | ContractWithCustomerLiabilityCurrent | CommercialPaper | LongTermDebtCurrent | LiabilitiesCurrent | LongTermDebtNoncurrent | OtherLiabilitiesNoncurrent | LiabilitiesNoncurrent | Liabilities | CommitmentsAndContingencies | CommonStocksIncludingAdditionalPaidInCapital | RetainedEarningsAccumulatedDeficit | AccumulatedOtherComprehensiveIncomeLossNetOfTax | StockholdersEquity | LiabilitiesAndStockholdersEquity ||---:|-------------:|:-----------|----------------------------------------:|------------------------------:|-------------------------------:|---------------:|-----------------------------:|---------------------:|----------------:|---------------------------------:|-------------------------------:|------------------------:|-------------------:|--------------:|-------------------------:|--------------------------:|---------------------------------------:|------------------:|----------------------:|---------------------:|-------------------------:|-----------------------------:|------------------------:|--------------:|------------------------------:|-----------------------------------------------:|-------------------------------------:|--------------------------------------------------:|---------------------:|-----------------------------------:|| 0 | 0 | 2017-09-30 | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | 134047000000 | nan || 1 | 1 | 2018-09-29 | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | nan | 107147000000 | nan || 2 | 2 | 2019-09-28 | 4.8844e+10 | 5.1713e+10 | 2.2926e+10 | 4.106e+09 | 2.2878e+10 | 1.2352e+10 | 1.62819e+11 | 1.05341e+11 | 3.7378e+10 | 3.2978e+10 | 1.75697e+11 | 3.38516e+11 | 4.6236e+10 | 3.772e+10 | 5.522e+09 | 5.98e+09 | 1.026e+10 | 1.05718e+11 | 9.1807e+10 | 5.0503e+10 | 1.4231e+11 | 2.48028e+11 | 0 | 4.5174e+10 | 4.5898e+10 | -5.84e+08 | 90488000000 | 3.38516e+11 |\n```\n\nExample:\n```text\n1for qsn,val in question_dict.items():2 print(f'question:{qsn}')3 question = val[0]4 output = cohere_agent(question, preamble, tools, verbose=True)\n```\n\nExample:\n```text\nquestion:q1running 0th step.I will use the provided code to find the ratio of the largest stockholders equity to the smallest revenue.running 1th step.= running tool run_python_code, with parameters: {'code': 'import pandas as pd\\n\\n# Read the CSV files into dataframes\\nincome_statement = pd.read_csv(\\'income_statement.csv\\')\\nbalance_sheet = pd.read_csv(\\'balance_sheet.csv\\')\\n\\n# Find the smallest revenue\\nsmallest_revenue = income_statement[\\'RevenueFromContractWithCustomerExcludingAssessedTax\\'].min()\\n\\n# Find the largest stockholders equity\\nlargest_stockholders_equity = balance_sheet[\\'StockholdersEquity\\'].max()\\n\\n# Calculate the ratio\\nratio = largest_stockholders_equity / smallest_revenue\\nprint(f\"The ratio of the largest stockholders equity to the smallest revenue is {ratio:.2f}\")'}== tool results: [{'python_answer': 'The ratio of the largest stockholders equity to the smallest revenue is 2.49\\n'}]The ratio of the largest stockholders equity to the smallest revenue is 2.49.\n```\n\nExample:\n```text\n1preamble = \"\"\"2You are an expert who answers the user's question. You are working with a pandas dataframe in Python. The name of the dataframe is `income_statement.csv`.3\"\"\"45question1 = \"what is the highest value of cost of goods and service?\"67output = cohere_agent(question1, preamble, tools, verbose=True)\n```\n\nExample:\n```text\nrunning 0th step.I will use Python to find the highest value of 'Cost of Goods and Service' in the 'income_statement.csv' file.running 1th step.= running tool run_python_code, with parameters: {'code': 'import pandas as pd\\n\\ndf = pd.read_csv(\\'income_statement.csv\\')\\n\\n# Find the highest value of \\'Cost of Goods and Service\\'\\nhighest_cost = df[\\'Cost of Goods and Service\\'].max()\\n\\nprint(f\"The highest value of \\'Cost of Goods and Service\\' is: {highest_cost}\")'}== tool results: [{'python_answer': \"KeyError('Cost of Goods and Service')\"}]Sorry, there is no column named 'Cost of Goods and Service' in the 'income_statement.csv' file.\n```\n\nExample:\n```text\n1preamble = \"\"\"2You are an expert who answers the user's question. You are working with a pandas dataframe in Python. The name of the dataframe is `income_statement.csv`.3If you run into error, keep trying until you fix it. You may need to view the data to understand the error.4\"\"\"56question1 = \"what is the highest value of cost of goods and service?\"78output = cohere_agent(question1, preamble, tools, verbose=True)\n```\n\nExample:\n```text\nrunning 0th step.I will use Python to find the highest value of cost of goods and service.running 1th step.= running tool run_python_code, with parameters: {'code': 'import pandas as pd\\n\\ndf = pd.read_csv(\\'income_statement.csv\\')\\n\\n# Find the highest value of \\'Cost of Goods and Services\\'\\nhighest_cost = df[\\'Cost of Goods and Services\\'].max()\\n\\nprint(f\"The highest value of \\'Cost of Goods and Services\\' is: {highest_cost}\")'}== tool results: [{'python_answer': \"KeyError('Cost of Goods and Services')\"}]I encountered an error. I will now view the data to understand the error.running 2th step.= running tool run_python_code, with parameters: {'code': \"import pandas as pd\\n\\ndf = pd.read_csv('income_statement.csv')\\n\\nprint(df.columns)\"}== tool results: [{'python_answer': \"Index(['Unnamed: 0', 'index',\\n 'RevenueFromContractWithCustomerExcludingAssessedTax',\\n 'CostOfGoodsAndServicesSold', 'GrossProfit',\\n 'ResearchAndDevelopmentExpense',\\n 'SellingGeneralAndAdministrativeExpense', 'OperatingExpenses',\\n 'OperatingIncomeLoss', 'NonoperatingIncomeExpense',\\n 'IncomeLossFromContinuingOperationsBeforeIncomeTaxesExtraordinaryItemsNoncontrollingInterest',\\n 'IncomeTaxExpenseBenefit', 'NetIncomeLoss', 'EarningsPerShareBasic',\\n 'EarningsPerShareDiluted',\\n 'WeightedAverageNumberOfSharesOutstandingBasic',\\n 'WeightedAverageNumberOfDilutedSharesOutstanding'],\\n dtype='object')\\n\"}]I found that the column name is 'CostOfGoodsAndServicesSold'. I will now use this to find the highest value.running 3th step.= running tool run_python_code, with parameters: {'code': 'import pandas as pd\\n\\ndf = pd.read_csv(\\'income_statement.csv\\')\\n\\n# Find the highest value of \\'Cost of Goods and Services\\'\\nhighest_cost = df[\\'CostOfGoodsAndServicesSold\\'].max()\\n\\nprint(f\"The highest value of \\'Cost of Goods and Services\\' is: {highest_cost}\")'}== tool results: [{'python_answer': \"The highest value of 'Cost of Goods and Services' is: 169559000000.0\\n\"}]The highest value of 'Cost of Goods and Services' is 169559000000.0.\n```\n\nExample:\n```text\n1def view_csv_data(path: str) -> dict:2 \"\"\"3 Function to view the head, tail and shape of a given csv file.4 \"\"\"5 df = pd.read_csv(path)67 return {8 \"head\": df.head().to_string(),9 \"tail\": df.tail().to_string(),10 \"shape\": str(df.shape),11 }1213functions_map = {14 \"run_python_code\": run_python_code,15 \"view_csv_data\": view_csv_data16}1718tools = [19 {20 \"name\": \"run_python_code\",21 \"description\": \"given a python code, runs it\",22 \"parameter_definitions\": {23 \"code\": {24 \"description\": \"executable python code\",25 \"type\": \"str\",26 \"required\": True27 }28 }29 },30 {31 \"name\": \"view_csv_data\",32 \"description\": \"give path to csv data and get head, tail and shape of the data\",33 \"parameter_definitions\": {34 \"path\": {35 \"description\": \"path to csv\",36 \"type\": \"str\",37 \"required\": True38 }39 }40 },41]\n```\n\nExample:\n```text\n1preamble = \"\"\"2You are an expert who answers the user's question. You are working with a pandas dataframe in Python. The name of the dataframe is `income_statement.csv`.3Always view the data first to write flawless code.4\"\"\"56question1 = \"what is the highest value of cost of goods and service?\"78output = cohere_agent(question1, preamble, tools, verbose=True)\n```\n\nExample:\n```text\nrunning 0th step.I will first view the data and then write and execute Python code to find the highest value of cost of goods and service.running 1th step.= running tool view_csv_data, with parameters: {'path': 'income_statement.csv'}== tool results: [{'head': ' Unnamed: 0 index RevenueFromContractWithCustomerExcludingAssessedTax CostOfGoodsAndServicesSold GrossProfit ResearchAndDevelopmentExpense SellingGeneralAndAdministrativeExpense OperatingExpenses OperatingIncomeLoss NonoperatingIncomeExpense IncomeLossFromContinuingOperationsBeforeIncomeTaxesExtraordinaryItemsNoncontrollingInterest IncomeTaxExpenseBenefit NetIncomeLoss EarningsPerShareBasic EarningsPerShareDiluted WeightedAverageNumberOfSharesOutstandingBasic WeightedAverageNumberOfDilutedSharesOutstanding\\n0 0 2017-10-01-2018-09-29 265595000000 1.637560e+11 101839000000 1.423600e+10 1.670500e+10 3.094100e+10 7.089800e+10 2.005000e+09 7.290300e+10 1.337200e+10 59531000000 3.00 2.98 1.982151e+10 2.000044e+10\\n1 1 2018-09-30-2018-12-29 84310000000 NaN 32031000000 NaN NaN NaN NaN NaN NaN NaN 19965000000 1.05 1.05 NaN NaN\\n2 2 2018-09-30-2019-09-28 260174000000 1.617820e+11 98392000000 1.621700e+10 1.824500e+10 3.446200e+10 6.393000e+10 1.807000e+09 6.573700e+10 1.048100e+10 55256000000 2.99 2.97 1.847134e+10 1.859565e+10\\n3 3 2018-12-30-2019-03-30 58015000000 NaN 21821000000 NaN NaN NaN NaN NaN NaN NaN 11561000000 0.62 0.61 NaN NaN\\n4 4 2019-03-31-2019-06-29 53809000000 NaN 20227000000 NaN NaN NaN NaN NaN NaN NaN 10044000000 0.55 0.55 NaN NaN', 'tail': ' Unnamed: 0 index RevenueFromContractWithCustomerExcludingAssessedTax CostOfGoodsAndServicesSold GrossProfit ResearchAndDevelopmentExpense SellingGeneralAndAdministrativeExpense OperatingExpenses OperatingIncomeLoss NonoperatingIncomeExpense IncomeLossFromContinuingOperationsBeforeIncomeTaxesExtraordinaryItemsNoncontrollingInterest IncomeTaxExpenseBenefit NetIncomeLoss EarningsPerShareBasic EarningsPerShareDiluted WeightedAverageNumberOfSharesOutstandingBasic WeightedAverageNumberOfDilutedSharesOutstanding\\n6 6 2019-09-29-2019-12-28 91819000000 NaN 35217000000 NaN NaN NaN NaN NaN NaN NaN 22236000000 1.26 1.25 NaN NaN\\n7 7 2019-09-29-2020-09-26 274515000000 1.695590e+11 104956000000 1.875200e+10 1.991600e+10 3.866800e+10 6.628800e+10 803000000.0 6.709100e+10 9.680000e+09 57411000000 3.31 3.28 1.735212e+10 1.752821e+10\\n8 8 2019-12-29-2020-03-28 58313000000 NaN 22370000000 NaN NaN NaN NaN NaN NaN NaN 11249000000 0.64 0.64 NaN NaN\\n9 9 2020-03-29-2020-06-27 59685000000 NaN 22680000000 NaN NaN NaN NaN NaN NaN NaN 11253000000 0.65 0.65 NaN NaN\\n10 10 2020-06-28-2020-09-26 64698000000 NaN 24689000000 NaN NaN NaN NaN NaN NaN NaN 12673000000 0.74 0.73 NaN NaN', 'shape': '(11, 17)'}]The column name is 'CostOfGoodsAndServicesSold'. I will now write and execute Python code to find the highest value in this column.running 2th step.= running tool run_python_code, with parameters: {'code': \"import pandas as pd\\n\\ndf = pd.read_csv('income_statement.csv')\\n\\nprint(df['CostOfGoodsAndServicesSold'].max())\"}== tool results: [{'python_answer': '169559000000.0\\n'}]The highest value of cost of goods and services is 169559000000.0.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.373Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":29,"totalLines":148,"estimatedTokens":7597}}197{"id":"doc-long_form_text_strategies_with_cohere_cohere-c5d9b82e","source":"documentation","title":"Long-Form Text Strategies with Cohere | Cohere","url":"https://docs.cohere.com/page/long-form-general-strategies","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1%%capture2!pip install cohere3!pip install python-dotenv4!pip install tokenizers5!pip install langchain6!pip install nltk7!pip install networkx8!pip install pypdf2\n```\n\nExample:\n```text\n1import os2import requests3from collections import deque4from typing import List, Tuple56import cohere78import numpy as np910import PyPDF211from dotenv import load_dotenv1213from tokenizers import Tokenizer1415import nltk16nltk.download('punkt') # Download the necessary data for sentence tokenization17from nltk.tokenize import sent_tokenize1819import networkx as nx20from getpass import getpass\n```\n\nExample:\n```text\n[nltk_data] Downloading package punkt to[nltk_data] /home/anna_cohere_com/nltk_data...[nltk_data] Package punkt is already up-to-date!\n```\n\nExample:\n```text\n1# Set up Cohere client2co_model = 'command-a-03-2025'3co_api_key = getpass(\"Enter your Cohere API key: \")4co = cohere.Client(api_key=co_api_key)\n```\n\nExample:\n```text\n1def load_long_pdf(file_path):2 \"\"\"3 Load a long PDF file and extract its text content.45 Args:6 file_path (str): The path to the PDF file.78 Returns:9 str: The extracted text content of the PDF file.10 \"\"\"11 with open(file_path, 'rb') as file:12 pdf_reader = PyPDF2.PdfReader(file)13 num_pages = len(pdf_reader.pages)14 full_text = ''15 for page_num in range(num_pages):16 page = pdf_reader.pages[page_num]17 full_text += page.extract_text()18 return full_text1920def save_pdf_from_url(pdf_url, save_path):21 try:22 # Send a GET request to the PDF URL23 response = requests.get(pdf_url, stream=True)24 response.raise_for_status() # Raise an exception for HTTP errors2526 # Open the local file for writing in binary mode27 with open(save_path, 'wb') as file:28 # Write the content of the response to the local file29 for chunk in response.iter_content(chunk_size=8192):30 file.write(chunk)3132 print(f\"PDF saved successfully to '{save_path}'\")33 except requests.exceptions.RequestException as e:34 print(f\"Error downloading PDF: {e}\")\n```\n\nExample:\n```text\n1# Download the PDF file from the URL2pdf_url = 'https://data.consilium.europa.eu/doc/document/ST-5662-2024-INIT/en/pdf'3save_path = 'example.pdf'4save_pdf_from_url(pdf_url, save_path)56# Load the PDF file and extract its text content7long_text = load_long_pdf(save_path)8long_text = long_text.replace('\\n', ' ')910# Print the length of the document11print(\"Document length - #tokens:\", len(co.tokenize(text=long_text, model=co_model).tokens))\n```\n\nExample:\n```text\nPDF saved successfully to 'example.pdf'Document length - #tokens: 134184\n```\n\nExample:\n```text\n1def generate_response(message, max_tokens=300, temperature=0.2, k=0):2 \"\"\"3 A wrapper around the Cohere API to generate a response based on a given prompt.45 Args:6 messsage (str): The input message for generating the response.7 max_tokens (int, optional): The maximum number of tokens in the generated response. Defaults to 300.8 temperature (float, optional): Controls the randomness of the generated response. Higher values (e.g., 1.0) make the output more random, while lower values (e.g., 0.2) make it more deterministic. Defaults to 0.2.9 k (int, optional): Controls the diversity of the generated response. Higher values (e.g., 5) make the output more diverse, while lower values (e.g., 0) make it more focused. Defaults to 0.1011 Returns:12 str: The generated response.1314 \"\"\"15 response = co.chat(16 model = co_model,17 message=message,18 max_tokens=max_tokens,19 temperature=temperature20 )21 return response.text\n```\n\nExample:\n```text\n1# Example summary prompt.2prompt_template = \"\"\"3## Instruction4Summarize the following Document in 3-5 sentences. Only answer based on the information provided in the document.56## Document7{document}89## Summary10\"\"\".strip()\n```\n\nExample:\n```text\n1prompt = prompt_template.format(document=long_text)2# print(generate_response(message=prompt))\n```\n\nExample:\n```text\n1# The new Cohere model has a context limit of 128k tokens. However, for the purpose of this exercise, we will assume a smaller context window.2# Employing a smaller context window also has the additional benefit of reducing the cost per request, especially if billed by the number of tokens.34MAX_TOKENS = 4000056def truncate(long: str, max_tokens: int) -> str:7 \"\"\"8 Shortens `long` by brutally truncating it to the first `max_tokens` tokens.9 This can break up sentences, passages, etc.10 \"\"\"1112 tokenized = co.tokenize(text=long, model=co_model).token_strings13 truncated = tokenized[:max_tokens]14 short = \"\".join(truncated)15 return short\n```\n\nExample:\n```text\n1short_text = truncate(long_text, MAX_TOKENS)23prompt = prompt_template.format(document=short_text)4print(generate_response(message=prompt))\n```\n\nExample:\n```text\n1def split_text_into_sentences(text) -> List[str]:2 \"\"\"3 Split the input text into a list of sentences.4 \"\"\"5 sentences = sent_tokenize(text)67 return sentences89def group_sentences_into_passages(sentence_list, n_sentences_per_passage=5):10 \"\"\"11 Group sentences into passages of n_sentences sentences.12 \"\"\"13 passages = []14 passage = \"\"15 for i, sentence in enumerate(sentence_list):16 passage += sentence + \" \"17 if (i + 1) % n_sentences_per_passage == 0:18 passages.append(passage)19 passage = \"\"20 return passages2122def build_simple_chunks(text, n_sentences=5):23 \"\"\"24 Build chunks of text from the input text.25 \"\"\"26 sentences = split_text_into_sentences(text)27 chunks = group_sentences_into_passages(sentences, n_sentences_per_passage=n_sentences)28 return chunks\n```\n\nExample:\n```text\n1sentences = split_text_into_sentences(long_text)2passages = group_sentences_into_passages(sentences, n_sentences_per_passage=5)3print('Example sentence:', np.random.choice(np.asarray(sentences), size=1, replace=False))4print()5print('Example passage:', np.random.choice(np.asarray(passages), size=1, replace=False))\n```\n\nExample:\n```text\nExample sentence: ['The European Data Protection Supervisor may also establish an AI regulatory sandbox for the EU institutions, bodies and agencies and exercise the roles and the tasks of national competent authorities in accordance with this chapter.']Example passage: ['This flexibility could mean, for example a decision by the provider to integrate a part of the necessary testing and reporting processes, information and documentation required under this Regulation into already existing documentation and procedu res required under the existing Union harmonisation legislation listed in Annex II, Section A. This however should not in any way undermine the obligation of the provider to comply with all the applicable requirements. (42a) The risk management system shou ld consist of a continuous, iterative process that is planned and run throughout the entire lifecycle of a high - risk AI system. This process should be aimed at identifying and mitigating the relevant risks of artificial intelligence systems on health, safe ty and fundamental rights. The risk management system should be regularly reviewed and updated to ensure its continuing effectiveness, as well as justification and documentation of any significant decisions and actions taken subject to this Regulation. ']\n```\n\nExample:\n```text\n1def _add_chunks_by_priority(2 chunks: List[str],3 idcs_sorted_by_priority: List[int],4 max_tokens: int,5) -> List[Tuple[int, str]]:6 \"\"\"7 Given chunks of text and their indices sorted by priority (highest priority first), this function8 fills the model context window with as many highest-priority chunks as possible.910 The output is a list of (index, chunk) pairs, ordered by priority. To stitch back the chunks into11 a cohesive text that preserves chronological order, sort the output on its index.12 \"\"\"1314 selected = []15 num_tokens = 016 idcs_queue = deque(idcs_sorted_by_priority)1718 while num_tokens < max_tokens and len(idcs_queue) > 0:19 next_idx = idcs_queue.popleft()20 num_tokens += len(co.tokenize(text=chunks[next_idx], model=co_model).tokens)21 # keep index and chunk, to reorder chronologically22 selected.append((next_idx, chunks[next_idx]))23 if num_tokens > max_tokens:24 selected.pop()2526 return selected2728def query_based_retrieval(29 long: str,30 max_tokens: int,31 query: str,32 n_setences_per_passage: int = 5,33) -> str:34 \"\"\"35 Performs query-based retrieval on a long text document.36 \"\"\"37 # 1. Chunk text into units38 chunks = build_simple_chunks(long, n_setences_per_passage)3940 # 2. Use co.rerank to rank chunks vs. query41 chunks_reranked = co.rerank(query=query, documents=chunks, model=\"rerank-english-v3.0\")42 idcs_sorted_by_relevance = [43 chunk.index for chunk in sorted(chunks_reranked.results, key=lambda c: c.relevance_score, reverse=True)44 ]4546 # 3. Add chunks back in order of relevance47 selected = _add_chunks_by_priority(chunks, idcs_sorted_by_relevance, max_tokens)4849 # 4. Put condensed text back in original order50 separator = \" \"51 short = separator.join([chunk for index, chunk in sorted(selected, key=lambda item: item[0], reverse=False)])52 return short\n```\n\nExample:\n```text\n1# Example prompt2prompt_template = \"\"\"3## Instruction4{query}56## Document7{document}89## Answer10\"\"\".strip()\n```\n\nExample:\n```text\n1query = \"What does the report say about biometric identification? Answer only based on the document.\"2short_text = query_based_retrieval(long_text, MAX_TOKENS, query)3prompt = prompt_template.format(query=query, document=short_text)4print(generate_response(message=prompt, max_tokens=300))\n```\n\nExample:\n```text\nThe report outlines several key points regarding biometric identification within the context of the proposed Artificial Intelligence Act:1. **Prohibition of Real-Time Biometric Identification in Public Spaces**: The report proposes a ban on real-time biometric identification by law enforcement authorities in publicly accessible spaces, with specific exceptions. These exceptions are detailed in Article 5(1)(d) and are subject to safeguards, including monitoring, oversight, and limited reporting obligations at the EU level.2. **Exceptions to the Prohibition**: The exceptions to the ban on real-time biometric identification include: - Search for victims of specific crimes (e.g., abduction, trafficking, sexual exploitation). - Prevention of imminent threats to life or physical safety, including terrorist attacks. - Localization or identification of suspects for serious criminal offenses (as defined in Annex IIa) punishable by a custodial sentence of at least four years.3. **Safeguards and Conditions**: The use of real-time biometric identification systems in these exceptional cases must comply with specific safeguards and conditions, including: - A fundamental rights impact assessment. - Registration of the system in a database. - Prior authorization by a judicial or independent administrative authority, except in urgent situations where authorization can be sought within 24 hours. - Limitation to what is strictly necessary in terms of time, geography, and personal scope.4. **Post-Remote Biometric Identification**: The use of post-remote biometric identification\n```\n\nExample:\n```text\n1def text_rank(text: str, max_tokens: int, n_setences_per_passage: int) -> str:2 \"\"\"3 Shortens text by extracting key units of text from it based on their centrality.4 The output is the concatenation of those key units, in their original order.5 \"\"\"67 # 1. Chunk text into units8 chunks = build_simple_chunks(text, n_setences_per_passage)910 # 2. Embed and construct similarity matrix11 embeddings = np.array(12 co.embed(13 texts=chunks,14 model=\"embed-v4.0\",15 input_type=\"clustering\",16 ).embeddings17 )18 similarities = np.dot(embeddings, embeddings.T)1920 # 3. Compute centrality and sort sentences by centrality21 # Easiest to use networkx's `degree` function with similarity as weight22 g = nx.from_numpy_array(similarities, edge_attr=\"weight\")23 centralities = g.degree(weight=\"weight\")24 idcs_sorted_by_centrality = [node for node, degree in sorted(centralities, key=lambda item: item[1], reverse=True)]2526 # 4. Add chunks back in order of centrality27 selected = _add_chunks_by_priority(chunks, idcs_sorted_by_centrality, max_tokens)2829 # 5. Put condensed text back in original order30 short = \" \".join([chunk for index, chunk in sorted(selected, key=lambda item: item[0], reverse=False)])3132 return short\n```\n\nExample:\n```text\n1short_text = text_rank(long_text, MAX_TOKENS, 5)2prompt = prompt_template.format(document=short_text)3print(generate_response(message=prompt, max_tokens=600))\n```\n\nExample:\n```text\nThe document outlines the European Union's regulatory framework for artificial intelligence (AI) systems, focusing on high-risk AI applications. It establishes rules for placing AI systems on the market, including prohibitions on certain practices, requirements for high-risk systems, and transparency obligations. The regulation defines high-risk AI systems based on their intended use and potential risks to health, safety, and fundamental rights. Providers of high-risk AI systems must comply with specific requirements, such as risk management, data governance, and human oversight. The regulation also mandates conformity assessments, registration in an EU database, and post-market monitoring. It emphasizes the importance of AI literacy, prohibits manipulative or exploitative AI practices, and ensures compliance through market surveillance and enforcement mechanisms. Additionally, the regulation addresses general-purpose AI models, requiring providers to meet specific obligations, especially for models with systemic risks. The framework aims to promote trustworthy AI while safeguarding public interests and supporting innovation.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.374Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":22,"totalLines":113,"estimatedTokens":3607}}198{"id":"doc-wikipedia_semantic_search_with_cohere_embedding_-74d21736","source":"documentation","title":"Wikipedia Semantic Search with Cohere Embedding Archives | Cohere","url":"https://docs.cohere.com/page/wikipedia-semantic-search","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from datasets import load_dataset2import torch3import cohere4s5co = cohere.Client(\"\")67#Load at max 1000 documents + embeddings8max_docs = 10009docs_stream = load_dataset(f\"Cohere/wikipedia-22-12-simple-embeddings\", split=\"train\", streaming=True)1011docs = []12doc_embeddings = []1314for doc in docs_stream:15 docs.append(doc)16 doc_embeddings.append(doc['emb'])17 if len(docs) >= max_docs:18 break1920doc_embeddings = torch.tensor(doc_embeddings)\n```\n\nExample:\n```text\nDownloading: 0%| | 0.00/1.29k [00:00<?, ?B/s]Using custom data configuration Cohere--wikipedia-22-12-simple-embeddings-94deea3d55a22093\n```\n\nExample:\n```text\n1doc_embeddings.shape\n```\n\nExample:\n```text\ntorch.Size([1000, 768])\n```\n\nExample:\n```text\n1query = 'Who founded Wikipedia'2response = co.embed(texts=[query], model='embed-v4.0')3query_embedding = response.embeddings4query_embedding = torch.tensor(query_embedding)56dot_scores = torch.mm(query_embedding, doc_embeddings.transpose(0, 1))7top_k = torch.topk(dot_scores, k=3)89print(\"Query:\", query)10for doc_id in top_k.indices[0].tolist():11 print(docs[doc_id]['title'])12 print(docs[doc_id]['text'], \"\\n\")\n```\n\nExample:\n```text\nQuery: Who founded WikipediaWikipediaLarry Sanger and Jimmy Wales are the ones who started Wikipedia. Wales is credited with defining the goals of the project. Sanger created the strategy of using a wiki to reach Wales' goal. On January 10, 2001, Larry Sanger proposed on the Nupedia mailing list to create a wiki as a \"feeder\" project for Nupedia. Wikipedia was launched on January 15, 2001. It was launched as an English-language edition at www.wikipedia.com, and announced by Sanger on the Nupedia mailing list. Wikipedia's policy of \"neutral point-of-view\" was enforced in its initial months, and was similar to Nupedia's earlier \"nonbiased\" policy. Otherwise, there weren't very many rules initially, and Wikipedia operated independently of Nupedia.WikipediaWikipedia began as a related project for Nupedia. Nupedia was a free English-language online encyclopedia project. Nupedia's articles were written and owned by Bomis, Inc which was a web portal company. The important people of the company were Jimmy Wales, the person in charge of Bomis, and Larry Sanger, the editor-in-chief of Nupedia. Nupedia was first licensed under the Nupedia Open Content License which was changed to the GNU Free Documentation License before Wikipedia was founded and made their first article when Richard Stallman requested them.WikipediaWikipedia was started on January 10, 2001, by Jimmy Wales and Larry Sanger as part of an earlier online encyclopedia named Nupedia. On January 15, 2001, Wikipedia became a separate website of its own. It is a wiki that uses the software MediaWiki (like all other Wikimedia Foundation projects).\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.375Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":6,"totalLines":33,"estimatedTokens":754}}199{"id":"doc-creating_a_qa_bot_from_technical_documentation_c-70223470","source":"documentation","title":"Creating a QA Bot From Technical Documentation | Cohere","url":"https://docs.cohere.com/page/creating-a-qa-bot","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1%%capture2!pip install cohere datasets llama_index llama-index-llms-cohere llama-index-embeddings-cohere\n```\n\nExample:\n```text\n1import cohere2import datasets3from llama_index.core import StorageContext, VectorStoreIndex, load_index_from_storage4from llama_index.core.schema import TextNode5from llama_index.embeddings.cohere import CohereEmbedding6import pandas as pd78import json9from pathlib import Path10from tqdm import tqdm11from typing import List\n```\n\nExample:\n```text\n1api_key = \"\" # <your api=\"\" key=\"\">2co = cohere.Client(api_key=api_key)\n```\n\nExample:\n```text\n1data = datasets.load_dataset(\"sauravjoshi23/aws-documentation-chunked\")2print(data)34map_id2index = {sample[\"id\"]: index for index, sample in enumerate(data[\"train\"])}\n```\n\nExample:\n```text\n/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_token.py:88: UserWarning:The secret `HF_TOKEN` does not exist in your Colab secrets.To authenticate with the Hugging Face Hub, create a token in your settings tab (https://huggingface.co/settings/tokens), set it as secret in your Google Colab and restart your session.You will be able to reuse this secret in all of your notebooks.Please note that authentication is recommended but still optional to access public models or datasets. warnings.warn(DatasetDict({ train: Dataset({ features: ['id', 'text', 'source'], num_rows: 187147 })})\n```\n\nExample:\n```text\n1overwrite = True # only compute index if it doesn't exist2path_index = Path(\".\") / \"aws-documentation_index_cohere\"34embed_model = CohereEmbedding(5 cohere_api_key=api_key,6 model_name=\"embed-v4.0\",7)89if not path_index.exists() or overwrite:10 # Documents are prechunked. Keep them as-is for now11 stub_len = len(\"https://github.com/siagholami/aws-documentation/tree/main/documents/\")12 documents = [13 # -- for indexing full dataset --14 TextNode(15 text=sample[\"text\"],16 title=sample[\"source\"][stub_len:], # save source minus stub17 id_=sample[\"id\"],18 ) for sample in data[\"train\"]19 # -- for testing on subset --20 # TextNode(21 # text=data[\"train\"][index][\"text\"],22 # title=data[\"train\"][index][\"source\"][stub_len:],23 # id_=data[\"train\"][index][\"id\"],24 # ) for index in range(1_000)25 ]26 index = VectorStoreIndex(documents, embed_model=embed_model)27 index.storage_context.persist(path_index)2829else:30 storage_context = StorageContext.from_defaults(persist_dir=path_index)31 index = load_index_from_storage(storage_context, embed_model=embed_model)\n```\n\nExample:\n```text\n1retriever = index.as_retriever(similarity_top_k=top_k)\n```\n\nExample:\n```text\n1class RetrieverWithRerank:2 def __init__(self, retriever, api_key):3 self.retriever = retriever4 self.co = cohere.Client(api_key=api_key)56 def retrieve(self, query: str, top_n: int):7 # First call to the retriever fetches the closest indices8 nodes = self.retriever.retrieve(query)9 nodes = [10 {11 \"text\": node.node.text,12 \"llamaindex_id\": node.node.id_,13 }14 for node15 in nodes16 ]17 # Call co.rerank to improve the relevance of retrieved documents18 reranked = self.co.rerank(query=query, documents=nodes, model=\"rerank-english-v3.0\", top_n=top_n)19 nodes = [nodes[node.index] for node in reranked.results]20 return nodes212223top_k = 60 # how many documents to fetch on first pass24top_n = 20 # how many documents to sub-select with rerank2526retriever = RetrieverWithRerank(27 index.as_retriever(similarity_top_k=top_k),28 api_key=api_key,29)\n```\n\nExample:\n```text\n1query = \"What happens to my Amazon EC2 instances if I delete my Auto Scaling group?\"23documents = retriever.retrieve(query, top_n=top_n)45resp = co.chat(message=query, model=\"command-r-08-2024\", temperature=0., documents=documents)6print(resp.text)\n```\n\nExample:\n```text\n1def build_answer_with_citations(response):2 \"\"\" \"\"\"3 text = response.text4 citations = response.citations56 # Construct text_with_citations adding citation spans as we iterate through citations7 end = 08 text_with_citations = \"\"910 for citation in citations:11 # Add snippet between last citatiton and current citation12 start = citation.start13 text_with_citations += text[end : start]14 end = citation.end # overwrite15 citation_blocks = \" [\" + \", \".join([stub[4:] for stub in citation.document_ids]) + \"] \"16 text_with_citations += text[start : end] + citation_blocks17 # Add any left-over18 text_with_citations += text[end:]1920 return text_with_citations2122grounded_answer = build_answer_with_citations(resp)23print(grounded_answer)\n```\n\nExample:\n```text\n1url = \"https://github.com/siagholami/aws-documentation/blob/main/QA_true.csv?raw=true\"2qa_pairs = pd.read_csv(url)3qa_pairs.sample(2)\n```\n\nExample:\n```text\n1LLM_EVAL_TEMPLATE = \"\"\"## References2{references}34QUESTION: based on the above reference documents, answer the following question: {question}5ANSWER: {answer}6STUDENT RESPONSE: {completion}78Based on the question and answer above, grade the studen't reponse. A correct response will contain exactly \\9the same information as in the answer, even if it is worded differently. If the student's reponse is correct, \\10give it a score of 1. Otherwise, give it a score of 0. Let's think step by step. Return your answer as \\11as a compilable JSON with the following structure:12{{13 \"reasoning\": <reasoning>,14 \"score: <score 0=\"\" 1=\"\" of=\"\" or=\"\">,15}}\"\"\"161718def get_rank_of_golden_within_retrieved(golden: str, retrieved: List[dict]) -> int:19 \"\"\"20 Returns the rank that the golden document (single) has within the retrieved documents21 * `golden` contains the source of the document, e.g. 'amazon-ec2-user-guide/EBSEncryption.md'22 * `retrieved` has a list of responses with key 'llamaindex_id', which links back to document sources23 \"\"\"24 # Create {document: rank} map using llamaindex_id (count first occurrence of any document; they can25 # appear multiple times because they're chunked)26 doc_to_rank = {}27 for rank, doc in enumerate(retrieved):28 # retrieve source of document29 _id = doc[\"llamaindex_id\"]30 source = data[\"train\"][map_id2index[_id]][\"source\"]31 # format as in dataset32 source = source[stub_len:] # remove stub33 source = source.replace(\"/doc_source\", \"\") # remove /doc_source/34 if source not in doc_to_rank:35 doc_to_rank[source] = rank + 13637 # Return rank of `golden`, defaulting to len(retrieved) + 1 if it's absent38 return doc_to_rank.get(golden, len(retrieved) + 1)\n```\n\nExample:\n```text\n1from tqdm import tqdm23answers = []4golden_answers = []5ranks = []6grading_prompts = [] # best computed in batch78for _, row in tqdm(qa_pairs.iterrows(), total=len(qa_pairs)):9 query, golden_answer, golden_doc = row[\"Question\"], row[\"Answer_True\"], row[\"Document_True\"]10 golden_answers.append(golden_answer)1112 # --- Produce answer using retriever ---13 documents = retriever.retrieve(query, top_n=top_n)14 resp = co.chat(message=query, model=\"command-r-08-2024\", temperature=0., documents=documents)15 answer = resp.text16 answers.append(answer)1718 # --- Do some prework for evaluation later ---19 # Rank20 rank = get_rank_of_golden_within_retrieved(golden_doc, documents)21 ranks.append(rank)22 # Score: construct the grading prompts for LLM evals, then evaluate in batch23 # Need to reformat documents slightly24 documents = [{\"index\": str(i), \"text\": doc[\"text\"]} for i, doc in enumerate(documents)]25 references_text = \"\\n\\n\".join(\"\\n\".join([f\"{k}: {v}\" for k, v in doc.items()]) for doc in documents)26 # ^ snippet looks complicated, but all it does it unpack all kwargs from `documents`27 # into text separated by \\n\\n28 grading_prompt = LLM_EVAL_TEMPLATE.format(29 references=references_text, question=query, answer=golden_answer, completion=answer,30 )31 grading_prompts.append(grading_prompt)\n```\n\nExample:\n```text\n1results = pd.DataFrame()2results[\"answer\"] = answers3results[\"golden_answer\"] = qa_pairs[\"Answer_True\"]4results[\"rank\"] = ranks\n```\n\nExample:\n```text\n1scores = []2reasonings = []34def remove_backticks(text: str) -> str:5 \"\"\"6 Some models are trained to output JSON in Markdown formatting:7 ```json {json object}```8 Remove the backticks from those model responses so that they become9 parasable by json.loads.10 \"\"\"11 if text.startswith(\"```json\"):12 text = text[7:]13 if text.endswith(\"```\"):14 text = text[:-3]15 return text161718for prompt in tqdm(grading_prompts, total=len(grading_prompts)):19 resp = co.chat(message=prompt, model=\"command-a-03-2025\", temperature=0.)20 # Convert response to JSON to extract the `score` and `reasoning` fields21 # We remove backticks for compatibility with different LLMs22 parsed = json.loads(remove_backticks(resp.text))23 scores.append(parsed[\"score\"])24 reasonings.append(parsed[\"reasoning\"])\n```\n\nExample:\n```text\n1results[\"score\"] = scores2results[\"reasoning\"] = reasonings\n```\n\nExample:\n```text\n1print(f\"Average score: {results['score'].mean():.3f}\")\n```\n\nExample:\n```text\n1import matplotlib.pyplot as plt2import seaborn as sns34sns.set_theme(style=\"darkgrid\", rc={\"grid.color\": \".8\"})56results[\"rank_shifted_left\"] = results[\"rank\"] - 0.17results[\"rank_shifted_right\"] = results[\"rank\"] + 0.189f, ax = plt.subplots(figsize=(5, 3))10sns.histplot(data=results.loc[results[\"score\"] == 1], x=\"rank_shifted_left\", color=\"skyblue\", label=\"Correct answer\", binwidth=1)11sns.histplot(data=results.loc[results[\"score\"] == 0], x=\"rank_shifted_right\", color=\"red\", label=\"False answer\", binwidth=1)1213ax.set_xticks([1, 5, 0, 10, 15, 20])14ax.set_title(\"Rank of golden document (max means golden doc. wasn't retrieved)\")15ax.set_xlabel(\"Rank\")16ax.legend();\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.376Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":18,"totalLines":93,"estimatedTokens":2562}}200{"id":"doc-pondr_fostering_connection_through_good_conversa-d0a9d993","source":"documentation","title":"Pondr, Fostering Connection through Good Conversation | Cohere","url":"https://docs.cohere.com/page/pondr","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere2from cohere.responses.classify import Example3import pandas as pd\n```\n\nExample:\n```text\n1co=cohere.Client('YOUR_API_KEY')\n```\n\nExample:\n```text\n1#user_input is hardcoded for this example2user_input='I am meeting up with a coworker. We are meeting at a fancy restaurant. I wanna ask some interesting questions. These questions should be deep.'3prompt=user_input+'\\nHere are 10 interesting questions to ask:\\n1)'4response=co.generate(model='xlarge', prompt=prompt, max_tokens=200, temperature=5).generations[0].text5response\n```\n\nExample:\n```text\n1def generation_to_df(generation):2 generation=response.split('\\n')3 clean_questions=[]4 for i in range(10):5 curr_q=generation[i]6 clean_questions.append(curr_q[curr_q.find(')')+1:])7 clean_q_df=pd.DataFrame(clean_questions, columns=['questions'])8 return clean_q_df\n```\n\nExample:\n```text\n1clean_q_df = generation_to_df(response)2pd.options.display.max_colwidth=1503clean_q_df\n```\n\nExample:\n```text\n1interestingness=[2 Example(\"What do you think is the hardest part of what I do for a living?\", \"Not Interesting\"),3 Example(\"What\\'s the first thing you noticed about me?\", \"Interesting\"),4 Example(\"Do you think plants thrive or die in my care?\", \"Interesting\"),5 Example(\"Do I seem like more of a creative or analytical type?\", \"Interesting\"),6 Example(\"What subject do you think I thrived at in school?\", \"Not Interesting\"),7 Example(\"What\\'s been your happiest memory this past year?\", \"Interesting\"),8 Example(\"What lesson took you the longest to un-learn?\", \"Not Interesting\"),9 Example(\"How can you become a better person?\", \"Not Interesting\"),10 Example(\"Do you think I intimidate others? Why or why not?\", \"Interesting\"),11 Example(\"What\\'s the most embarrassing thing that happened to you on a date?\", \"Not Interesting\"),12 Example(\"How would you describe what you think my type is in three words?\", \"Interesting\"),13 Example(\"What do you think I\\'m most likely to splurge on?\", \"Interesting\"),14 Example(\"As a child what do you think I wanted to be when I grow up?\", \"Interesting\"),15 Example(\"Do you think you are usually early, on time, or late to events?\", \"Not Interesting\"),16 Example(\"Do you think I was popular at school?\", \"Interesting\"),17 Example(\"What questions are you trying to answer most in your life right now?\", \"Not Interesting\")]18specificity=[19 Example(\"What\\'s the first thing you noticed about me?\", \"Specific\"),20 Example(\"Do you think plants thrive or die in my care?\", \"Specific\"),21 Example(\"Do I seem like more of a creative or analytical type?\", \"Not Specific\"),22 Example(\"How would you describe what you think my type is in three words?\", \"Not Specific\"),23 Example(\"What do you think I\\'m most likely to splurge on?\", \"Specific\"),24 Example(\"What subject do you think I thrived at in school?\", \"Not Specific\"),25 Example(\"As a child what do you think I wanted to be when I grow up?\", \"Specific\"),26 Example(\"Do you think I was popular at school?\", \"Specific\"),27 Example(\"Do you think you\\'re usually early, on time, or late to events?\", \"Specific\"),28 Example(\"Do you think I intimidate others? Why or why not?\", \"Specific\"),29 Example(\"What\\'s been your happiest memory this past year?\", \"Not Specific\"),30 Example(\"What subject do you think I thrived at in school?\", \"Specific\"),31 Example(\"What\\'s the biggest mistake that you think you needed to make to become who you are now?\", \"Specific\"),32 Example(\"Is there anything you\\'ve done recently that you\\'re incredibly proud of?\", \"Not Specific\"),33 Example(\"How are you and your siblings similar?\", \"Not Specific\"),34 Example(\"What\\'s the worst pain you have ever been in that wasn\\'t physical?\", \"Specific\"),35 Example(\"Has a stranger ever changed your life?\", \"Specific\"),36 Example(\"Do you think the image you have of yourself matches the image other people see you as?\", \"Specific\"),37 Example(\"What would your younger self not believe about your life today?\", \"Specific\")]\n```\n\nExample:\n```text\n1def add_attribute(df, attribute, name, target):23 response = co.classify(4 model='medium',5 inputs=list(df['questions']),6 examples=attribute)78 q_conf=[]9 for q in response.classifications:10 q_conf.append(q.labels[target].confidence)1112 df[name]=q_conf\n```\n\nExample:\n```text\n1add_attribute(clean_q_df, interestingness, 'interestingness', 'Interesting')2add_attribute(clean_q_df, specificity, 'specificity', 'Specific')3clean_q_df['average']= clean_q_df.iloc[:,1:].mean(axis=1)4clean_q_df.sort_values(by='average', ascending=False)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.376Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":8,"totalLines":43,"estimatedTokens":1212}}201{"id":"doc-effective_chunking_strategies_for_rag_cohere-5e4472d9","source":"documentation","title":"Effective Chunking Strategies for RAG | Cohere","url":"https://docs.cohere.com/page/chunking-strategies","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1%%capture2!pip install cohere3!pip install -qU langchain-text-splitters4!pip install llama-index-embeddings-cohere5!pip install llama-index-postprocessor-cohere-rerank\n```\n\nExample:\n```text\n1import requests2from typing import List34from bs4 import BeautifulSoup56import cohere7from getpass import getpass8from IPython.display import HTML, display910from langchain_text_splitters import CharacterTextSplitter11from langchain_text_splitters import RecursiveCharacterTextSplitter1213from llama_index.core import Document14from llama_index.embeddings.cohere import CohereEmbedding15from llama_index.postprocessor.cohere_rerank import CohereRerank16from llama_index.core import VectorStoreIndex, ServiceContext\n```\n\nExample:\n```text\n1co_model = 'command-a-03-2025'2co_api_key = getpass(\"Enter Cohere API key: \")3co = cohere.Client(api_key=co_api_key)\n```\n\nExample:\n```text\nEnter Cohere API key: ··········\n```\n\nExample:\n```text\n1def set_css(*args, **kwargs):2 display(HTML('''3 <style>4 pre {5 white-space: pre-wrap;6 }7 </style>8 '''))9get_ipython().events.register('pre_run_cell', set_css)1011set_css()\n```\n\nExample:\n```text\n1def insert_citations(text: str, citations: List[dict]):2 \"\"\"3 A helper function to pretty print citations.4 \"\"\"5 offset = 06 # Process citations in the order they were provided7 for citation in citations:8 # Adjust start/end with offset9 start, end = citation.start + offset, citation.end + offset10 placeholder = \"[\" + \", \".join(doc[4:] for doc in citation.document_ids) + \"]\"11 # ^ doc[4:] removes the 'doc_' prefix, and leaves the quoted document12 modification = f'{text[start:end]} {placeholder}'13 # Replace the cited text with its bolded version + placeholder14 text = text[:start] + modification + text[end:]15 # Update the offset for subsequent replacements16 offset += len(modification) - (end - start)1718 return text1920def build_retreiver(documents, top_n=5):21 # Create the embedding model22 embed_model = CohereEmbedding(23 cohere_api_key=co_api_key,24 model_name=\"embed-v4.0\",25 input_type=\"search_query\",26 )2728 # Load the data, for this example data needs to be in a test file29 index = VectorStoreIndex.from_documents(30 documents,31 embed_model=embed_model32 )3334 # Create a cohere reranker35 cohere_rerank = CohereRerank(api_key=co_api_key)3637 # Create the retriever38 retriever = index.as_retriever(node_postprocessors=[cohere_rerank], similarity_top_k=top_n)39 return retriever\n```\n\nExample:\n```text\n1# Get all investement memos (19) in bvp repository2url_path = 'https://www.fool.com/earnings/call-transcripts/2024/01/24/tesla-tsla-q4-2023-earnings-call-transcript/'3response = requests.get(url_path)4soup = BeautifulSoup(response.content, 'html.parser')56target_divs = soup.find(\"div\", {\"class\": \"article-body\"}).find_all(\"p\")[2:]7print('Length of the script: ', len(target_divs))89print()10print('Example of processed text:')11text = '\\n\\n'.join([div.get_text() for div in target_divs])12print(text[:500])\n```\n\nExample:\n```text\nLength of the script: 385Example of processed text:Martin ViechaGood afternoon, everyone, and welcome to Tesla's fourth-quarter 2023 Q&A webcast. My name is Martin Viecha, VP of investor relations, and I'm joined today by Elon Musk, Vaibhav Taneja, and a number of other executives. Our Q4 results were announced at about 3 p.m. Central Time in the update that we published at the same link as this webcast.During this call, we will discuss our business outlook and make forward-looking statements. These comments are based on our predictions and\n```\n\nExample:\n```text\nElon Musk -- Chief Executive Officer and Product ArchitectYeah. The creators of Westworld, Jonathan Nolan, Lisa Joy Nolan, are friends -- are all friends of mine, actually. And I invited them to come see the lab and, like, well, come see it, hopefully soon. It's pretty well -- especially the sort of subsystem test stands where you've just got like one leg on a test stand just doing repetitive exercises and one arm on a test stand pretty well.\n```\n\nExample:\n```text\n1# Define the question2question = \"Who mentions Jonathan Nolan?\"\n```\n\nExample:\n```text\n1# Define the chunking function2def get_chunks(text, chunk_size, chunk_overlap):3 text_splitter = RecursiveCharacterTextSplitter(4 chunk_size=chunk_size,5 chunk_overlap=chunk_overlap,6 length_function=len,7 is_separator_regex=False,8 )910 documents = text_splitter.create_documents([text])11 documents = [Document(text=doc.page_content) for doc in documents]1213 return documents\n```\n\nExample:\n```text\n1chunk_size = 5002chunk_overlap = 03documents = get_chunks(text, chunk_size, chunk_overlap)4retriever = build_retreiver(documents)56source_nodes = retriever.retrieve(question)7print('Number of docuemnts: ',len(source_nodes))8source_nodes= [{\"text\": ni.get_content()}for ni in source_nodes]91011response = co.chat(12 message=question,13 documents=source_nodes,14 model=co_model15)16response = response17print(response.text)\n```\n\nExample:\n```text\nNumber of docuemnts: 5Elon Musk mentions Jonathan Nolan.\n```\n\nExample:\n```text\n1print(insert_citations(response.text, response.citations))\n```\n\nExample:\n```text\nElon Musk [0] mentions Jonathan Nolan.\n```\n\nExample:\n```text\n1print(source_nodes[0])\n```\n\nExample:\n```text\n1{'text': \"Yeah. The creators of Westworld, Jonathan Nolan, Lisa Joy Nolan, are friends -- are all friends of mine, actually. And I invited them to come see the lab and, like, well, come see it, hopefully soon. It's pretty well -- especially the sort of subsystem test stands where you've just got like one leg on a test stand just doing repetitive exercises and one arm on a test stand pretty well.\\n\\nYeah.\\n\\nUnknown speaker\\n\\nWe're not entering Westworld anytime soon.\"}\n```\n\nExample:\n```text\n1chunk_size = 5002chunk_overlap = 1003documents = get_chunks(text,chunk_size, chunk_overlap)4retriever = build_retreiver(documents)56source_nodes = retriever.retrieve(question)7print('Number of docuemnts: ',len(source_nodes))8source_nodes= [{\"text\": ni.get_content()}for ni in source_nodes]91011response = co.chat(12 message=question,13 documents=source_nodes,14 model=co_model15)16response = response17print(response.text)\n```\n\nExample:\n```text\n1source_nodes[0]\n```\n\nExample:\n```text\n1{'text': \"Yeah, not the best reference.\\n\\nElon Musk -- Chief Executive Officer and Product Architect\\n\\nYeah. The creators of Westworld, Jonathan Nolan, Lisa Joy Nolan, are friends -- are all friends of mine, actually. And I invited them to come see the lab and, like, well, come see it, hopefully soon. It's pretty well -- especially the sort of subsystem test stands where you've just got like one leg on a test stand just doing repetitive exercises and one arm on a test stand pretty well.\\n\\nYeah.\"}\n```\n\nExample:\n```text\n1print('HTML text')2print(target_divs[:3])3print('-------------------\\n')45text_custom = []6for div in target_divs:7 if div.get_text() is None:8 continue9 if str(div).startswith('<p><strong>'):10 text_custom.append(f'### {div.get_text()}')11 else:12 text_custom.append(div.get_text())1314text_custom = '\\n'.join(text_custom)15print(text_custom[:500])\n```\n\nExample:\n```text\nHTML text[<p><strong>Martin Viecha</strong></p>, <p>Good afternoon, everyone, and welcome to Tesla's fourth-quarter 2023 Q&A webcast. My name is Martin Viecha, VP of investor relations, and I'm joined today by Elon Musk, Vaibhav Taneja, and a number of other executives. Our Q4 results were announced at about 3 p.m. Central Time in the update that we published at the same link as this webcast.</p>, <p>During this call, we will discuss our business outlook and make forward-looking statements. These comments are based on our predictions and expectations as of today. Actual events or results could differ materially due to a number of risks and uncertainties, including those mentioned in our most recent filings with the SEC. [Operator instructions] But before we jump into Q&A, Elon has some opening remarks.</p>]-------------------### Martin ViechaGood afternoon, everyone, and welcome to Tesla's fourth-quarter 2023 Q&A webcast. My name is Martin Viecha, VP of investor relations, and I'm joined today by Elon Musk, Vaibhav Taneja, and a number of other executives. Our Q4 results were announced at about 3 p.m. Central Time in the update that we published at the same link as this webcast.During this call, we will discuss our business outlook and make forward-looking statements. These comments are based on our predictions an\n```\n\nExample:\n```text\n1separator = \"###\"2chunk_size = 10003chunk_overlap = 045text_splitter = CharacterTextSplitter(6 separator = separator,7 chunk_size=chunk_size,8 chunk_overlap=chunk_overlap,9 length_function=len,10 is_separator_regex=False,11)1213documents = text_splitter.create_documents([text_custom])14documents = [Document(text=doc.page_content) for doc in documents]1516retriever = build_retreiver(documents)1718source_nodes = retriever.retrieve(question)19print('Number of docuemnts: ',len(source_nodes))20source_nodes= [{\"text\": ni.get_content()}for ni in source_nodes]2122response = co.chat(23 message=question,24 documents=source_nodes,25 model=co_model26)27response = response28print(response.text)\n```\n\nExample:\n```text\nCreated a chunk of size 5946, which is longer than the specified 1000Created a chunk of size 4092, which is longer than the specified 1000Created a chunk of size 1782, which is longer than the specified 1000Created a chunk of size 1392, which is longer than the specified 1000Created a chunk of size 2046, which is longer than the specified 1000Created a chunk of size 1152, which is longer than the specified 1000Created a chunk of size 1304, which is longer than the specified 1000Created a chunk of size 1295, which is longer than the specified 1000Created a chunk of size 2090, which is longer than the specified 1000Created a chunk of size 1251, which is longer than the specified 1000Number of docuemnts: 5Elon Musk mentions Jonathan Nolan.\n```\n\nExample:\n```text\n1{'text': \"Elon Musk -- Chief Executive Officer and Product Architect\\nYeah. The creators of Westworld, Jonathan Nolan, Lisa Joy Nolan, are friends -- are all friends of mine, actually. And I invited them to come see the lab and, like, well, come see it, hopefully soon. It's pretty well -- especially the sort of subsystem test stands where you've just got like one leg on a test stand just doing repetitive exercises and one arm on a test stand pretty well.\\nYeah.\\n### Unknown speaker\\nWe're not entering Westworld anytime soon.\\n### Elon Musk -- Chief Executive Officer and Product Architect\\nRight, right. Yeah. I take -- take safety very very seriously.\\n### Martin Viecha\\nThank you. The next question from Norman is: How many Cybertruck orders are in the queue? And when do you anticipate to be able to fulfill existing orders?\"}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.377Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":25,"totalLines":128,"estimatedTokens":2800}}202{"id":"doc-text_classification_using_embeddings_cohere-6711db49","source":"documentation","title":"Text Classification Using Embeddings | Cohere","url":"https://docs.cohere.com/page/text-classification-using-embeddings","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1#!pip install --upgrade cohere\n```\n\nExample:\n```text\n1import cohere2from sklearn.model_selection import train_test_split34import pandas as pd5pd.set_option('display.max_colwidth', None)67df = pd.read_csv('https://github.com/clairett/pytorch-sentiment-classification/raw/master/data/SST2/train.tsv', delimiter='\\t', header=None)\n```\n\nExample:\n```text\n1df.head()\n```\n\nExample:\n```text\n1num_examples = 5002df_sample = df.sample(num_examples)34sentences_train, sentences_test, labels_train, labels_test = train_test_split(5 list(df_sample[0]), list(df_sample[1]), test_size=0.25, random_state=0)678sentences_train = sentences_train[:95]9sentences_test = sentences_test[:95]1011labels_train = labels_train[:95]12labels_test = labels_test[:95]\n```\n\nExample:\n```text\n1model_name = \"embed-v4.0\"2api_key = \"\"34input_type = \"classification\"56co = cohere.Client(api_key)\n```\n\nExample:\n```text\n1embeddings_train = co.embed(texts=sentences_train,2 model=model_name,3 input_type=input_type4 ).embeddings56embeddings_test = co.embed(texts=sentences_test,7 model=model_name,8 input_type=input_type9 ).embeddings\n```\n\nExample:\n```text\n1print(f\"Review text: {sentences_train[0]}\")2print(f\"Embedding vector: {embeddings_train[0][:10]}\")\n```\n\nExample:\n```text\nReview text: the script was reportedly rewritten a dozen times either 11 times too many or else too fewEmbedding vector: [1.1531117, -0.8543223, -1.2496399, -0.28317127, -0.75870246, 0.5373464, 0.63233083, 0.5766576, 1.8336298, 0.44203663]\n```\n\nExample:\n```text\n1from sklearn.svm import SVC2from sklearn.pipeline import make_pipeline3from sklearn.preprocessing import StandardScaler456svm_classifier = make_pipeline(StandardScaler(), SVC(class_weight='balanced'))78svm_classifier.fit(embeddings_train, labels_train)\n```\n\nExample:\n```text\nPipeline(steps=[('standardscaler', StandardScaler()), ('svc', SVC(class_weight='balanced'))])\n```\n\nExample:\n```text\n1score = svm_classifier.score(embeddings_test, labels_test)2print(f\"Validation accuracy on is {100*score}%!\")\n```\n\nExample:\n```text\nValidation accuracy on Large is 91.2%!\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.377Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":614}}203{"id":"doc-multilingual_search_with_cohere_and_langchain_co-0dabf9bf","source":"documentation","title":"Multilingual Search with Cohere and Langchain | Cohere","url":"https://docs.cohere.com/page/multilingual-search","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1from langchain.embeddings.cohere import CohereEmbeddings2from langchain.llms import Cohere3from langchain.prompts import PromptTemplate4from langchain.text_splitter import RecursiveCharacterTextSplitter5from langchain.chains.question_answering import load_qa_chain6from langchain.chains import RetrievalQA7from langchain.vectorstores import Qdrant8from langchain.document_loaders import TextLoader9import textwrap as tr10import random11import dotenv12import os1314dotenv.load_dotenv(\".env\") # Upload an '.env' file containing an environment variable named 'COHERE_API_KEY' using your Cohere API Key\n```\n\nExample:\n```text\nTrue\n```\n\nExample:\n```text\n1import tensorflow_datasets as tfds2dataset = tfds.load('trec', split='train')3texts = [item['text'].decode('utf-8') for item in tfds.as_numpy(dataset)]4print(f\"Number of documents: {len(texts)}\")\n```\n\nExample:\n```text\nDownloading and preparing dataset 350.79 KiB (download: 350.79 KiB, generated: 636.90 KiB, total: 987.69 KiB) to /root/tensorflow_datasets/trec/1.0.0...Dl Completed...: 0 url [00:00, ? url/s]Dl Size...: 0 MiB [00:00, ? MiB/s]Extraction completed...: 0 file [00:00, ? file/s]Generating splits...: 0%| | 0/2 [00:00<?, ? splits/s]Generating train examples...: 0%| | 0/5452 [00:00<?, ? examples/s]Shuffling /root/tensorflow_datasets/trec/1.0.0.incompleteWOR5EP/trec-train.tfrecord*...: 0%| | 0/54…Generating test examples...: 0%| | 0/500 [00:00<?, ? examples/s]Shuffling /root/tensorflow_datasets/trec/1.0.0.incompleteWOR5EP/trec-test.tfrecord*...: 0%| | 0/500…Dataset trec downloaded and prepared to /root/tensorflow_datasets/trec/1.0.0. Subsequent calls will reuse this data.Number of documents: 5452\n```\n\nExample:\n```text\n1random.seed(11)2for item in random.sample(texts, 5):3 print(item)\n```\n\nExample:\n```text\nWhat is the starting salary for beginning lawyers ?Where did Bill Gates go to college ?What task does the Bouvier breed of dog perform ?What are the top boy names in the U.S. ?What is a female rabbit called ?\n```\n\nExample:\n```text\n1embeddings = CohereEmbeddings(model = \"multilingual-22-12\")23db = Qdrant.from_texts(texts, embeddings, location=\":memory:\", collection_name=\"my_documents\", distance_func=\"Dot\")\n```\n\nExample:\n```text\n1queries = [\"How to get in touch with Bill Gates\",2 \"Comment entrer en contact avec Bill Gates\",3 \"Cara menghubungi Bill Gates\"]45queries_lang = [\"English\", \"French\", \"Indonesian\"]\n```\n\nExample:\n```text\n1answers = []2for query in queries:3 docs = db.similarity_search(query)4 answers.append(docs[0].page_content)\n```\n\nExample:\n```text\n1for idx,query in enumerate(queries):2 print(f\"Query language: {queries_lang[idx]}\")3 print(f\"Query: {query}\")4 print(f\"Most similar existing question: {answers[idx]}\")5 print(\"-\"*20,\"\\n\")\n```\n\nExample:\n```text\nQuery language: EnglishQuery: How to get in touch with Bill GatesMost similar existing question: What is Bill Gates of Microsoft E-mail address ?--------------------Query language: FrenchQuery: Comment entrer en contact avec Bill GatesMost similar existing question: What is Bill Gates of Microsoft E-mail address ?--------------------Query language: IndonesianQuery: Cara menghubungi Bill GatesMost similar existing question: What is Bill Gates of Microsoft E-mail address ?--------------------\n```\n\nExample:\n```text\n1!wget 'https://docs.google.com/uc?export=download&id=1f1INWOfJrHTFmbyF_0be5b4u_moz3a4F' -O steve-jobs-commencement.txt\n```\n\nExample:\n```text\n--2023-06-08 06:11:19-- https://docs.google.com/uc?export=download&id=1f1INWOfJrHTFmbyF_0be5b4u_moz3a4FResolving docs.google.com (docs.google.com)... 74.125.200.101, 74.125.200.138, 74.125.200.102, ...Connecting to docs.google.com (docs.google.com)|74.125.200.101|:443... connected.HTTP request sent, awaiting response... 303 See OtherLocation: https://doc-0g-84-docs.googleusercontent.com/docs/securesc/ha0ro937gcuc7l7deffksulhg5h7mbp1/84t4moii9dmg08hmrh6rfpp8ecrjh6jq/1686204675000/12721472133292131824/*/1f1INWOfJrHTFmbyF_0be5b4u_moz3a4F?e=download&uuid=a26288c7-ad0c-4707-ae0b-72cb94c224dc [following]Warning: wildcards not supported in HTTP.--2023-06-08 06:11:19-- https://doc-0g-84-docs.googleusercontent.com/docs/securesc/ha0ro937gcuc7l7deffksulhg5h7mbp1/84t4moii9dmg08hmrh6rfpp8ecrjh6jq/1686204675000/12721472133292131824/*/1f1INWOfJrHTFmbyF_0be5b4u_moz3a4F?e=download&uuid=a26288c7-ad0c-4707-ae0b-72cb94c224dcResolving doc-0g-84-docs.googleusercontent.com (doc-0g-84-docs.googleusercontent.com)... 74.125.130.132, 2404:6800:4003:c01::84Connecting to doc-0g-84-docs.googleusercontent.com (doc-0g-84-docs.googleusercontent.com)|74.125.130.132|:443... connected.HTTP request sent, awaiting response... 200 OKLength: 11993 (12K) [text/plain]Saving to: ‘steve-jobs-commencement.txt’steve-jobs-commence 100%[===================>] 11.71K --.-KB/s in 0s2023-06-08 06:11:20 (115 MB/s) - ‘steve-jobs-commencement.txt’ saved [11993/11993]\n```\n\nExample:\n```text\n1loader = TextLoader(\"steve-jobs-commencement.txt\")2documents = loader.load()3text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)4texts = text_splitter.split_documents(documents)\n```\n\nExample:\n```text\n1embeddings = CohereEmbeddings(model = \"multilingual-22-12\")2db = Qdrant.from_documents(texts, embeddings, location=\":memory:\", collection_name=\"my_documents\", distance_func=\"Dot\")\n```\n\nExample:\n```text\n1questions = [2 \"What did the author liken The Whole Earth Catalog to?\",3 \"What was Reed College great at?\",4 \"What was the author diagnosed with?\",5 \"What is the key lesson from this article?\",6 \"What did the article say about Michael Jackson?\",7 ]\n```\n\nExample:\n```text\n1prompt_template = \"\"\"Text: {context}23Question: {question}45Answer the question based on the text provided. If the text doesn't contain the answer, reply that the answer is not available.\"\"\"67PROMPT = PromptTemplate(8 template=prompt_template, input_variables=[\"context\", \"question\"]9)\n```\n\nExample:\n```text\n1chain_type_kwargs = {\"prompt\": PROMPT}23qa = RetrievalQA.from_chain_type(llm=Cohere(model=\"command\", temperature=0),4 chain_type=\"stuff\",5 retriever=db.as_retriever(),6 chain_type_kwargs=chain_type_kwargs,7 return_source_documents=True)89for question in questions:10 answer = qa({\"query\": question})11 result = answer[\"result\"].replace(\"\\n\",\"\").replace(\"Answer:\",\"\")12 sources = answer['source_documents']13 print(\"-\"*150,\"\\n\")14 print(f\"Question: {question}\")15 print(f\"Answer: {result}\")1617 ### COMMENT OUT THE 4 LINES BELOW TO HIDE THE SOURCES18 print(f\"\\nSources:\")19 for idx, source in enumerate(sources):20 source_wrapped = tr.fill(str(source.page_content), width=150)21 print(f\"{idx+1}: {source_wrapped}\")\n```\n\nExample:\n```text\n------------------------------------------------------------------------------------------------------------------------------------------------------Question: What did the author liken The Whole Earth Catalog to?Answer: It was sort of like Google in paperback form, 35 years before Google came alongSources:1: When I was young, there was an amazing publication called The Whole Earth Catalog, which was one of the bibles of my generation. It was created by afellow named Stewart Brand not far from here in Menlo Park, and he brought it to life with his poetic touch. This was in the late 1960s, beforepersonal computers and desktop publishing, so it was all made with typewriters, scissors and Polaroid cameras. It was sort of like Google in paperbackform, 35 years before Google came along: It was2: Stewart and his team put out several issues of The Whole Earth Catalog, and then when it had run its course, they put out a final issue. It was themid-1970s, and I was your age. On the back cover of their final issue was a photograph of an early morning country road, the kind you might findyourself hitchhiking on if you were so adventurous. Beneath it were the words: “Stay Hungry. Stay Foolish.” It was their farewell message as theysigned off. Stay Hungry. Stay Foolish. And I have always3: idealistic, and overflowing with neat tools and great notions.4: beautiful, historical, artistically subtle in a way that science can’t capture, and I found it fascinating.------------------------------------------------------------------------------------------------------------------------------------------------------Question: What was Reed College great at?Answer: Reed College was great at calligraphy instruction.Sources:1: Reed College at that time offered perhaps the best calligraphy instruction in the country. Throughout the campus every poster, every label on everydrawer, was beautifully hand calligraphed. Because I had dropped out and didn’t have to take the normal classes, I decided to take a calligraphy classto learn how to do this. I learned about serif and sans serif typefaces, about varying the amount of space between different letter combinations,about what makes great typography great. It was2: I dropped out of Reed College after the first 6 months, but then stayed around as a drop-in for another 18 months or so before I really quit. So whydid I drop out?3: never dropped out, I would have never dropped in on this calligraphy class, and personal computers might not have the wonderful typography that theydo. Of course it was impossible to connect the dots looking forward when I was in college. But it was very, very clear looking backward 10 yearslater.4: OK. It was pretty scary at the time, but looking back it was one of the best decisions I ever made. The minute I dropped out I could stop taking therequired classes that didn’t interest me, and begin dropping in on the ones that looked interesting.------------------------------------------------------------------------------------------------------------------------------------------------------Question: What was the author diagnosed with?Answer: The author was diagnosed with cancer.Sources:1: I lived with that diagnosis all day. Later that evening I had a biopsy, where they stuck an endoscope down my throat, through my stomach and into myintestines, put a needle into my pancreas and got a few cells from the tumor. I was sedated, but my wife, who was there, told me that when they viewedthe cells under a microscope the doctors started crying because it turned out to be a very rare form of pancreatic cancer that is curable withsurgery. I had the surgery and I’m fine now.2: About a year ago I was diagnosed with cancer. I had a scan at 7:30 in the morning, and it clearly showed a tumor on my pancreas. I didn’t even knowwhat a pancreas was. The doctors told me this was almost certainly a type of cancer that is incurable, and that I should expect to live no longer thanthree to six months. My doctor advised me to go home and get my affairs in order, which is doctor’s code for prepare to die. It means to try to tellyour kids everything you thought you’d have the3: Stewart and his team put out several issues of The Whole Earth Catalog, and then when it had run its course, they put out a final issue. It was themid-1970s, and I was your age. On the back cover of their final issue was a photograph of an early morning country road, the kind you might findyourself hitchhiking on if you were so adventurous. Beneath it were the words: “Stay Hungry. Stay Foolish.” It was their farewell message as theysigned off. Stay Hungry. Stay Foolish. And I have always4: beautiful, historical, artistically subtle in a way that science can’t capture, and I found it fascinating.------------------------------------------------------------------------------------------------------------------------------------------------------Question: What is the key lesson from this article?Answer: The key lesson from this article is that you have to trust that the dots will somehow connect in your future. You have to trust in something -- your gut, destiny, life, karma, whatever. This approach has never let me down, and it has made all the difference in my life.Sources:1: Again, you can’t connect the dots looking forward; you can only connect them looking backward. So you have to trust that the dots will somehow connectin your future. You have to trust in something — your gut, destiny, life, karma, whatever. This approach has never let me down, and it has made allthe difference in my life. My second story is about love and loss.2: Remembering that I’ll be dead soon is the most important tool I’ve ever encountered to help me make the big choices in life. Because almost everything— all external expectations, all pride, all fear of embarrassment or failure — these things just fall away in the face of death, leaving only what istruly important. Remembering that you are going to die is the best way I know to avoid the trap of thinking you have something to lose. You arealready naked. There is no reason not to follow your3: Your time is limited, so don’t waste it living someone else’s life. Don’t be trapped by dogma — which is living with the results of other people’sthinking. Don’t let the noise of others’ opinions drown out your own inner voice. And most important, have the courage to follow your heart andintuition. They somehow already know what you truly want to become. Everything else is secondary.4: I really didn’t know what to do for a few months. I felt that I had let the previous generation of entrepreneurs down — that I had dropped the batonas it was being passed to me. I met with David Packard and Bob Noyce and tried to apologize for screwing up so badly. I was a very public failure, andI even thought about running away from the valley. But something slowly began to dawn on me — I still loved what I did. The turn of events at Applehad not changed that one bit. I had been rejected,------------------------------------------------------------------------------------------------------------------------------------------------------Question: What did the article say about Michael Jackson?Answer: The text did not provide information about Michael Jackson.Sources:1: baby boy; do you want him?” They said: “Of course.” My biological mother later found out that my mother had never graduated from college and that myfather had never graduated from high school. She refused to sign the final adoption papers. She only relented a few months later when my parentspromised that I would someday go to college.2: beautiful, historical, artistically subtle in a way that science can’t capture, and I found it fascinating.3: I really didn’t know what to do for a few months. I felt that I had let the previous generation of entrepreneurs down — that I had dropped the batonas it was being passed to me. I met with David Packard and Bob Noyce and tried to apologize for screwing up so badly. I was a very public failure, andI even thought about running away from the valley. But something slowly began to dawn on me — I still loved what I did. The turn of events at Applehad not changed that one bit. I had been rejected,4: This was the closest I’ve been to facing death, and I hope it’s the closest I get for a few more decades. Having lived through it, I can now say thisto you with a bit more certainty than when death was a useful but purely intellectual concept:\n```\n\nExample:\n```text\n1questions_fr = [2 \"À quoi se compare The Whole Earth Catalog ?\",3 \"Dans quoi Reed College était-il excellent ?\",4 \"De quoi l'auteur a-t-il été diagnostiqué ?\",5 \"Quelle est la leçon clé de cet article ?\",6 \"Que disait l'article sur Michael Jackson ?\",7 ]\n```\n\nExample:\n```text\n1chain_type_kwargs = {\"prompt\": PROMPT}23qa = RetrievalQA.from_chain_type(llm=Cohere(model=\"command\", temperature=0),4 chain_type=\"stuff\",5 retriever=db.as_retriever(),6 chain_type_kwargs=chain_type_kwargs,7 return_source_documents=True)89for question in questions_fr:10 answer = qa({\"query\": question})11 result = answer[\"result\"].replace(\"\\n\",\"\").replace(\"Answer:\",\"\")12 sources = answer['source_documents']13 print(\"-\"*20,\"\\n\")14 print(f\"Question: {question}\")15 print(f\"Answer: {result}\")\n```\n\nExample:\n```text\n--------------------Question: À quoi se compare The Whole Earth Catalog ?Answer: The Whole Earth Catalog was like Google in paperback form, 35 years before Google came along.--------------------Question: Dans quoi Reed College était-il excellent ?Answer: Reed College offered the best calligraphy instruction in the country.--------------------Question: De quoi l'auteur a-t-il été diagnostiqué ?Answer: The author was diagnosed with a very rare form of pancreatic cancer that is curable with surgery.--------------------Question: Quelle est la leçon clé de cet article ?Answer: The key lesson of this article is that remembering that you will die soon is the most important tool to help one make the big choices in life.--------------------Question: Que disait l'article sur Michael Jackson ?Answer: The text does not contain the answer to the question.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.379Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":22,"totalLines":113,"estimatedTokens":4372}}204{"id":"doc-migrating_away_from_create_csv_agent_in_langchai-d0bfd908","source":"documentation","title":"Migrating away from create_csv_agent in langchain-cohere | Cohere","url":"https://docs.cohere.com/page/migrate-csv-agent","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n!pip install langchain langchain-core langchain-experimental langchain-cohere pandas -qq\n```\n\nExample:\n```text\n1# Import packages2from datetime import datetime3from io import IOBase4from typing import List, Optional, Union5import pandas as pd6from pydantic import BaseModel, Field78from langchain.agents import AgentExecutor, create_tool_calling_agent9from langchain_core.language_models import BaseLanguageModel10from langchain_core.messages import (11 BaseMessage,12 HumanMessage,13 SystemMessage,14)15from langchain_core.prompts import (16 ChatPromptTemplate,17 MessagesPlaceholder,18)1920from langchain_core.tools import Tool, BaseTool21from langchain_core.prompts.chat import (22 BaseMessagePromptTemplate,23 HumanMessagePromptTemplate,24)2526from langchain_experimental.tools.python.tool import PythonAstREPLTool27from langchain_cohere.chat_models import ChatCohere\n```\n\nExample:\n```text\n1# Replace this cell with your actual cohere api key2os.env[\"COHERE_API_KEY\"] = \"cohere_api_key\"\n```\n\nExample:\n```text\n1# Define prompts that we want to use in the csv agent2FUNCTIONS_WITH_DF = \"\"\"3This is the result of `print(df.head())`:4{df_head}56Do note that the above df isn't the complete df. It is only the first {number_of_head_rows} rows of the df.7Use this as a sample to understand the structure of the df. However, donot use this to make any calculations directly!89The complete path for the csv files are for the corresponding dataframe is:10{csv_path}11\"\"\" # noqa E5011213FUNCTIONS_WITH_MULTI_DF = \"\"\"14This is the result of `print(df.head())` for each dataframe:15{dfs_head}1617Do note that the above dfs aren't the complete df. It is only the first {number_of_head_rows} rows of the df.18Use this as a sample to understand the structure of the df. However, donot use this to make any calculations directly!1920The complete path for the csv files are for the corresponding dataframes are:21{csv_paths}22\"\"\" # noqa E5012324PREFIX_FUNCTIONS = \"\"\"25You are working with a pandas dataframe in Python. The name of the dataframe is `df`.\"\"\" # noqa E5012627MULTI_DF_PREFIX_FUNCTIONS = \"\"\"28You are working with {num_dfs} pandas dataframes in Python named df1, df2, etc.\"\"\" # noqa E5012930CSV_PREAMBLE = \"\"\"## Task And Context31You use your advanced complex reasoning capabilities to help people by answering their questions and other requests interactively. You will be asked a very wide array of requests on all kinds of topics. You will be equipped with a wide range of search engines or similar tools to help you, which you use to research your answer. You may need to use multiple tools in parallel or sequentially to complete your task. You should focus on serving the user's needs as best you can, which will be wide-ranging. The current date is {current_date}3233## Style Guide34Unless the user asks for a different style of answer, you should answer in full sentences, using proper grammar and spelling35\"\"\" # noqa E501\n```\n\nExample:\n```text\n1# Define tools that we want the csv agent to have access to234def get_file_peek_tool() -> Tool:5 def file_peek(filename: str, num_rows: int = 5) -> str:6 \"\"\"Returns the first textual contents of an uploaded file78 Args:9 table_path: the table path10 num_rows: the number of rows of the table to preview.11 \"\"\" # noqa E50112 if \".csv\" in filename:13 return pd.read_csv(filename).head(num_rows).to_markdown()14 else:15 return \"the table_path was not recognised\"1617 class file_peek_inputs(BaseModel):18 filename: str = Field(19 description=\"The name of the attached file to show a peek preview.\"20 )2122 file_peek_tool = Tool(23 name=\"file_peek\",24 description=\"The name of the attached file to show a peek preview.\", # noqa E50125 func=file_peek,26 args_schema=file_peek_inputs,27 )2829 return file_peek_tool303132def get_file_read_tool() -> Tool:33 def file_read(filename: str) -> str:34 \"\"\"Returns the textual contents of an uploaded file, broken up in text chunks3536 Args:37 filename (str): The name of the attached file to read.38 \"\"\" # noqa E50139 if \".csv\" in filename:40 return pd.read_csv(filename).to_markdown()41 else:42 return \"the table_path was not recognised\"4344 class file_read_inputs(BaseModel):45 filename: str = Field(46 description=\"The name of the attached file to read.\"47 )4849 file_read_tool = Tool(50 name=\"file_read\",51 description=\"Returns the textual contents of an uploaded file, broken up in text chunks\", # noqa E50152 func=file_read,53 args_schema=file_read_inputs,54 )5556 return file_read_tool575859def get_python_tool() -> Tool:60 \"\"\"Returns a tool that will execute python code and return the output.\"\"\"6162 def python_interpreter(code: str) -> str:63 \"\"\"A function that will return the output of the python code.6465 Args:66 code: the python code to run.67 \"\"\"68 return python_repl.run(code)6970 python_repl = PythonAstREPLTool()71 python_tool = Tool(72 name=\"python_interpreter\",73 description=\"Executes python code and returns the result. The code runs in a static sandbox without interactive mode, so print output or save output to a file.\", # noqa E50174 func=python_interpreter,75 )7677 class PythonToolInput(BaseModel):78 code: str = Field(description=\"Python code to execute.\")7980 python_tool.args_schema = PythonToolInput81 return python_tool\n```\n\nExample:\n```text\n1def create_prompt(2 system_message: Optional[BaseMessage] = SystemMessage(3 content=\"You are a helpful AI assistant.\"4 ),5 extra_prompt_messages: Optional[6 List[BaseMessagePromptTemplate]7 ] = None,8) -> ChatPromptTemplate:9 \"\"\"Create prompt for this agent.1011 Args:12 system_message: Message to use as the system message that will be the13 first in the prompt.14 extra_prompt_messages: Prompt messages that will be placed between the15 system message and the new human input.1617 Returns:18 A prompt template to pass into this agent.19 \"\"\"20 _prompts = extra_prompt_messages or []21 messages: List[Union[BaseMessagePromptTemplate, BaseMessage]]22 if system_message:23 messages = [system_message]24 else:25 messages = []2627 messages.extend(28 [29 *_prompts,30 HumanMessagePromptTemplate.from_template(\"{input}\"),31 MessagesPlaceholder(variable_name=\"agent_scratchpad\"),32 ]33 )34 return ChatPromptTemplate(messages=messages)353637def _get_csv_head_str(path: str, number_of_head_rows: int) -> str:38 with open(path, \"r\") as file:39 lines = []40 for _ in range(number_of_head_rows):41 lines.append(file.readline().strip(\"\\n\"))42 # validate that the head contents are well formatted csv4344 return \" \".join(lines)454647def _get_prompt(48 path: Union[str, List[str]], number_of_head_rows: int49) -> ChatPromptTemplate:50 if isinstance(path, str):51 lines = _get_csv_head_str(path, number_of_head_rows)52 prompt_message = f\"The user uploaded the following attachments:\\nFilename: {path}\\nWord Count: {count_words_in_file(path)}\\nPreview: {lines}\" # noqa: E5015354 elif isinstance(path, list):55 prompt_messages = []56 for file_path in path:57 lines = _get_csv_head_str(file_path, number_of_head_rows)58 prompt_messages.append(59 f\"The user uploaded the following attachments:\\nFilename: {file_path}\\nWord Count: {count_words_in_file(file_path)}\\nPreview: {lines}\" # noqa: E50160 )61 prompt_message = \" \".join(prompt_messages)6263 prompt = create_prompt(64 system_message=HumanMessage(prompt_message)65 )66 return prompt676869def count_words_in_file(file_path: str) -> int:70 try:71 with open(file_path, \"r\") as file:72 content = file.readlines()73 words = [len(sentence.split()) for sentence in content]74 return sum(words)75 except FileNotFoundError:76 print(\"File not found.\")77 return 078 except Exception as e:79 print(\"An error occurred:\", str(e))80 return 0\n```\n\nExample:\n```text\n1# Build the agent abstraction itself2def create_csv_agent(3 llm: BaseLanguageModel,4 path: Union[str, List[str]],5 extra_tools: List[BaseTool] = [],6 pandas_kwargs: Optional[dict] = None,7 prompt: Optional[ChatPromptTemplate] = None,8 number_of_head_rows: int = 5,9 verbose: bool = True,10 return_intermediate_steps: bool = True,11) -> AgentExecutor:12 \"\"\"Create csv agent with the specified language model.1314 Args:15 llm: Language model to use for the agent.16 path: A string path, or a list of string paths17 that can be read in as pandas DataFrames with pd.read_csv().18 number_of_head_rows: Number of rows to display in the prompt for sample data19 include_df_in_prompt: Display the DataFrame sample values in the prompt.20 pandas_kwargs: Named arguments to pass to pd.read_csv().21 prefix: Prompt prefix string.22 suffix: Prompt suffix string.23 prompt: Prompt to use for the agent. This takes precedence over the other prompt arguments, such as suffix and prefix.24 temp_path_dir: Temporary directory to store the csv files in for the python repl.25 delete_temp_path: Whether to delete the temporary directory after the agent is done. This only works if temp_path_dir is not provided.2627 Returns:28 An AgentExecutor with the specified agent_type agent and access to29 a PythonREPL and any user-provided extra_tools.3031 Example:32 .. code-block:: python3334 from langchain_cohere import ChatCohere, create_csv_agent3536 llm = ChatCohere(model=\"command-a-03-2025\", temperature=0)37 agent_executor = create_csv_agent(38 llm,39 \"titanic.csv\"40 )41 resp = agent_executor.invoke({\"input\":\"How many people were on the titanic?\"})42 print(resp.get(\"output\"))43 \"\"\" # noqa: E50144 try:45 import pandas as pd46 except ImportError:47 raise ImportError(48 \"pandas package not found, please install with `pip install pandas`.\"49 )5051 _kwargs = pandas_kwargs or {}52 if isinstance(path, (str)):53 df = pd.read_csv(path, **_kwargs)5455 elif isinstance(path, list):56 df = []57 for item in path:58 if not isinstance(item, (str, IOBase)):59 raise ValueError(60 f\"Expected str or file-like object, got {type(path)}\"61 )62 df.append(pd.read_csv(item, **_kwargs))63 else:64 raise ValueError(65 f\"Expected str, list, or file-like object, got {type(path)}\"66 )6768 if not prompt:69 prompt = _get_prompt(path, number_of_head_rows)7071 final_tools = [72 get_file_read_tool(),73 get_file_peek_tool(),74 get_python_tool(),75 ] + extra_tools76 if \"preamble\" in llm.__dict__ and not llm.__dict__.get(77 \"preamble\"78 ):79 llm = ChatCohere(**llm.__dict__)80 llm.preamble = CSV_PREAMBLE.format(81 current_date=datetime.now().strftime(82 \"%A, %B %d, %Y %H:%M:%S\"83 )84 )8586 agent = create_tool_calling_agent(87 llm=llm, tools=final_tools, prompt=prompt88 )89 agent_executor = AgentExecutor(90 agent=agent,91 tools=final_tools,92 verbose=verbose,93 return_intermediate_steps=return_intermediate_steps,94 )95 return agent_executor\n```\n\nExample:\n```text\n1import csv23# Data to be written to the CSV file4data = [5 [\"movie\", \"name\", \"num_tickets\"],6 [\"The Shawshank Redemption\", \"John\", 2],7 [\"The Shawshank Redemption\", \"Jerry\", 2],8 [\"The Shawshank Redemption\", \"Jack\", 4],9 [\"The Shawshank Redemption\", \"Jeremy\", 2],10 [\"Finding Nemo\", \"Darren\", 3],11 [\"Finding Nemo\", \"Jones\", 2],12 [\"Finding Nemo\", \"King\", 1],13 [\"Finding Nemo\", \"Penelope\", 5],14]1516file_path = \"movies_tickets.csv\"1718with open(file_path, \"w\", newline=\"\") as file:19 writer = csv.writer(file)20 writer.writerows(data)2122print(f\"CSV file created successfully at {file_path}.\")\n```\n\nExample:\n```text\n1# Try out an example2llm = ChatCohere(model=\"command-a-03-2025\", temperature=0)3agent_executor = create_csv_agent(llm, \"movies_tickets.csv\")4resp = agent_executor.invoke(5 {\"input\": \"Who all watched Shawshank redemption?\"}6)7print(resp.get(\"output\"))\n```\n\nExample:\n```text\nJohn, Jerry, Jack and Jeremy watched Shawshank Redemption.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.380Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":10,"totalLines":53,"estimatedTokens":3284}}205{"id":"doc-pdf_extractor_with_native_multi_step_tool_use_co-225e756a","source":"documentation","title":"PDF Extractor with Native Multi Step Tool Use | Cohere","url":"https://docs.cohere.com/page/pdf-extractor","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import os23import cohere4import pandas as pd5import json6from unstructured.partition.pdf import partition_pdf\n```\n\nExample:\n```text\n1# uncomment to install dependencies2# !pip install cohere unstructured\n```\n\nExample:\n```text\n1# versions2print('cohere version:', cohere.__version__)\n```\n\nExample:\n```text\ncohere version: 5.5.1\n```\n\nExample:\n```text\n1COHERE_API_KEY = os.environ.get(\"CO_API_KEY\")2COHERE_MODEL = 'command-a-03-2025'3co = cohere.Client(api_key=COHERE_API_KEY)\n```\n\nExample:\n```text\n1def convert_to_json(text: str) -> dict:2 \"\"\"3 Given text files, convert to json object and saves to csv.45 Args:6 text (str): The text to extract information from.78 Returns:9 dict: A dictionary containing the result of the conversion process.10 \"\"\"1112 MANDATORY_FIELDS = [13 \"total_amount\",14 \"invoice_number\",15 ]1617 message = \"\"\"# Instruction18 Given the text, convert to json object with the following keys:19 total_amount, invoice_number2021 # Output format json:22 {{23 \"total_amount\": \"<extracted amount=\"\" invoice=\"\" total=\"\">\",24 \"invoice_number\": \"<extracted invoice=\"\" number=\"\">\",25 }}2627 Do not output code blocks.2829 # Extracted PDF30 {text}31 \"\"\"3233 result = co.chat(34 message=message.format(text=text), model=COHERE_MODEL, preamble=None35 ).text3637 try:38 result = json.loads(result)39 # check if all keys are present40 if not all(i in result.keys() for i in MANDATORY_FIELDS):41 return {\"result\": f\"ERROR: Keys are missing. Please check your result {result}\"}4243 df = pd.DataFrame(result, index=[0])44 df.to_csv(\"output.csv\", index=False)45 return {\"result\": \"SUCCESS. All steps have been completed.\"}4647 except Exception as e:48 return {\"result\": f\"ERROR: Could not load the result as json. Please check the result: {result} and ERROR: {e}\"}\n```\n\nExample:\n```text\n1def cohere_agent(2 message: str,3 preamble: str,4 verbose: bool = False,5) -> str:6 \"\"\"7 Function to handle multi-step tool use api.89 Args:10 message (str): The message to send to the Cohere AI model.11 preamble (str): The preamble or context for the conversation.12 verbose (bool, optional): Whether to print verbose output. Defaults to False.1314 Returns:15 str: The final response from the call.16 \"\"\"1718 functions_map = {19 \"convert_to_json\": convert_to_json,20 }2122 tools = [23 {24 \"name\": \"convert_to_json\",25 \"description\": \"Given a text, convert it to json object.\",26 \"parameter_definitions\": {27 \"text\": {28 \"description\": \"text to be converted into json\",29 \"type\": \"str\",30 \"required\": True,31 },32 },33 }34 ]3536 counter = 13738 response = co.chat(39 model=COHERE_MODEL,40 message=message,41 preamble=preamble,42 tools=tools,43 )4445 if verbose:46 print(f\"\\nrunning step 0\")47 print(response.text)4849 while response.tool_calls:50 tool_results = []5152 if verbose:53 print(f\"\\nrunning step {counter}\")54 for tool_call in response.tool_calls:55 print(\"tool_call.parameters:\", tool_call.parameters)56 if tool_call.parameters:57 output = functions_map[tool_call.name](**tool_call.parameters)58 else:59 output = functions_map[tool_call.name]()6061 outputs = [output]62 tool_results.append({\"call\": tool_call, \"outputs\": outputs})6364 if verbose:65 print(66 f\"= running tool {tool_call.name}, with parameters: {tool_call.parameters}\"67 )68 print(f\"== tool results: {outputs}\")6970 response = co.chat(71 model=COHERE_MODEL,72 message=\"\",73 chat_history=response.chat_history,74 preamble=preamble,75 tools=tools,76 tool_results=tool_results,77 )7879 if verbose:80 print(response.text)81 counter += 18283 return response.text\n```\n\nExample:\n```text\n1def extract_pdf(path):2 \"\"\"3 Function to extract text from a PDF file.4 \"\"\"5 elements = partition_pdf(path)6 return \"\\n\".join([str(el) for el in elements])789def pdf_extractor(pdf_path):10 \"\"\"11 Main function that extracts pdf and calls the cohere agent.12 \"\"\"13 pdf_text = extract_pdf(pdf_path)1415 prompt = f\"\"\"16 # Instruction17 You are expert at extracting invoices from PDF. The text of the PDF file is given below.1819 You must follow the steps below:20 1. Summarize the text and extract only the most information: total amount billed and invoice number.21 2. Using the summary above, call convert_to_json tool, which uses the summary from step 1.22 If you run into issues. Identifiy the issue and retry.23 You are not done unless you see SUCCESS in the tool output.2425 # File Name:26 {pdf_path}2728 # Extracted Text:29 {pdf_text}30 \"\"\"31 output = cohere_agent(prompt, None, verbose=True)32 print(f\"Finished extracting: {pdf_path}\")3334 print('Please check the output below')35 print(pd.read_csv('output.csv'))363738pdf_extractor('simple_invoice.pdf')\n```\n\nExample:\n```text\nrunning step 0I will summarise the text and then use the convert_to_json tool to format the summary.running step 1tool_call.parameters: {'text': 'Total amount billed: $115.00\\nInvoice number: 0852'}= running tool convert_to_json, with parameters: {'text': 'Total amount billed: $115.00\\nInvoice number: 0852'}== tool results: [{'result': 'SUCCESS. All steps have been completed.'}]SUCCESS.Finished extracting: simple_invoice.pdfPlease check the output below total_amount invoice_number0 $115.00 852\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.380Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":9,"totalLines":48,"estimatedTokens":1545}}206{"id":"doc-deploy_your_finetuned_model_on_aws_marketplace_c-d21052d7","source":"documentation","title":"Deploy your finetuned model on AWS Marketplace | Cohere","url":"https://docs.cohere.com/page/deploy-finetuned-model-aws-marketplace","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1pip install \"cohere>=5.11.0\"\n```\n\nExample:\n```text\n1import cohere2import os3import sagemaker as sage45from sagemaker.s3 import S3Uploader\n```\n\nExample:\n```text\n1# Change \"<aws_profile>\" to your own AWS profile name2os.environ[\"AWS_PROFILE\"] = \"<aws_profile>\"\n```\n\nExample:\n```text\n1# The AWS region2region = \"<region>\"34# Get the arn of the bring your own finetuning algorithm by region5cohere_package = (6 \"cohere-command-r-v2-byoft-8370167e649c32a1a5f00267cd334c2c\"7)8algorithm_map = {9 \"us-east-1\": f\"arn:aws:sagemaker:us-east-1:865070037744:algorithm/{cohere_package}\",10 \"us-east-2\": f\"arn:aws:sagemaker:us-east-2:057799348421:algorithm/{cohere_package}\",11 \"us-west-2\": f\"arn:aws:sagemaker:us-west-2:594846645681:algorithm/{cohere_package}\",12 \"eu-central-1\": f\"arn:aws:sagemaker:eu-central-1:446921602837:algorithm/{cohere_package}\",13 \"ap-southeast-1\": f\"arn:aws:sagemaker:ap-southeast-1:192199979996:algorithm/{cohere_package}\",14 \"ap-southeast-2\": f\"arn:aws:sagemaker:ap-southeast-2:666831318237:algorithm/{cohere_package}\",15 \"ap-northeast-1\": f\"arn:aws:sagemaker:ap-northeast-1:977537786026:algorithm/{cohere_package}\",16 \"ap-south-1\": f\"arn:aws:sagemaker:ap-south-1:077584701553:algorithm/{cohere_package}\",17}18if region not in algorithm_map:19 raise Exception(f\"Current region {region} is not supported.\")20arn = algorithm_map[region]2122# The local directory of your adapter weights. No need to specify this, if you bring your own merged weights23adapter_weights_dir = \"<adapter_weights_dir>\"2425# The local directory you want to save the merged weights. Or the local directory of your own merged weights, if you bring your own merged weights26merged_weights_dir = \"<merged_weights_dir>\"2728# The S3 directory you want to save the merged weights29s3_checkpoint_dir = \"<s3_checkpoint_dir>\"3031# The S3 directory you want to save the exported TensorRT-LLM engine. Make sure you do not reuse the same S3 directory across multiple runs32s3_output_dir = \"<s3_output_dir>\"3334# The name of the export35export_name = \"<export_name>\"3637# The name of the SageMaker endpoint38endpoint_name = \"<endpoint_name>\"3940# The instance type for export and inference. Now \"ml.p4de.24xlarge\" and \"ml.p5.48xlarge\" are supported41instance_type = \"<instance_type>\"\n```\n\nExample:\n```text\n1import torch23from peft import PeftModel4from transformers import CohereForCausalLM567def load_and_merge_model(8 base_model_name_or_path: str, adapter_weights_dir: str9):10 \"\"\"11 Load the base model and the model finetuned by PEFT, and merge the adapter weights to the base weights to get a model with merged weights12 \"\"\"13 base_model = CohereForCausalLM.from_pretrained(14 base_model_name_or_path15 )16 peft_model = PeftModel.from_pretrained(17 base_model, adapter_weights_dir18 )19 merged_model = peft_model.merge_and_unload()20 return merged_model212223def save_hf_model(output_dir: str, model, tokenizer=None, args=None):24 \"\"\"25 Save a HuggingFace model (and optionally tokenizer as well as additional args) to a local directory26 \"\"\"27 os.makedirs(output_dir, exist_ok=True)28 model.save_pretrained(29 output_dir, state_dict=None, safe_serialization=True30 )31 if tokenizer is not None:32 tokenizer.save_pretrained(output_dir)33 if args is not None:34 torch.save(35 args, os.path.join(output_dir, \"training_args.bin\")36 )\n```\n\nExample:\n```text\n1# Get the merged model from adapter weights2merged_model = load_and_merge_model(3 \"CohereForAI/c4ai-command-r-08-2024\", adapter_weights_dir4)56# Save the merged weights to your local directory7save_hf_model(merged_weights_dir, merged_model)\n```\n\nExample:\n```text\n1sess = sage.Session()2merged_weights = S3Uploader.upload(3 merged_weights_dir, s3_checkpoint_dir, sagemaker_session=sess4)5print(\"merged_weights\", merged_weights)\n```\n\nExample:\n```text\n1co = cohere.SagemakerClient(aws_region=region)2co.sagemaker_finetuning.export_finetune(3 arn=arn,4 name=export_name,5 s3_checkpoint_dir=s3_checkpoint_dir,6 s3_output_dir=s3_output_dir,7 instance_type=instance_type,8 role=\"ServiceRoleSagemaker\",9)\n```\n\nExample:\n```text\n1co.sagemaker_finetuning.create_endpoint(2 arn=arn,3 endpoint_name=endpoint_name,4 s3_models_dir=s3_output_dir,5 recreate=True,6 instance_type=instance_type,7 role=\"ServiceRoleSagemaker\",8)\n```\n\nExample:\n```text\n1# If the endpoint is already deployed, you can directly connect to it2co.sagemaker_finetuning.connect_to_endpoint(3 endpoint_name=endpoint_name4)56message = \"Classify the following text as either very negative, negative, neutral, positive or very positive: mr. deeds is , as comedy goes , very silly -- and in the best way.\"7result = co.sagemaker_finetuning.chat(message=message)8print(result)\n```\n\nExample:\n```text\n1import json2from tqdm import tqdm34eval_data_path = \"<path_to_scienceQA_eval.jsonl>\"56total = 07correct = 08for line in tqdm(open(eval_data_path).readlines()):9 total += 110 question_answer_json = json.loads(line)11 question = question_answer_json[\"messages\"][0][\"content\"]12 answer = question_answer_json[\"messages\"][1][\"content\"]13 model_ans = co.sagemaker_finetuning.chat(14 message=question, temperature=015 ).text16 if model_ans == answer:17 correct += 11819print(f\"Accuracy of finetuned model is %.3f\" % (correct / total))\n```\n\nExample:\n```text\n1co.sagemaker_finetuning.delete_endpoint()2co.sagemaker_finetuning.close()\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.381Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":1433}}207{"id":"doc-finetuning_cohere_models_on_aws_sagemaker_cohere-ae3db06b","source":"documentation","title":"Finetuning Cohere Models on AWS Sagemaker | Cohere","url":"https://docs.cohere.com/page/finetune-on-sagemaker","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1pip install \"cohere>=5.11.0\"\n```\n\nExample:\n```text\n1import cohere2import boto33import sagemaker as sage4from sagemaker.s3 import S3Uploader\n```\n\nExample:\n```text\n1region = boto3.Session().region_name23cohere_package = \"\"4# cohere_package = \"cohere-command-r-ft-v-0-1-2-bae2282f0f4a30bca8bc6fea9efeb7ca\"56# Mapping for algorithms7algorithm_map = {8 \"us-east-1\": f\"arn:aws:sagemaker:us-east-1:865070037744:algorithm/{cohere_package}\",9 \"us-east-2\": f\"arn:aws:sagemaker:us-east-2:057799348421:algorithm/{cohere_package}\",10 \"us-west-2\": f\"arn:aws:sagemaker:us-west-2:594846645681:algorithm/{cohere_package}\",11 \"eu-central-1\": f\"arn:aws:sagemaker:eu-central-1:446921602837:algorithm/{cohere_package}\",12 \"ap-southeast-1\": f\"arn:aws:sagemaker:ap-southeast-1:192199979996:algorithm/{cohere_package}\",13 \"ap-southeast-2\": f\"arn:aws:sagemaker:ap-southeast-2:666831318237:algorithm/{cohere_package}\",14 \"ap-northeast-1\": f\"arn:aws:sagemaker:ap-northeast-1:977537786026:algorithm/{cohere_package}\",15 \"ap-south-1\": f\"arn:aws:sagemaker:ap-south-1:077584701553:algorithm/{cohere_package}\",16}17if region not in algorithm_map.keys():18 raise Exception(19 f\"Current boto3 session region {region} is not supported.\"20 )2122arn = algorithm_map[region]\n```\n\nExample:\n```text\n1s3_data_dir = \"s3://...\" # Do not add a trailing slash otherwise the upload will not work\n```\n\nExample:\n```text\n{ \"messages\": [ { \"role\": \"System\", \"content\": \"You are a chatbot trained to answer to my every question.\" }, { \"role\": \"User\", \"content\": \"Hello\" }, { \"role\": \"Chatbot\", \"content\": \"Greetings! How can I help you?\" }, { \"role\": \"User\", \"content\": \"What makes a good running route?\" }, { \"role\": \"Chatbot\", \"content\": \"A sidewalk-lined road is ideal so that you\\u2019re up and off the road away from vehicular traffic.\" } ]}\n```\n\nExample:\n```text\n1sess = sage.Session()2# TODO[Optional]: change it to your data3# You can download following example datasets from https://github.com/cohere-ai/cohere-developer-experience/tree/main/notebooks/data and upload them4# to the root of this juyter notebook5train_dataset = S3Uploader.upload(6 \"./scienceQA_train.jsonl\", s3_data_dir, sagemaker_session=sess7)8# optional eval dataset9eval_dataset = S3Uploader.upload(10 \"./scienceQA_eval.jsonl\", s3_data_dir, sagemaker_session=sess11)12print(\"traint_dataset\", train_dataset)13print(\"eval_dataset\", eval_dataset)\n```\n\nExample:\n```text\n1# TODO update this with a custom S3 path2# DO NOT add a trailing slash at the end3s3_models_dir = f\"s3://...\"\n```\n\nExample:\n```text\n1co = cohere.SagemakerClient(region_name=region)\n```\n\nExample:\n```text\n1# Example of how to pass hyperparameters to the fine-tuning job2train_parameters = {3 \"train_epochs\": 1,4 \"early_stopping_patience\": 2,5 \"early_stopping_threshold\": 0.001,6 \"learning_rate\": 0.01,7 \"train_batch_size\": 16,8}\n```\n\nExample:\n```text\n1finetune_name = \"test-finetune\"2co.sagemaker_finetuning.create_finetune(3 arn=arn,4 name=finetune_name,5 train_data=train_dataset,6 eval_data=eval_dataset,7 s3_models_dir=s3_models_dir,8 instance_type=\"ml.p4de.24xlarge\",9 training_parameters=train_parameters,10 role=\"ServiceRoleSagemaker\",11)\n```\n\nExample:\n```text\n1endpoint_name = \"test-finetune\"2co.sagemaker_finetuning.create_endpoint(3 arn=arn,4 endpoint_name=endpoint_name,5 s3_models_dir=s3_models_dir,6 recreate=True,7 instance_type=\"ml.p4de.24xlarge\",8 role=\"ServiceRoleSagemaker\",9)1011# If the endpoint is already created, you just need to connect to it12co.connect_to_endpoint(endpoint_name=endpoint_name)\n```\n\nExample:\n```text\n1message = \"Classify the following text as either very negative, negative, neutral, positive or very positive: mr. deeds is , as comedy goes , very silly -- and in the best way.\"23result = co.sagemaker_finetuning.chat(message=message)4print(result)\n```\n\nExample:\n```text\n1import json2from tqdm import tqdm34total = 05correct = 06for line in tqdm(7 open(\"./sample_finetune_scienceQA_eval.jsonl\").readlines()8):9 total += 110 question_answer_json = json.loads(line)11 question = question_answer_json[\"messages\"][0][\"content\"]12 answer = question_answer_json[\"messages\"][1][\"content\"]13 model_ans = co.sagemaker_finetuning.chat(14 message=question, temperature=015 ).text16 if model_ans == answer:17 correct += 11819print(f\"Accuracy of finetuned model is %.3f\" % (correct / total))\n```\n\nExample:\n```text\n1co.delete_endpoint()2co.close()\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.382Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":14,"totalLines":73,"estimatedTokens":1199}}208{"id":"doc-topic_modeling_system_for_ai_papers_cohere-236bdd34","source":"documentation","title":"Topic Modeling System for AI Papers | Cohere","url":"https://docs.cohere.com/page/topic-modeling-ai-papers","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1pip install requests beautifulsoup4 cohere altair clean-text numpy pandas scikit-learn > /dev/null\n```\n\nExample:\n```text\n1import cohere23api_key = '<api_key>'4co = cohere.ClientV2(api_key=\"YOUR API KEY\")\n```\n\nExample:\n```text\n1## Getting the web content2import requests3from bs4 import BeautifulSoup45## Processing the content 6import pandas as pd7import numpy as np89## Handling the underlying NLP10from sklearn.decomposition import PCA11from sklearn.cluster import KMeans\n```\n\nExample:\n```text\n1URL = \"https://arxiv.org/list/cs.AI/new\"2page = requests.get(URL)\n```\n\nExample:\n```text\n1def make_raw_df(url):2 response=requests.get(url)3 soup=BeautifulSoup(response.content, \"html.parser\")45 titles=list()6 texts=list()78 # Extract titles from <div class=\"list-title mathjax\">9 title_tags = soup.find_all(class_=\"list-title mathjax\")10 for title_tag in title_tags:11 titles.append(title_tag.text.strip()) # Remove leading/trailing whitespace1213 # Extract abstracts from <p class=\"mathjax\">14 abstract_tags = soup.find_all('p', class_=\"mathjax\")#, tag=\"p\")15 for abstract_tag in abstract_tags:16 texts.append(abstract_tag.text.strip())1718 df = pd.DataFrame({\"titles\": titles, \"texts\": texts})19 return df\n```\n\nExample:\n```text\n1def get_embeddings(text,model='embed-v4.0'):2 output = co.embed(3 model=model,4 texts=[text],5 input_type=\"classification\",6 embedding_types=[\"float\"],)7 return output.embeddings.float_[0]89# Reduce embeddings to 2 principal components to aid visualization10# Function to return the principal components11def get_pc(arr,n):12 pca = PCA(n_components=n)13 embeds_transform = pca.fit_transform(arr)14 return embeds_transform1516def make_clusters(df,n_clusters):1718 # Get the embeddings for the text column19 df_clust = df.copy()20 df_clust['text_embeds'] = df_clust['texts'].apply(get_embeddings) # We've defined this function above.2122 # Convert the embeddings list to a numpy array23 embeddings_array = np.array(df_clust['text_embeds'].tolist())24 # Cluster the embeddings2526 kmeans_model = KMeans(n_clusters=n_clusters, random_state=0, n_init='auto')27 classes = kmeans_model.fit_predict(embeddings_array).tolist()28 df_clust['cluster'] = (list(map(str,classes)))2930 df_clust.columns.astype(str)31 return df_clust3233def create_cluster_names(essences_dict):34 cluster_names = {}35 for cluster_num, description in essences_dict.items():36 # Take the first sentence and limit to first N characters37 short_name = description.split('.')[0][:30].strip() + '...'38 cluster_names[cluster_num] = short_name39 return cluster_names\n```\n\nExample:\n```text\n1def get_essence(df):23 clusters = sorted(df['cluster'].unique())4 cluster_descriptions = {}56 for cluster_num in clusters:7 8 cluster_df = df[df['cluster'] == cluster_num]9 # Combine titles and texts10 titles = ' '.join(cluster_df['titles'].fillna(''))11 texts = ' '.join(cluster_df['texts'].fillna(''))12 combined_text = f\"Titles: {titles}\\n\\nTexts: {texts}\"1314 system_message = \"\"\"15 ## Task & Context16 You are a world-class language model that's especially good at capturing the essence of complex text.1718 ## Style Guide19 Unless the user asks for a different style of answer, you should answer in concise text with proper grammar and spelling.20 \"\"\"2122 message=f\"\"\"Based on the following titles and texts from academic papers, provide 3-4 words that describe what this category of papers is about. Think of this like a word cloud.23 Focus on the main theme or topic that unifies these papers.24 Please do not use the words 'AI', 'Artificial Intelligence,' 'Machine Learning,' or 'ML' in your response.25 26 {combined_text}27 28 Description:\"\"\"2930 messages = [31 {\"role\": \"system\", \"content\": system_message},32 {\"role\": \"user\", \"content\": message},33 ]3435 essence = co.chat(36 model=\"command-a-03-2025,37 messages=messages38 )3940 description = essence.message.content[0].text.strip() + \".\"41 cluster_descriptions[cluster_num] = description 4243 return cluster_descriptions\n```\n\nExample:\n```text\n1import altair as alt2# Function to generate the 2D plot3def generate_chart(df,xcol,ycol,lbl='off',color='basic',title='', cluster_names=None):45 ## IMPORTANT6 ## We're using this function to create the 'x' and 'y' columns for the chart.7 ## We don't actually use the principal components anywhere else in the code.8 embeds = np.array(df['text_embeds'].tolist())9 embeds_pc2 = get_pc(embeds,2)10 # Add the principal components to dataframe11 df = pd.concat([df, pd.DataFrame(embeds_pc2)], axis=1)12 ## END IMPORTANT1314 # Add cluster names to the dataframe if provided15 if cluster_names:16 df['cluster_label'] = df['cluster'].map(cluster_names)17 else:18 df['cluster_label'] = df['cluster']19 20 # Plot the 2D embeddings on a chart21 df.columns = df.columns.astype(str)2223 if color == 'basic':24 color_encode = alt.value('#333293')25 else:26 color_encode = alt.Color('cluster_label:N',27 scale=alt.Scale(scheme='category20'),28 legend=alt.Legend(29 title=\"Topics\",30 symbolLimit=len(cluster_names) if cluster_names else None,31 orient='right',32 labelLimit=500, # Increase label width limit (default is 200)33 columns=1 # Force single column layout34 ))353637 chart = alt.Chart(df).mark_circle(size=500).encode(38 x=alt.X(xcol,39 scale=alt.Scale(zero=False),40 axis=alt.Axis(labels=False, ticks=False, domain=False)41 ),42 y=alt.Y(ycol,43 scale=alt.Scale(zero=False),44 axis=alt.Axis(labels=False, ticks=False, domain=False)45 ),46 color=color_encode,47 tooltip=['titles', 'cluster_label'] # Updated to show cluster label in tooltip48 )4950 if lbl == 'on':51 text = chart.mark_text(align='left', baseline='middle',dx=15, size=13,color='black').encode(text='title', color= alt.value('black'))52 else:53 text = chart.mark_text(align='left', baseline='middle',dx=10).encode()5455 result = (chart + text).configure(background=\"#FDF7F0\"56 ).properties(57 width=800,58 height=500,59 title=title60 ).configure_legend(61 orient='right',62 titleFontSize=18,63 labelFontSize=10,64 padding=5, # Add some padding around the legend65 offset=5, # Move legend away from chart66 labelLimit=500 # Also need to set it here67 )68 69 return result\n```\n\nExample:\n```text\n1### Creating the baseline dataframe.2df = make_raw_df(\"https://arxiv.org/list/cs.AI/new\")34### Defining our cluster number and making a 'cluster' dataframe.5n_clusters = 126df_clust = make_clusters(df,n_clusters)78### Get the topic essences and cluster names9overview = get_essence(df_clust)10cluster_names = create_cluster_names(overview)1112### Generate the chart13generate_chart(df_clust,'0','1',lbl='off',color='cluster',title=f'Clustering with {n_clusters} Clusters', cluster_names=cluster_names)\n```\n\nExample:\n```text\n1from sklearn.metrics.pairwise import cosine_similarity23def get_similarity(target,candidates):4 # Turn list into array5 candidates = np.array(candidates)6 target = np.expand_dims(np.array(target),axis=0)78 # Calculate cosine similarity9 sim = cosine_similarity(target,candidates)10 sim = np.squeeze(sim).tolist()11 sort_index = np.argsort(sim)[::-1]12 sort_score = [sim[i] for i in sort_index]13 similarity_scores = zip(sort_index,sort_score)1415 # Return similarity scores16 return similarity_scores\n```\n\nExample:\n```text\n1# Add new query2new_query = \"Anything on AI personalities?\"34# Get embeddings of the new query5new_query_embeds = get_embeddings(new_query)67embeds = np.array(df_clust['text_embeds'].tolist()) # We defined these embeddings earlier and are pulling them out now for the function.89# Get the similarity between the search query and existing queries10similarity = get_similarity(new_query_embeds, embeds)11#print(list(similarity))12# View the top 5 articles13print('Query:')14print(new_query,'\\n')1516print('Similar queries:')17for idx,sim in similarity:18 print(f'Similarity: {sim:.2f};')19 print(df.iloc[idx]['titles'])20 print(df.iloc[idx]['texts'])21 print()\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.383Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":11,"totalLines":58,"estimatedTokens":2204}}209{"id":"doc-finetuning_on_cohere_s_platform_cohere-785b157e","source":"documentation","title":"Finetuning on Cohere's Platform | Cohere","url":"https://docs.cohere.com/page/convfinqa-finetuning-wandb","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install cohere\n```\n\nExample:\n```text\n1import os2import cohere3from cohere.finetuning import (4 Hyperparameters,5 Settings,6 WandbConfig,7 FinetunedModel,8 BaseModel,9)1011# fill in your Cohere API key here12os.environ[\"COHERE_API_KEY\"] = \"<COHERE_API_KEY>\"1314# instantiate the Cohere client15co = cohere.ClientV2(os.environ[\"COHERE_API_KEY\"])\n```\n\nExample:\n```text\n1{2 \"messages\": [3 {4 \"role\": \"System\",5 \"content\": \"stock-based awards under the plan stock options 2013 marathon grants stock options under the 2007 plan and previously granted options under the 2003 plan .\\nmarathon 2019s stock options represent the right to purchase shares of common stock at the fair market value of the common stock on the date of grant .\\nthrough 2004 , certain stock options were granted under the 2003 plan with a tandem stock appreciation right , which allows the recipient to instead elect to receive cash and/or common stock equal to the excess of the fair market value of shares of common stock , as determined in accordance with the 2003 plan , over the option price of the shares .\\nin general , stock options granted under the 2007 plan and the 2003 plan vest ratably over a three-year period and have a maximum term of ten years from the date they are granted .\\nstock appreciation rights 2013 prior to 2005 , marathon granted sars under the 2003 plan .\\nno stock appreciation rights have been granted under the 2007 plan .\\nsimilar to stock options , stock appreciation rights represent the right to receive a payment equal to the excess of the fair market value of shares of common stock on the date the right is exercised over the grant price .\\nunder the 2003 plan , certain sars were granted as stock-settled sars and others were granted in tandem with stock options .\\nin general , sars granted under the 2003 plan vest ratably over a three-year period and have a maximum term of ten years from the date they are granted .\\nstock-based performance awards 2013 prior to 2005 , marathon granted stock-based performance awards under the 2003 plan .\\nno stock-based performance awards have been granted under the 2007 plan .\\nbeginning in 2005 , marathon discontinued granting stock-based performance awards and instead now grants cash-settled performance units to officers .\\nall stock-based performance awards granted under the 2003 plan have either vested or been forfeited .\\nas a result , there are no outstanding stock-based performance awards .\\nrestricted stock 2013 marathon grants restricted stock and restricted stock units under the 2007 plan and previously granted such awards under the 2003 plan .\\nin 2005 , the compensation committee began granting time-based restricted stock to certain u.s.-based officers of marathon and its consolidated subsidiaries as part of their annual long-term incentive package .\\nthe restricted stock awards to officers vest three years from the date of grant , contingent on the recipient 2019s continued employment .\\nmarathon also grants restricted stock to certain non-officer employees and restricted stock units to certain international employees ( 201crestricted stock awards 201d ) , based on their performance within certain guidelines and for retention purposes .\\nthe restricted stock awards to non-officers generally vest in one-third increments over a three-year period , contingent on the recipient 2019s continued employment .\\nprior to vesting , all restricted stock recipients have the right to vote such stock and receive dividends thereon .\\nthe non-vested shares are not transferable and are held by marathon 2019s transfer agent .\\ncommon stock units 2013 marathon maintains an equity compensation program for its non-employee directors under the 2007 plan and previously maintained such a program under the 2003 plan .\\nall non-employee directors other than the chairman receive annual grants of common stock units , and they are required to hold those units until they leave the board of directors .\\nwhen dividends are paid on marathon common stock , directors receive dividend equivalents in the form of additional common stock units .\\nstock-based compensation expense 2013 total employee stock-based compensation expense was $ 80 million , $ 83 million and $ 111 million in 2007 , 2006 and 2005 .\\nthe total related income tax benefits were $ 29 million , $ 31 million and $ 39 million .\\nin 2007 and 2006 , cash received upon exercise of stock option awards was $ 27 million and $ 50 million .\\ntax benefits realized for deductions during 2007 and 2006 that were in excess of the stock-based compensation expense recorded for options exercised and other stock-based awards vested during the period totaled $ 30 million and $ 36 million .\\ncash settlements of stock option awards totaled $ 1 million and $ 3 million in 2007 and 2006 .\\nstock option awards granted 2013 during 2007 , 2006 and 2005 , marathon granted stock option awards to both officer and non-officer employees .\\nthe weighted average grant date fair value of these awards was based on the following black-scholes assumptions: .\\nThe weighted average exercise price per share of 2007, 2006, 2005 are $ 60.94, $ 37.84, $ 25.14. The expected annual dividends per share of 2007, 2006, 2005 are $ 0.96, $ 0.80, $ 0.66. The expected life in years of 2007, 2006, 2005 are 5.0, 5.1, 5.5. The expected volatility of 2007, 2006, 2005 are 27% ( 27 % ), 28% ( 28 % ), 28% ( 28 % ). The risk-free interest rate of 2007, 2006, 2005 are 4.1% ( 4.1 % ), 5.0% ( 5.0 % ), 3.8% ( 3.8 % ). The weighted average grant date fair value of stock option awards granted of 2007, 2006, 2005 are $ 17.24, $ 10.19, $ 6.15.\\n.\",6 },7 {8 \"role\": \"User\",9 \"content\": \"what was the weighted average exercise price per share in 2007?\",10 },11 {\"role\": \"Chatbot\", \"content\": \"60.94\"},12 {\"role\": \"User\", \"content\": \"and what was it in 2005?\"},13 {\"role\": \"Chatbot\", \"content\": \"25.14\"},14 {15 \"role\": \"User\",16 \"content\": \"what was, then, the change over the years?\",17 },18 {\"role\": \"Chatbot\", \"content\": \"subtract(60.94, 25.14)\"},19 {20 \"role\": \"User\",21 \"content\": \"what was the weighted average exercise price per share in 2005?\",22 },23 {\"role\": \"Chatbot\", \"content\": \"25.14\"},24 {25 \"role\": \"User\",26 \"content\": \"and how much does that change represent in relation to this 2005 weighted average exercise price?\",27 },28 {29 \"role\": \"Chatbot\",30 \"content\": \"subtract(60.94, 25.14), divide(#0, 25.14)\",31 },32 ]33}\n```\n\nExample:\n```text\n1chat_dataset = co.datasets.create(2 name=\"cfqa-ft-dataset\",3 data=open(\"data/convfinqa-train-chat.jsonl\", \"rb\"),4 eval_data=open(\"data/convfinqa-eval-chat.jsonl\", \"rb\"),5 type=\"chat-finetune-input\",6)7print(8 chat_dataset.id9) # we will use this id to refer to the dataset when creating a finetuning job\n```\n\nExample:\n```text\n1co.wait(2 chat_dataset3) # wait for the dataset to be processed and validated\n```\n\nExample:\n```text\n1hp_config = Hyperparameters(2 train_batch_size=16,3 train_epochs=1,4 learning_rate=0.0001,5)\n```\n\nExample:\n```text\n1wnb_config = WandbConfig(2 project=\"test-project\",3 api_key=\"<wandb_api_key>\",4 entity=\"test-entity\", # must be a valid enitity associated with the provided API key5)\n```\n\nExample:\n```text\n1cfqa_finetune = co.finetuning.create_finetuned_model(2 request=FinetunedModel(3 name=\"cfqa-command-r-ft\",4 settings=Settings(5 base_model=BaseModel(6 base_type=\"BASE_TYPE_CHAT\", # specifies this is a chat finetuning7 ),8 dataset_id=chat_dataset.id, # the id of the dataset we created above9 hyperparameters=hp_config,10 wandb=wnb_config,11 ),12 ),13)14print(15 cfqa_finetune.finetuned_model.id16) # we will use this id to refer to the finetuned model when making predictions/getting status/etc.\n```\n\nExample:\n```text\n1response = co.finetuning.get_finetuned_model(2 cfqa_finetune.finetuned_model.id3)4print(5 response.finetuned_model.status6) # when the job finished this will be STATUS_READY\n```\n\nExample:\n```text\n1train_step_metrics = co.finetuning.list_training_step_metrics(2 finetuned_model_id=cfqa_finetune.finetuned_model.id3)45for metric in train_step_metrics.step_metrics:6 print(metric.metrics)\n```\n\nExample:\n```text\n1response = co.chat(2 model=cfqa_finetune.finetuned_model.id + \"-ft\",3 messages=[4 {5 \"role\": \"system\",6 \"content\": \"in the ordinary course of business , based on our evaluations of certain geologic trends and prospective economics , we have allowed certain lease acreage to expire and may allow additional acreage to expire in the future .\\nif production is not established or we take no other action to extend the terms of the leases , licenses or concessions , undeveloped acreage listed in the table below will expire over the next three years .\\nwe plan to continue the terms of certain of these licenses and concession areas or retain leases through operational or administrative actions ; however , the majority of the undeveloped acres associated with other africa as listed in the table below pertains to our licenses in ethiopia and kenya , for which we executed agreements in 2015 to sell .\\nthe kenya transaction closed in february 2016 and the ethiopia transaction is expected to close in the first quarter of 2016 .\\nsee item 8 .\\nfinancial statements and supplementary data - note 5 to the consolidated financial statements for additional information about this disposition .\\nnet undeveloped acres expiring year ended december 31 .\\nThe u.s . of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 68, 89, 128. The e.g . of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 2014, 92, 36. The other africa of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 189, 4352, 854. The total africa of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 189, 4444, 890. The other international of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 2014, 2014, 2014. The total of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 257, 4533, 1018.\\n.\",7 },8 {9 \"role\": \"user\",10 \"content\": \"what percentage of undeveloped acres were in the us in 2018?\",11 },12 {13 \"role\": \"assistant\",14 \"content\": \"divide(128, 1018)\",15 },16 {17 \"role\": \"user\",18 \"content\": \"what was the total african and us net undeveloped acres expiring in 2016?\",19 },20 ],21)22print(\"#### Model response ####\")23print(response.text)24print(\"########################\")\n```\n\nExample:\n```text\n#### Model response ####add(189, 68)########################\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-r-08-24\",3 messages=[4 {5 \"role\": \"system\",6 \"content\": \"in the ordinary course of business , based on our evaluations of certain geologic trends and prospective economics , we have allowed certain lease acreage to expire and may allow additional acreage to expire in the future .\\nif production is not established or we take no other action to extend the terms of the leases , licenses or concessions , undeveloped acreage listed in the table below will expire over the next three years .\\nwe plan to continue the terms of certain of these licenses and concession areas or retain leases through operational or administrative actions ; however , the majority of the undeveloped acres associated with other africa as listed in the table below pertains to our licenses in ethiopia and kenya , for which we executed agreements in 2015 to sell .\\nthe kenya transaction closed in february 2016 and the ethiopia transaction is expected to close in the first quarter of 2016 .\\nsee item 8 .\\nfinancial statements and supplementary data - note 5 to the consolidated financial statements for additional information about this disposition .\\nnet undeveloped acres expiring year ended december 31 .\\nThe u.s . of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 68, 89, 128. The e.g . of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 2014, 92, 36. The other africa of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 189, 4352, 854. The total africa of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 189, 4444, 890. The other international of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 2014, 2014, 2014. The total of net undeveloped acres expiring year ended december 31 , 2016, net undeveloped acres expiring year ended december 31 , 2017, net undeveloped acres expiring year ended december 31 , 2018 are 257, 4533, 1018.\\n.\",7 },8 {9 \"role\": \"user\",10 \"content\": \"what percentage of undeveloped acres were in the us in 2018?\",11 },12 {13 \"role\": \"assistant\",14 \"content\": \"divide(128, 1018)\",15 },16 {17 \"role\": \"user\",18 \"content\": \"what was the total african and us net undeveloped acres expiring in 2016?\",19 },20 ],21)2223print(\"#### Model response ####\")24print(base_response.text)25print(\"########################\")\n```\n\nExample:\n```text\n#### Model response ####The total African undeveloped acres expiring in 2016 is 189 acres, while the US undeveloped acres expiring in the same year is 68 acres. Adding these together gives a total of 257 acres.########################\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.386Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":14,"totalLines":73,"estimatedTokens":3839}}210{"id":"doc-rag_with_chat_embed_and_rerank_via_pinecone_cohe-a6f9f44a","source":"documentation","title":"RAG With Chat Embed and Rerank via Pinecone | Cohere","url":"https://docs.cohere.com/page/rag-with-chat-embed","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1! pip install cohere hnswlib unstructured python-dotenv -q\n```\n\nExample:\n```text\n1import cohere2from pinecone import Pinecone, PodSpec3import uuid4import hnswlib5from typing import List, Dict6from unstructured.partition.html import partition_html7from unstructured.chunking.title import chunk_by_title89co = cohere.Client(\"COHERE_API_KEY\") # Get your API key here: https://dashboard.cohere.com/api-keys10pc = Pinecone(api_key=\"PINECONE_API_KEY\") # (get API key at app.pinecone.io)\n```\n\nExample:\n```text\n1import cohere2import os3import dotenv45dotenv.load_dotenv()6co = cohere.Client(os.getenv(\"COHERE_API_KEY\"))7pc = Pinecone(8 api_key=os.getenv(\"PINECONE_API_KEY\")9)\n```\n\nExample:\n```text\n1raw_documents = [2 {3 \"title\": \"Text Embeddings\",4 \"url\": \"https://docs.cohere.com/docs/text-embeddings\"},5 {6 \"title\": \"Similarity Between Words and Sentences\",7 \"url\": \"https://docs.cohere.com/docs/similarity-between-words-and-sentences\"},8 {9 \"title\": \"The Attention Mechanism\",10 \"url\": \"https://docs.cohere.com/docs/the-attention-mechanism\"},11 {12 \"title\": \"Transformer Models\",13 \"url\": \"https://docs.cohere.com/docs/transformer-models\"}14]\n```\n\nExample:\n```text\n1class Vectorstore:2 \"\"\"3 A class representing a collection of documents indexed into a vectorstore.45 Parameters:6 raw_documents (list): A list of dictionaries representing the sources of the raw documents. Each dictionary should have 'title' and 'url' keys.78 Attributes:9 raw_documents (list): A list of dictionaries representing the raw documents.10 docs (list): A list of dictionaries representing the chunked documents, with 'title', 'text', and 'url' keys.11 docs_embs (list): A list of the associated embeddings for the document chunks.12 docs_len (int): The number of document chunks in the collection.13 idx (hnswlib.Index): The index used for document retrieval.1415 Methods:16 load_and_chunk(): Loads the data from the sources and partitions the HTML content into chunks.17 embed(): Embeds the document chunks using the Cohere API.18 index(): Indexes the document chunks for efficient retrieval.19 retrieve(): Retrieves document chunks based on the given query.20 \"\"\"2122 def __init__(self, raw_documents: List[Dict[str, str]]):23 self.raw_documents = raw_documents24 self.docs = []25 self.docs_embs = []26 self.retrieve_top_k = 1027 self.rerank_top_k = 328 self.load_and_chunk()29 self.embed()30 self.index()313233 def load_and_chunk(self) -> None:34 \"\"\"35 Loads the text from the sources and chunks the HTML content.36 \"\"\"37 print(\"Loading documents...\")3839 for raw_document in self.raw_documents:40 elements = partition_html(url=raw_document[\"url\"])41 chunks = chunk_by_title(elements)42 for chunk in chunks:43 self.docs.append(44 {45 \"title\": raw_document[\"title\"],46 \"text\": str(chunk),47 \"url\": raw_document[\"url\"],48 }49 )5051 def embed(self) -> None:52 \"\"\"53 Embeds the document chunks using the Cohere API.54 \"\"\"55 print(\"Embedding document chunks...\")5657 batch_size = 9058 self.docs_len = len(self.docs)59 for i in range(0, self.docs_len, batch_size):60 batch = self.docs[i : min(i + batch_size, self.docs_len)]61 texts = [item[\"text\"] for item in batch]62 docs_embs_batch = co.embed(63 texts=texts, model=\"embed-v4.0\", input_type=\"search_document\"64 ).embeddings65 self.docs_embs.extend(docs_embs_batch)6667 def index(self) -> None:68 \"\"\"69 Indexes the documents for efficient retrieval.70 \"\"\"71 print(\"Indexing documents...\")7273 index_name = 'rag-01'7475 # If the index does not exist, we create it76 if index_name not in pc.list_indexes().names():77 pc.create_index(78 name=index_name,79 dimension=len(self.docs_embs[0]),80 metric=\"cosine\",81 spec=PodSpec(82 environment=\"gcp-starter\"83 )84 )8586 # connect to index87 self.idx = pc.Index(index_name)8889 batch_size = 1289091 ids = [str(i) for i in range(len(self.docs))]92 # create list of metadata dictionaries93 meta = self.docs9495 # create list of (id, vector, metadata) tuples to be upserted96 to_upsert = list(zip(ids, self.docs_embs, meta))9798 for i in range(0, len(self.docs), batch_size):99 i_end = min(i+batch_size, len(self.docs))100 self.idx.upsert(vectors=to_upsert[i:i_end])101102 # let's view the index statistics103 print(\"Indexing complete\")104105106 def retrieve(self, query: str) -> List[Dict[str, str]]:107 \"\"\"108 Retrieves document chunks based on the given query.109110 Parameters:111 query (str): The query to retrieve document chunks for.112113 Returns:114 List[Dict[str, str]]: A list of dictionaries representing the retrieved document chunks, with 'title', 'text', and 'url' keys.115 \"\"\"116117 docs_retrieved = []118 query_emb = co.embed(119 texts=[query], model=\"embed-v4.0\", input_type=\"search_query\"120 ).embeddings121122123 res = self.idx.query(vector=query_emb, top_k=self.retrieve_top_k, include_metadata=True)124 docs_to_rerank = [match['metadata']['text'] for match in res['matches']]125126 rerank_results = co.rerank(127 query=query,128 documents=docs_to_rerank,129 top_n=self.rerank_top_k,130 model=\"rerank-english-v2.0\",131 )132133 docs_reranked = [res['matches'][result.index] for result in rerank_results.results]134135 for doc in docs_reranked:136 docs_retrieved.append(doc['metadata'])137138 return docs_retrieved\n```\n\nExample:\n```text\n1vectorstore = Vectorstore(raw_documents)\n```\n\nExample:\n```text\nLoading documents...Embedding document chunks...Indexing documents...Indexing complete\n```\n\nExample:\n```text\n1vectorstore.retrieve(\"multi-head attention definition\")\n```\n\nExample:\n```text\n[{'text': 'The attention step used in transformer models is actually much more powerful, and it’s called multi-head attention. In multi-head attention, several different embeddings are used to modify the vectors and add context to them. Multi-head attention has helped language models reach much higher levels of efficacy when processing and generating text.', 'title': 'Transformer Models', 'url': 'https://docs.cohere.com/docs/transformer-models'}, {'text': \"What you learned in this chapter is simple self-attention. However, we can do much better than that. There is a method called multi-head attention, in which one doesn't only consider one embedding, but several different ones. These are all obtained from the original by transforming it in different ways. Multi-head attention has been very successful at the task of adding context to text. If you'd like to learn more about the self and multi-head attention, you can check out the following two\", 'title': 'The Attention Mechanism', 'url': 'https://docs.cohere.com/docs/the-attention-mechanism'}, {'text': 'Attention helps give context to each word, based on the other words in the sentence (or text).', 'title': 'Transformer Models', 'url': 'https://docs.cohere.com/docs/transformer-models'}]\n```\n\nExample:\n```text\n1class Chatbot:2 def __init__(self, vectorstore: Vectorstore):3 \"\"\"4 Initializes an instance of the Chatbot class.56 Parameters:7 vectorstore (Vectorstore): An instance of the Vectorstore class.89 \"\"\"10 self.vectorstore = vectorstore11 self.conversation_id = str(uuid.uuid4())1213 def run(self):14 \"\"\"15 Runs the chatbot application.1617 \"\"\"18 while True:19 # Get the user message20 message = input(\"User: \")2122 # Typing \"quit\" ends the conversation23 if message.lower() == \"quit\":24 print(\"Ending chat.\")25 break26 # else: # Uncomment for Google Colab to avoid printing the same thing twice27 # print(f\"User: {message}\") # Uncomment for Google Colab to avoid printing the same thing twice2829 # Generate search queries (if any)30 response = co.chat(message=message,31 model=\"command-r\",32 search_queries_only=True)3334 # If there are search queries, retrieve document chunks and respond35 if response.search_queries:36 print(\"Retrieving information...\", end=\"\")3738 # Retrieve document chunks for each query39 documents = []40 for query in response.search_queries:41 documents.extend(self.vectorstore.retrieve(query.text))4243 # Use document chunks to respond44 response = co.chat_stream(45 message=message,46 model=\"command-r\",47 documents=documents,48 conversation_id=self.conversation_id,49 )5051 # If there is no search query, directly respond52 else:53 response = co.chat_stream(54 message=message,55 model=\"command-r\",56 conversation_id=self.conversation_id,57 )5859 # Print the chatbot response, citations, and documents60 print(\"\\nChatbot:\")61 citations = []62 cited_documents = []6364 # Display response65 for event in response:66 if event.event_type == \"text-generation\":67 print(event.text, end=\"\")68 elif event.event_type == \"citation-generation\":69 citations.extend(event.citations)70 elif event.event_type == \"search-results\":71 cited_documents = event.documents7273 # Display citations and source documents74 if citations:75 print(\"\\n\\nCITATIONS:\")76 for citation in citations:77 print(citation)7879 print(\"\\nDOCUMENTS:\")80 for document in cited_documents:81 print(document)8283 print(f\"\\n{'-'*100}\\n\")\n```\n\nExample:\n```text\n1chatbot = Chatbot(vectorstore)23chatbot.run()\n```\n\nExample:\n```text\nChatbot:Hello! What's your question? I'm here to help you in any way I can.----------------------------------------------------------------------------------------------------Retrieving information...Chatbot:Word embeddings associate words with lists of numbers, so that similar words are close to each other and dissimilar words are further away.Sentence embeddings do the same thing, but for sentences. Each sentence is associated with a vector of numbers in a coherent way, so that similar sentences are assigned similar vectors, and different sentences are given different vectors.CITATIONS:start=0 end=15 text='Word embeddings' document_ids=['doc_0']start=16 end=53 text='associate words with lists of numbers' document_ids=['doc_0']start=63 end=100 text='similar words are close to each other' document_ids=['doc_0']start=105 end=139 text='dissimilar words are further away.' document_ids=['doc_0']start=140 end=159 text='Sentence embeddings' document_ids=['doc_0', 'doc_2']start=160 end=177 text='do the same thing' document_ids=['doc_0', 'doc_2']start=198 end=211 text='Each sentence' document_ids=['doc_0', 'doc_2']start=215 end=250 text='associated with a vector of numbers' document_ids=['doc_0', 'doc_2']start=256 end=264 text='coherent' document_ids=['doc_2']start=278 end=295 text='similar sentences' document_ids=['doc_0', 'doc_2']start=300 end=324 text='assigned similar vectors' document_ids=['doc_0', 'doc_2']start=330 end=349 text='different sentences' document_ids=['doc_0', 'doc_2']start=354 end=378 text='given different vectors.' document_ids=['doc_0', 'doc_2']DOCUMENTS:{'id': 'doc_0', 'text': 'In the previous chapters, you learned about word and sentence embeddings and similarity between words and sentences. In short, a word embedding is a way to associate words with lists of numbers (vectors) in such a way that similar words are associated with numbers that are close by, and dissimilar words with numbers that are far away from each other. A sentence embedding does the same thing, but associating a vector to every sentence. Similarity is a way to measure how similar two words (or', 'title': 'The Attention Mechanism', 'url': 'https://docs.cohere.com/docs/the-attention-mechanism'}{'id': 'doc_1', 'text': 'Sentence embeddings\\n\\nSo word embeddings seem to be pretty useful, but in reality, human language is much more complicated than simply a bunch of words put together. Human language has structure, sentences, etc. How would one be able to represent, for instance, a sentence? Well, here’s an idea. How about the sums of scores of all the words? For example, say we have a word embedding that assigns the following scores to these words:\\n\\nNo: [1,0,0,0]\\n\\nI: [0,2,0,0]\\n\\nAm: [-1,0,1,0]\\n\\nGood: [0,0,1,3]', 'title': 'Text Embeddings', 'url': 'https://docs.cohere.com/docs/text-embeddings'}{'id': 'doc_2', 'text': 'This is where sentence embeddings come into play. A sentence embedding is just like a word embedding, except it associates every sentence with a vector full of numbers, in a coherent way. By coherent, I mean that it satisfies similar properties as a word embedding. For instance, similar sentences are assigned to similar vectors, different sentences are assigned to different vectors, and most importantly, each of the coordinates of the vector identifies some (whether clear or obscure) property of', 'title': 'Text Embeddings', 'url': 'https://docs.cohere.com/docs/text-embeddings'}----------------------------------------------------------------------------------------------------Retrieving information...Chatbot:The similarities between words and sentences are both quantitative measures of how close the two given items are. There are two types of similarities that can be defined: dot product similarity, and cosine similarity. These methods can determine how similar two words, or sentences, are.CITATIONS:start=54 end=75 text='quantitative measures' document_ids=['doc_0']start=79 end=88 text='how close' document_ids=['doc_0']start=124 end=133 text='two types' document_ids=['doc_0', 'doc_4']start=171 end=193 text='dot product similarity' document_ids=['doc_0', 'doc_4']start=199 end=217 text='cosine similarity.' document_ids=['doc_0', 'doc_4']start=236 end=257 text='determine how similar' document_ids=['doc_0', 'doc_4']DOCUMENTS:{'id': 'doc_0', 'text': 'Now that we know embeddings quite well, let’s move on to using them to find similarities. There are two types of similarities we’ll define in this post: dot product similarity and cosine similarity. Both are very similar and very useful to determine if two words (or sentences) are similar.', 'title': 'Similarity Between Words and Sentences', 'url': 'https://docs.cohere.com/docs/similarity-between-words-and-sentences'}{'id': 'doc_1', 'text': 'But let me add some numbers to this reasoning to make it more clear. Imagine that we calculate similarities for the words in each sentence, and we get the following:\\n\\nThis similarity makes sense in the following ways:\\n\\nThe similarity between each word and itself is 1.\\n\\nThe similarity between any irrelevant word (“the”, “of”, etc.) and any other word is 0.\\n\\nThe similarity between “bank” and “river” is 0.11.\\n\\nThe similarity between “bank” and “money” is 0.25.', 'title': 'The Attention Mechanism', 'url': 'https://docs.cohere.com/docs/the-attention-mechanism'}{'id': 'doc_2', 'text': 'And the results are:\\n\\nThe similarity between sentences 1 and 2: 6738.2858668486715\\n\\nThe similarity between sentences 1 and 3: -122.22666955510499\\n\\nThe similarity between sentences 2 and 3: -3.494608113647928\\n\\nThese results certainly confirm our predictions. The similarity between sentences 1 and 2 is 6738, which is high. The similarities between sentences 1 and 3, and 2 and 3, are -122 and -3.5 (dot products are allowed to be negative too!), which are much lower.', 'title': 'Similarity Between Words and Sentences', 'url': 'https://docs.cohere.com/docs/similarity-between-words-and-sentences'}{'id': 'doc_3', 'text': 'But let me add some numbers to this reasoning to make it more clear. Imagine that we calculate similarities for the words in each sentence, and we get the following:\\n\\nThis similarity makes sense in the following ways:\\n\\nThe similarity between each word and itself is 1.\\n\\nThe similarity between any irrelevant word (“the”, “of”, etc.) and any other word is 0.\\n\\nThe similarity between “bank” and “river” is 0.11.\\n\\nThe similarity between “bank” and “money” is 0.25.', 'title': 'The Attention Mechanism', 'url': 'https://docs.cohere.com/docs/the-attention-mechanism'}{'id': 'doc_4', 'text': 'Now that we know embeddings quite well, let’s move on to using them to find similarities. There are two types of similarities we’ll define in this post: dot product similarity and cosine similarity. Both are very similar and very useful to determine if two words (or sentences) are similar.', 'title': 'Similarity Between Words and Sentences', 'url': 'https://docs.cohere.com/docs/similarity-between-words-and-sentences'}{'id': 'doc_5', 'text': 'And the results are:\\n\\nThe similarity between sentences 1 and 2: 6738.2858668486715\\n\\nThe similarity between sentences 1 and 3: -122.22666955510499\\n\\nThe similarity between sentences 2 and 3: -3.494608113647928\\n\\nThese results certainly confirm our predictions. The similarity between sentences 1 and 2 is 6738, which is high. The similarities between sentences 1 and 3, and 2 and 3, are -122 and -3.5 (dot products are allowed to be negative too!), which are much lower.', 'title': 'Similarity Between Words and Sentences', 'url': 'https://docs.cohere.com/docs/similarity-between-words-and-sentences'}----------------------------------------------------------------------------------------------------Ending chat.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.387Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":4698}}211{"id":"doc-evaluating_text_summarization_models_cohere-574df796","source":"documentation","title":"Evaluating Text Summarization Models | Cohere","url":"https://docs.cohere.com/page/summarization-evals","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1!pip install cohere datasets --quiet\n```\n\nExample:\n```text\n1import json2import random3import re4from typing import List, Optional56import cohere7from getpass import getpass8from datasets import load_dataset9import pandas as pd1011co_api_key = getpass(\"Enter your Cohere API key: \")12co_model = \"command-r\"13co = cohere.Client(api_key=co_api_key)\n```\n\nExample:\n```text\n1qmsum = load_dataset(\"MocktaiLEngineer/qmsum-processed\", split=\"validation\")2transcripts = [x for x in qmsum[\"meeting_transcript\"] if x is not None]\n```\n\nExample:\n```text\nGenerating train split: 0%| | 0/1095 [00:00<?, ? examples/s]Generating validation split: 0%| | 0/237 [00:00<?, ? examples/s]Generating test split: 0%| | 0/244 [00:00<?, ? examples/s]\n```\n\nExample:\n```text\n1prompt_template = \"\"\"## meeting transcript2{transcript}34## instructions5{instructions}\"\"\"\n```\n\nExample:\n```text\n1instruction_objectives = {2 \"general_summarization\": \"Summarize the meeting based on the transcript.\",3 \"action_items\": \"What are the follow-up items based on the meeting transcript?\",4}56format_length_modifiers = {7 \"paragraphs_short\": {8 \"text\": \"In paragraph form, output your response. Use at least 10 words and at most 50 words in total.\",9 \"objectives\": [\"general_summarization\"],10 \"eval_metadata\": {11 \"format\": \"paragraphs\",12 \"min_length\": 10,13 \"max_length\": 50,14 },15 },16 \"paragraphs_medium\": {17 \"text\": \"Return the answer in the form of paragraphs. Make sure your answer is between 50 and 200 words long.\",18 \"objectives\": [\"general_summarization\"],19 \"eval_metadata\": {20 \"format\": \"paragraphs\",21 \"min_length\": 50,22 \"max_length\": 200,23 },24 },25 \"bullets_short_3\": {26 \"text\": \"Format your answer in the form of bullets. Use exactly 3 bullets. Each bullet should be at least 10 words and at most 20 words.\",27 \"objectives\": [\"general_summarization\", \"action_items\"],28 \"eval_metadata\": {29 \"format\": \"bullets\",30 \"number\": 3,31 \"min_length\": 10,32 \"max_length\": 20,33 },34 },35 \"bullets_medium_2\": {36 \"text\": \"In bullets, output your response. Make sure to use exactly 2 bullets. Make sure each bullet is between 20 and 80 words long.\",37 \"objectives\": [\"general_summarization\", \"action_items\"],38 \"eval_metadata\": {39 \"format\": \"bullets\",40 \"number\": 2,41 \"min_length\": 20,42 \"max_length\": 80,43 },44 },45}\n```\n\nExample:\n```text\n1instructions = []2for obj_name, obj_text in instruction_objectives.items():3 for mod_data in format_length_modifiers.values():4 for mod_obj in mod_data[\"objectives\"]:5 if mod_obj == obj_name:6 instruction = {7 \"instruction\": f\"{obj_text} {mod_data['text']}\",8 \"eval_metadata\": mod_data[\"eval_metadata\"],9 \"objective\": obj_name,10 }11 instructions.append(instruction)1213print(json.dumps(instructions[:2], indent=4))\n```\n\nExample:\n```text\n1[2 {3 \"instruction\": \"Summarize the meeting based on the transcript. In paragraph form, output your response. Use at least 10 words and at most 50 words in total.\",4 \"eval_metadata\": {5 \"format\": \"paragraphs\",6 \"min_length\": 10,7 \"max_length\": 508 },9 \"objective\": \"general_summarization\"10 },11 {12 \"instruction\": \"Summarize the meeting based on the transcript. Return the answer in the form of paragraphs. Make sure your answer is between 50 and 200 words long.\",13 \"eval_metadata\": {14 \"format\": \"paragraphs\",15 \"min_length\": 50,16 \"max_length\": 20017 },18 \"objective\": \"general_summarization\"19 }20]\n```\n\nExample:\n```text\n1data = pd.DataFrame(instructions)23transcripts = sorted(transcripts, key=lambda x: len(x), reverse=True)[:int(len(transcripts) * 0.25)]4random.seed(42)5random.shuffle(transcripts)6data[\"transcript\"] = transcripts[:len(data)]78data[\"prompt\"] = data.apply(lambda x: prompt_template.format(transcript=x[\"transcript\"], instructions=x[\"instruction\"]), axis=1)\n```\n\nExample:\n```text\n1data[\"transcript_token_len\"] = [len(x) for x in co.batch_tokenize(data[\"transcript\"].tolist(), model=co_model)]\n```\n\nExample:\n```text\n1print(data[\"prompt\"][0])\n```\n\nExample:\n```text\n## meeting transcriptPhD F: As opposed to the rest of usPhD D: Well comment OK I I remind that me my first objective eh in the project is to to study difference parameters to to find a a good solution to detect eh the overlapping zone in eh speech recorded But eh tsk comment ehhh comment In that way comment I I I begin to to study and to analyze the ehn the recorded speech eh the different session to to find and to locate and to mark eh the the different overlapping zone And eh so eh I was eh I am transcribing the the first session and I I have found eh eh one thousand acoustic events eh besides the overlapping zones eh I I I mean the eh breaths eh aspiration eh eh talk eh eh clap eh comment I do not know what is the different names eh you use to to name the the pause n speechGrad G: Oh I do not think we ve been doing it at that level of detail SoPhD D: Eh I I I do I do not need to to to mmm to m to label the the different acoustic but I prefer because eh I would like to to study if eh I I will find eh eh a good eh parameters eh to detect overlapping I would like to to to test these parameters eh with the another eh eh acoustic events to nnn to eh to find what is the ehm the false eh the false eh hypothesis eh nnn which eh are produced when we use the the ehm this eh parameter eh I mean pitch eh eh difference eh featurePhD A: You know I think some of these that are the nonspeech overlapping events may be difficult even for humans to tell that there s two there I mean if it s a tapping sound you would not necessarily or you know something like that it would be it might be hard to know that it was two separate eventsGrad G: Well You were not talking about just overlaps were you ? You were just talking about acoustic eventsPhD D: I I I I t I t I talk eh about eh acoustic events in general but eh my my objective eh will be eh to study eh overlapping zone Eh ? comment n Eh in twelve minutes I found eh eh one thousand acoustic eventsProfessor E: How many overlaps were there in it ? No no how many of them were the overlaps of speech though ?PhD D: How many ? Eh almost eh three hundred eh in one session in five eh in forty five minutes Alm Three hundred overlapping zone With the overlapping zone overlapping speech speech what eh different durationPostdoc B: Does this ? So if you had an overlap involving three people how many times was that counted ?PhD D: three people two people Eh I would like to consider eh one people with difference noise eh in the background beProfessor E: No no but I think what she s asking is pause if at some particular for some particular stretch you had three people talking instead of two did you call that one event ?PhD D: Oh Oh I consider one event eh for th for that eh for all the zone This th I I I con I consider I consider eh an acoustic event the overlapping zone the period where three speaker or eh are talking togetherGrad G: So let s say me and Jane are talking at the same time and then Liz starts talking also over all of us How many events would that be ?PhD D: So I do not understandGrad G: So two people are talking comment and then a third person starts talking Is there an event right here ?PhD D: Eh no No no For me is the overlapping zone because because you you have s you have more one eh more one voice eh eh produced in a in in a momentGrad G: So i if two or more people are talkingProfessor E: OK So I think We just wanted to understand how you are defining it So then in the region between since there there is some continuous region in between regions where there is only one person speaking And one contiguous region like that you are calling an event Is it Are you calling the beginning or the end of it the event or are you calling the entire length of it the event ?PhD D: I consider the the nnn the nnn nnn eh the entirety eh eh all all the time there were the voice has overlapped This is the idea But eh I I do not distinguish between the the numbers of eh speaker I m not considering eh the the ehm eh the fact of eh eh for example what did you say ? Eh at first eh eh two talkers are eh speaking and eh eh a third person eh join to to that For me it s eh it s eh all overlap zone with eh several numbers of speakers is eh eh the same acoustic event Wi but without any mark between the zone of the overlapping zone with two speakers eh speaking together and the zone with the three speakersPostdoc B: That would j just be onePhD D: Eh with eh a beginning mark and the ending mark Because eh for me is the is the zone with eh some kind of eh distortion the spectral I do not mind By the moment by the momentGrad G: Well but But you could imagine that three people talking has a different spectral characteristic than twoPhD D: I I do not but eh but eh I have to study comment What will happen in a general wayGrad G: So You had to start somewherePhD C: So there s a lot of overlapPhD D: I I do not know what eh will will happen with theGrad G: That s a lot of overlapProfessor E: So again that s that s three three hundred in forty five minutes that are that are speakers just speakersPostdoc B: But a a a thProfessor E: So that s about eight per minutePostdoc B: But a thousand events in twelve minutes that sPhD C: But that can include tapsPostdoc B: Well but a thousand taps in eight minutes is a l in twelve minutes is a lotPhD D: I I con I consider I consider acoustic events eh the silent tooGrad G: Silence starting or silence endingPhD D: silent ground to bec to detect eh because I consider acoustic event all the things are not eh speech In ge in in in a general point of viewProfessor E: OK so how many of those thousand were silence ?PhD F: Not speech not speech or too much speechProfessor E: Right So how many of those thousand were silence silent sections ?PhD D: silent I I I I do not I I have not the eh I I would like to to do a stylistic study## instructionsSummarize the meeting based on the transcript. In paragraph form, output your response. Use at least 10 words and at most 50 words in total.\n```\n\nExample:\n```text\n1grading_prompt_template = \"\"\"You are an AI grader that given a prompt, a completion, and a criterion, grades the completion based on the prompt and criterion. Below is a prompt, a completion, and a criterion with which to grade the completion. You need to respond according to the criterion instructions.23PROMPT:4{prompt}56COMPLETION:7{completion}89CRITERION:10{criteria}1112Do not provide an explanation or respond with additional text. Only respond with the final score in the format SCORE: <final_score>1314SCORE:\"\"\"151617criteria_completeness = \"\"\"Please judge the above completion on whether or not all the completion fully addresses the task posed in the prompt. The completion should contain all the information that is asked for by the prompt while also upholding the constraints mentioned in the prompt. Please return a grade between 1 and 5, where 1 means the completion does not address the prompt at all, 2 means the completion mostly does not address the prompt, 3 is neutral, 4 means the completion mostly addresses the prompt, and 5 means the completion fully addresses the prompt.\"\"\"1819criteria_correctness = \"\"\"Please judge the above completion on whether or not all information was derived from the prompt and that no information is false. Return a grade between 1 and 5, where 1 means the answers are entirely untruthful relative to the prompt, 2 means the answers are somewhat untruthful relative to the prompt, 3 means neutral, 4 means the answers are somewhat truthful relative to the prompt, and 5 means the answers are entirely truthful relative to the prompt. Evaluate only the truthfulness of the answers, not whether or not they capture all the relevant information in the prompt.\"\"\"2021criteria_conciseness = \"\"\"Please judge the above completion on whether or not the completion contains any unnecessary information or wordiness that does not help answer the specific instruction given in the prompt. Return a grade between 1 and 5, where 1 means the completion contains many unnecessary details and wordiness that do not answer the specific instruction given in the prompt, 2 means the completion contains some unnecessary details or wordiness, 3 means neutral, 4 means the completion contains few unnecessary details or wordiness, and 5 means the completion contains only necessary details that answer the specific instruction given in the prompt.\"\"\"222324def score_llm(prompt: str, completion: str, criteria: str) -> int:25 \"\"\"26 Score a completion based on a prompt and a criterion using LLM Because we27 grade all completions on a scale of 1-5, we will normalize the scores by 5 so that the final score28 is between 0 and 1.29 \"\"\"30 grading_prompt = grading_prompt_template.format(31 prompt=prompt, completion=completion, criteria=criteria32 )33 # Use Cohere to grade the completion34 completion = co.chat(message=grading_prompt, model=co_model, temperature=0.2).text3536 ### Alternatively, use OpenAI to grade the completion (requires key)37 # import openai38 # completion = openai.OpenAI(api_key=\"INSERT OPENAI KEY HERE\").chat.completions.create(39 # model=\"gpt-4\",40 # messages=[{\"role\": \"user\", \"content\": grading_prompt}],41 # temperature=0.2,42 # ).choices[0].message.content4344 # Extract the score from the completion45 score = float(re.search(r\"[12345]\", completion).group()) / 546 return score\n```\n\nExample:\n```text\n1def score_format(completion: str, format_type: str) -> int:2 \"\"\"3 Returns 1 if the completion is in the correct format, 0 otherwise.4 \"\"\"5 if format_type == \"paragraphs\":6 return int(_is_only_paragraphs(completion))7 elif format_type == \"bullets\":8 return int(_is_only_bullets(completion))9 return 01011def score_length(12 completion: str,13 format_type: str,14 min_val: int,15 max_val: int,16 number: Optional[int] = None17) -> int:18 \"\"\"19 Returns 1 if the completion has the correct length for the given format, 0 otherwise. This20 includes both word count and number of items (optional).21 \"\"\"22 # Split into items (each bullet for bullets or each paragraph for paragraphs)23 if format_type == \"bullets\":24 items = _extract_markdown_bullets(completion, include_bullet=False)25 elif format_type == \"paragraphs\":26 items = completion.split(\"\\n\")2728 # Strip whitespace and remove empty items29 items = [item for item in items if item.strip() != \"\"]3031 # Check number of items if provided32 if number is not None and len(items) != number:33 return 03435 # Check length of each item36 for item in items:37 num_words = item.strip().split()38 if min_val is None and len(num_words) > max_val:39 return 040 elif max_val is None and len(num_words) < min_val:41 return 042 elif not min_val <= len(num_words) <= max_val:43 return 044 return 1454647def _is_only_bullets(text: str) -> bool:48 \"\"\"49 Returns True if text is only markdown bullets.50 \"\"\"51 bullets = _extract_markdown_bullets(text, include_bullet=True)5253 for bullet in bullets:54 text = text.replace(bullet, \"\")5556 return text.strip() == \"\"575859def _is_only_paragraphs(text: str) -> bool:60 \"\"\"61 Returns True if text is only paragraphs (no bullets).62 \"\"\"63 bullets = _extract_markdown_bullets(text, include_bullet=True)6465 return len(bullets) == 0666768def _extract_markdown_bullets(text: str, include_bullet: bool = False) -> List[str]:69 \"\"\"70 Extracts markdown bullets from text as a list. If include_bullet is True, the bullet will be71 included in the output. The list of accepted bullets is: -, *, +, •, and any number followed by72 a period.73 \"\"\"74 if include_bullet:75 return re.findall(r\"^[ \\t]*(?:[-*+•]|[\\d]+\\.).*\\w+.*$\", text, flags=re.MULTILINE)76 return re.findall(r\"^[ \\t]*(?:[-*+•]|[\\d]+\\.)(.*\\w+.*)$\", text, flags=re.MULTILINE)\n```\n\nExample:\n```text\n1completions = []2for prompt in data[\"prompt\"]:3 completion = co.chat(message=prompt, model=\"command-r\", temperature=0.2).text4 completions.append(completion)56data[\"completion\"] = completions\n```\n\nExample:\n```text\n1print(data[\"completion\"][0])\n```\n\nExample:\n```text\n1data[\"format_score\"] = data.apply(2 lambda x: score_format(x[\"completion\"], x[\"eval_metadata\"][\"format\"]), axis=13)45data[\"length_score\"] = data.apply(6 lambda x: score_length(7 x[\"completion\"],8 x[\"eval_metadata\"][\"format\"],9 x[\"eval_metadata\"].get(\"min_length\"),10 x[\"eval_metadata\"].get(\"max_length\"),11 ),12 axis=1,13)1415data[\"completeness_score\"] = data.apply(16 lambda x: score_llm(x[\"prompt\"], x[\"completion\"], criteria_completeness), axis=117)1819data[\"correctness_score\"] = data.apply(20 lambda x: score_llm(x[\"prompt\"], x[\"completion\"], criteria_correctness), axis=121)2223data[\"conciseness_score\"] = data.apply(24 lambda x: score_llm(x[\"prompt\"], x[\"completion\"], criteria_conciseness), axis=125)\n```\n\nExample:\n```text\n1data\n```\n\nExample:\n```text\n1avg_scores = data[[\"format_score\", \"length_score\", \"completeness_score\", \"correctness_score\", \"conciseness_score\"]].mean()2print(avg_scores)\n```\n\nExample:\n```text\nformat_score 1.000000length_score 0.833333completeness_score 0.800000correctness_score 1.000000conciseness_score 0.800000dtype: float64\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.388Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":20,"totalLines":103,"estimatedTokens":4552}}212{"id":"doc-deep_dive_into_evaluating_rag_outputs_cohere-d15b9ced","source":"documentation","title":"Deep Dive Into Evaluating RAG Outputs | Cohere","url":"https://docs.cohere.com/page/rag-evaluation-deep-dive","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1%%capture2!pip install llama-index cohere openai3!pip install mistralai\n```\n\nExample:\n```text\n1# required imports2from getpass import getpass3import os4import re5import numpy as np6from llama_index.core import SimpleDirectoryReader7from llama_index.core.llama_dataset import download_llama_dataset, LabelledRagDataset8from openai import Client9from mistralai.client import MistralClient\n```\n\nExample:\n```text\n1# Get keys2openai_api_key = getpass(\"Enter your OpenAI API Key: \")3# uncomment if you want to use mistral4#mistral_api_key = getpass[\"Enter your Mistral API Key: \"]56# Define the model you want to use - you can replace gpt-4 with any other gpt version7model = \"gpt-4\"8# uncomment if you want to use mistral9#model = \"mistral-large-latest\"\n```\n\nExample:\n```text\n1if model == \"gpt-4\":2 client = Client(api_key=openai_api_key)3else:4 client = MistralClient(api_key=mistral_api_key)\n```\n\nExample:\n```text\n1# let's define a function to get the model's response for a given input2def get_response(model, client, prompt):3 response = client.chat.completions.create(4 model=model,5 messages=[{\"role\": \"user\", \"content\": prompt}],6 temperature=0)7 return response.choices[0].message.content\n```\n\nExample:\n```text\n1# load the DocugamiKgRagSec10Q dataset2if os.path.exists(\"./data/source_files\") and os.path.exists(\"./data/rag_dataset.json\"):3 rag_dataset = LabelledRagDataset.from_json(\"./data/rag_dataset.json\")4 documents = SimpleDirectoryReader(input_dir=\"./data/source_files\").load_data(show_progress=True)5else:6 rag_dataset, documents = download_llama_dataset(\"DocugamiKgRagSec10Q\", \"./data\")\n```\n\nExample:\n```text\n1class RetrievalEvaluator:23 def compute_precision(self, retrieved_documents, golden_documents):4 # compute the percentage of retrieved documents found in the golden docs5 return len(set(retrieved_documents).intersection(golden_documents)) / len(retrieved_documents)67 def compute_recall(self, retrieved_documents, golden_documents):8 # compute the percentage of golden documents found in the retrieved docs9 return len(set(retrieved_documents).intersection(golden_documents)) / len(golden_documents)1011 def compute_mean_average_precision(self, retrieved_documents, golden_documents):12 # check which among the retrieved docs is found in the gold, keeping the order13 correct_retrieved_documents = [1 if x in golden_documents else 0 for x in retrieved_documents]14 # compute map15 map = np.mean([sum(correct_retrieved_documents[: i + 1]) / (i + 1) for i, v in enumerate(correct_retrieved_documents) if v == 1])16 return map1718 def run_evals(self, retrieved_documents, golden_documents):19 precision = round(self.compute_precision(retrieved_documents, golden_documents),2)20 recall = round(self.compute_recall(retrieved_documents, golden_documents),2)21 map = round(self.compute_mean_average_precision(retrieved_documents, golden_documents),2)22 results = {'precision': [precision],23 'recall': [recall],24 'map': [map]}25 for k,v in results.items():26 print(f\"{k}: {v[0]}\")\n```\n\nExample:\n```text\n1# select the index of a single datapoint - the first one in the dataset2idx = 034# select the query5query = rag_dataset[idx].query67# and the golden docs8golden_docs = rag_dataset[idx].reference_answer.split('SOURCE(S): ')[1].split(', ')910# let's assume we have the following set of retrieved docs11retrieved_docs = ['2022 Q3 AAPL.pdf', '2023 Q1 MSFT.pdf', '2023 Q1 AAPL.pdf']1213print(f'Query: {query}')14print(f'Golden docs: {golden_docs}')15print(f'Retrieved docs: {retrieved_docs}')\n```\n\nExample:\n```text\nQuery: How has Apple's total net sales changed over time?Golden docs: ['2022 Q3 AAPL.pdf', '2023 Q1 AAPL.pdf', '2023 Q2 AAPL.pdf', '2023 Q3 AAPL.pdf']Retrieved docs: ['2022 Q3 AAPL.pdf', '2023 Q1 MSFT.pdf', '2023 Q1 AAPL.pdf']\n```\n\nExample:\n```text\n1# we can now instantiate the evaluator2evaluate_retrieval = RetrievalEvaluator()34# and run the evaluation5evaluate_retrieval.run_evals(retrieved_docs,golden_docs)\n```\n\nExample:\n```text\nprecision: 0.67recall: 0.5map: 0.83\n```\n\nExample:\n```text\n1# first, let's define a function which extracts the claims from a response2def extract_claims(query, response, model, client):34 # define the instructions on how to extract the claims5 preamble = \"You are shown a prompt and a completion. You have to identify the main claims stated in the completion. A claim is any sentence or part of a sentence that expresses a verifiable fact. Please return a bullet list, in which every line includes one of the claims you identified. Do not add any further explanation to the bullet points.\"67 # build the prompt8 prompt = f\"{preamble}\\n\\nPROMPT: {query}\\n\\nCOMPLETION: {response}\"910 # get the claims11 claims = get_response(model, client, prompt)1213 return claims\n```\n\nExample:\n```text\n1# now, let's consider this answer, which we previously generated with command-r2response = \"Apple's total net sales experienced a decline over the last year. The three-month period ended July 1, 2023, saw a total net sale of $81,797 million, which was a 1% decrease from the same period in 2022. The nine-month period ended July 1, 2023, fared slightly better, with a 3% decrease in net sales compared to the first nine months of 2022.\\nThis downward trend continued into the three and six-month periods ending April 1, 2023. Apple's total net sales decreased by 3% and 4% respectively, compared to the same periods in 2022.\"34# let's extract the claims5claims = extract_claims(query, response, model, client)67# and see what the model returns8print(f\"List of claims extracted from the model's response:\\n\\n{claims}\")\n```\n\nExample:\n```text\nList of claims extracted from the model's response:- Apple's total net sales experienced a decline over the last year.- The three-month period ended July 1, 2023, saw a total net sale of $81,797 million.- This was a 1% decrease from the same period in 2022.- The nine-month period ended July 1, 2023, had a 3% decrease in net sales compared to the first nine months of 2022.- The downward trend continued into the three and six-month periods ending April 1, 2023.- Apple's total net sales decreased by 3% and 4% respectively, compared to the same periods in 2022.\n```\n\nExample:\n```text\n1# Let's create a function that checks each claim against a reference text,2# which here we will call \"context\". As you will see, we will use different contexts,3# depending on the metric we want to compute.45def assess_claims(query, claims, context, model, client):67 # define the instructions on how to perform the assessment.8 # the model has to append to each row a binary SUPPORTED tag9 preamble = \"You are shown a prompt, a context and a list of claims. You have to check which of the claims in the list are supported by the context. Please return the list of claims exactly as is it, just append to each row “SUPPORTED=1” if the claim is supported by the context, or “SUPPORTED=0” if the claim is not supported by the context. Do not add any further explanation to the bullet points.\"1011 # turn list into string12 context = '\\n'.join(context)1314 # build the prompt15 prompt = f\"{preamble}\\n\\nPROMPT: {query}\\n\\nCONTEXT:\\n{context}\\n\\nCLAIMS:\\n{claims}\"1617 # get the response18 assessment = get_response(model, client, prompt)1920 return assessment\n```\n\nExample:\n```text\n1# Let's start with Faithfulness: in this case, we want to assess the claims2# in the response against the retrieved documents (i.e., context = retrieved documents)34# for the sake of clarity, we report the actual text of the retrieved documents5retrieved_documents = ['Products and Services Performance\\nThe following table shows net sales by category for the three- and six-month periods ended April 1, 2023 and March 26, 2022 (dollars in millions):\\nThree Months Ended Six Months Ended\\nApril 1,\\n2023March 26,\\n2022 ChangeApril 1,\\n2023March 26,\\n2022 Change\\nNet sales by category:\\niPhone $ 51,334 $ 50,570 2 %$ 117,109 $ 122,198 (4)%\\nMac 7,168 10,435 (31)% 14,903 21,287 (30)%\\niPad 6,670 7,646 (13)% 16,066 14,894 8 %\\nWearables, Home and Accessories 8,757 8,806 (1)% 22,239 23,507 (5)%\\nServices 20,907 19,821 5 % 41,673 39,337 6 %\\nTotal net sales $ 94,836 $ 97,278 (3)%$ 211,990 $ 221,223 (4)%\\niPhone\\niPhone net sales were relatively flat during the second quarter of 2023 compared to the secon d quarter of 2022. Year-over-year iPhone net sales decreased\\nduring the first six months of 2023 due primarily to lower net sales from the Company’ s new iPhone models launched in the fourth quarter of 2022.\\nMac\\nMac net sales decreased during the second quarter and first six months of 2023 compared to the same periods in 2022 due primarily to lower net sales of\\nMacBook Pro.\\niPad\\niPad net sales decreased during the second quarter of 2023 compared to the second quarter of 2022 due primarily to lower net sales of iPad Pro and iPad Air.\\nYear-over-year iPad net sales increased during the first six months of 2023 due primarily to higher net sales of iPad, partially offset by lower net sales of iPad\\nmini .\\nWearables, Home and Accessories\\nWearables, Home and Accessories net sales were relatively flat during the second quarter of 2023 compared to the second quarter of 2022. Year-over-year\\nWearables, Home and Accessories net sales decreased during the first six months of 2023 due primarily to lower net sales of AirPods .\\nServices\\nServices net sales increased during the second quarter and first six months of 2023 compared to the same periods in 2022 due primarily to higher net sales from\\ncloud services, music and advertising.® ®\\n®\\n®\\nApple Inc. | Q2 2023 Form 10-Q | 16', 'Products and Services Performance\\nThe following table shows net sales by category for the three- and nine-month periods ended July 1, 2023 and June 25, 2022 (dollars in millions):\\nThree Months Ended Nine Months Ended\\nJuly 1,\\n2023June 25,\\n2022 ChangeJuly 1,\\n2023June 25,\\n2022 Change\\nNet sales by category:\\niPhone $ 39,669 $ 40,665 (2)%$ 156,778 $ 162,863 (4)%\\nMac 6,840 7,382 (7)% 21,743 28,669 (24)%\\niPad 5,791 7,224 (20)% 21,857 22,118 (1)%\\nWearables, Home and Accessories 8,284 8,084 2 % 30,523 31,591 (3)%\\nServices 21,213 19,604 8 % 62,886 58,941 7 %\\nTotal net sales $ 81,797 $ 82,959 (1)%$ 293,787 $ 304,182 (3)%\\niPhone\\niPhone net sales decreased during the third quarter and first nine months of 2023 compared to the same periods in 2022 due primarily to lower net sales from\\ncertain iPhone models, partially of fset by higher net sales of iPhone 14 Pro models.\\nMac\\nMac net sales decreased during the third quarter and first nine months of 2023 compared to the same periods in 2022 due primarily to lower net sales of laptops.\\niPad\\niPad net sales decreased during the third quarter of 2023 compared to the third quarter of 2022 due primarily to lower net sales across most iPad models. Year-\\nover-year iPad net sales were relatively flat during the first nine months of 2023.\\nWearables, Home and Accessories\\nWearables, Home and Accessories net sales increased during the third quarter of 2023 compare d to the third quarter of 2022 due primarily to higher net sales of\\nWearables, which includes AirPods , Apple Watch and Beats products, partially offset by lower net sales of accessories. Year-over-year Wearables, Home\\nand Accessories net sales decreased during the first nine months of 2023 due primarily to lower net sales of W earables and accessories.\\nServices\\nServices net sales increased during the third quarter of 2023 compared to the third quarter of 2022 due primarily to higher net sales from advertising, cloud\\nservices and the App Store . Year-over-year Services net sales increased during the first nine months of 2023 due primarily to higher net sales from cloud\\nservices, advertising and music.® ® ®\\n®\\nApple Inc. | Q3 2023 Form 10-Q | 16']67# get the Faithfulness assessment for each claim8assessed_claims_faithfulness = assess_claims(query=query,9 claims=claims,10 context=retrieved_documents,11 model=model,12 client=client)1314print(f\"Assessment of the claims extracted from the model's response:\\n\\n{assessed_claims_faithfulness}\")\n```\n\nExample:\n```text\nAssessment of the claims extracted from the model's response:- Apple's total net sales experienced a decline over the last year. SUPPORTED=1- The three-month period ended July 1, 2023, saw a total net sale of $81,797 million. SUPPORTED=1- This was a 1% decrease from the same period in 2022. SUPPORTED=1- The nine-month period ended July 1, 2023, had a 3% decrease in net sales compared to the first nine months of 2022. SUPPORTED=1- The downward trend continued into the three and six-month periods ending April 1, 2023. SUPPORTED=1- Apple's total net sales decreased by 3% and 4% respectively, compared to the same periods in 2022. SUPPORTED=1\n```\n\nExample:\n```text\n1# given the list of claims and their label, compute the final score2# as the proportion of correct claims over the full list of claims3def get_final_score(claims_list):4 supported = len(re.findall(\"SUPPORTED=1\", claims_list))5 non_supported = len(re.findall(\"SUPPORTED=0\", claims_list))6 score = supported / (supported+non_supported)7 return round(score, 2)\n```\n\nExample:\n```text\n1score_faithfulness = get_final_score(assessed_claims_faithfulness)2print(f'Faithfulness: {score_faithfulness}')\n```\n\nExample:\n```text\nFaithfulness: 1.0\n```\n\nExample:\n```text\n1# let's mess up the century, changing 2022 to 19222modified_response = response.replace('2022', '1922')34# extract the claims from the modified response5modified_claims = extract_claims(query, modified_response, model, client)67# and get assess the modified claims8assessed_modified_claims = assess_claims(query=query,9 claims=modified_claims,10 context=retrieved_documents,11 model=model,12 client=client)1314print(f\"Assessment of the modified claims:\\n\\n{assessed_modified_claims}\\n\")1516score_faithfulness_modified_claims = get_final_score(assessed_modified_claims)17print(f'Faithfulness: {score_faithfulness_modified_claims}')\n```\n\nExample:\n```text\nAssessment of the modified claims:- Apple's total net sales experienced a decline over the last year. SUPPORTED=1- The three-month period ended July 1, 2023, saw a total net sale of $81,797 million. SUPPORTED=1- This was a 1% decrease from the same period in 1922. SUPPORTED=0- The nine-month period ended July 1, 2023, had a 3% decrease in net sales compared to the first nine months of 1922. SUPPORTED=0- The downward trend continued into the three and six-month periods ending April 1, 2023. SUPPORTED=1- Apple's total net sales decreased by 3% and 4% respectively, compared to the same periods in 1922. SUPPORTED=0Faithfulness: 0.5\n```\n\nExample:\n```text\n1# let's get the gold answer from the dataset2golden_answer = rag_dataset[idx].reference_answer34# and check the claims in the response against the gold.5# note that assess_claims takes exactly the same args as with Faithfulness6# except for the context, that now is the golden_answer7assessed_claims_correctness = assess_claims(query=query,8 claims=claims,9 context=golden_answer, # note the different context10 model=model,11 client=client)121314print(f\"Assess the claims extracted from the model's response against the golden answer:\\n\\n{assessed_claims_correctness}\")\n```\n\nExample:\n```text\nAssess the claims extracted from the model's response against the golden answer:- Apple's total net sales experienced a decline over the last year. SUPPORTED=1- The three-month period ended July 1, 2023, saw a total net sale of $81,797 million. SUPPORTED=1- This was a 1% decrease from the same period in 2022. SUPPORTED=0- The nine-month period ended July 1, 2023, had a 3% decrease in net sales compared to the first nine months of 2022. SUPPORTED=0- The downward trend continued into the three and six-month periods ending April 1, 2023. SUPPORTED=1- Apple's total net sales decreased by 3% and 4% respectively, compared to the same periods in 2022. SUPPORTED=0\n```\n\nExample:\n```text\n1# we can now compute the final Correctness score2score_correctness = get_final_score(assessed_claims_correctness)3print(f'Correctness: {score_correctness}')\n```\n\nExample:\n```text\nCorrectness: 0.5\n```\n\nExample:\n```text\n1# let's extract the golden claims2gold_claims = extract_claims(query, golden_answer, model, client)34print(f\"List of claims extracted from the gold answer:\\n\\n{gold_claims}\")\n```\n\nExample:\n```text\nList of claims extracted from the gold answer:- For the quarterly period ended June 25, 2022, the total net sales were $82,959 million.- For the quarterly period ended December 31, 2022, the total net sales were $117,154 million.- For the quarterly period ended April 1, 2023, the total net sales were $94,836 million.- For the quarterly period ended July 1, 2023, the total net sales were $81,797 million.- There was an increase in total net sales from the quarter ended June 25, 2022, to the quarter ended December 31, 2022.- There was a decrease in total net sales in the quarters ended April 1, 2023, and July 1, 2023.\n```\n\nExample:\n```text\n1# note that in, this case, the context is the model's response2assessed_claims_coverage = assess_claims(query=query,3 claims=gold_claims,4 context=response,5 model=model,6 client=client)789print(f\"Assess which of the gold claims is in the model's response:\\n\\n{assessed_claims_coverage}\")\n```\n\nExample:\n```text\nAssess which of the gold claims is in the model's response:- For the quarterly period ended June 25, 2022, the total net sales were $82,959 million. SUPPORTED=0- For the quarterly period ended December 31, 2022, the total net sales were $117,154 million. SUPPORTED=0- For the quarterly period ended April 1, 2023, the total net sales were $94,836 million. SUPPORTED=0- For the quarterly period ended July 1, 2023, the total net sales were $81,797 million. SUPPORTED=1- There was an increase in total net sales from the quarter ended June 25, 2022, to the quarter ended December 31, 2022. SUPPORTED=0- There was a decrease in total net sales in the quarters ended April 1, 2023, and July 1, 2023. SUPPORTED=1\n```\n\nExample:\n```text\n1# we compute the final Coverage score2score_coverage = get_final_score(assessed_claims_coverage)3print(f'Coverage: {score_coverage}')\n```\n\nExample:\n```text\nCoverage: 0.33\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.391Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":32,"totalLines":163,"estimatedTokens":4807}}213{"id":"doc-document_translation_with_command_a_translate_co-180782af","source":"documentation","title":"Document Translation with Command A Translate | Cohere","url":"https://docs.cohere.com/page/command-a-translate","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1#!pip install --upgrade cohere\n```\n\nExample:\n```text\n1# 1. Set up your Cohere client, translation prompt and maximum words per chunk 2import cohere34co = cohere.ClientV2(\"<YOUR API KEY>\")5model = \"command-a-translate-08-2025\"67target_language = \"Spanish\"8prompt_template = \"Translate everything that follows into {target_language}:\\n\\n\"9max_words = 15 # Set your desired maximum number of words per chunk1011# 2. Your source text12text = (13 \"Enterprises rely on translation for some of their most sensitive and business-critical documents and cannot risk data leakage, compliance violations, or misunderstandings. Mistranslated documents can reduce trust and have strategic implications.\"14)151617# 3. Define the chunk_split function (from earlier in your notebook)18def chunk_split(text, max_words, threshold=0.8):1920 words = text.split() # Turn the text into a list of words21 chunks = [] # Initialize an empty list to store our chunks22 start = 0 # Starting index for slicing the words list2324 while start < len(words):25 # Determine the end index for the current chunk26 end = min(start + max_words, len(words))27 chunk_words = words[start:end]28 chunk_text = \" \".join(chunk_words) # Combine words back into a string2930 # If we're at the end of the text or the chunk is too short, add it as is31 if end == len(words) or len(chunk_words) < max_words * threshold:32 chunks.append(chunk_text.strip())33 break3435 # Try to find a natural breaking point within the chunk36 split_point = None37 for separator in [\"\\n\", \".\", \")\", \" \"]:38 idx = chunk_text.rfind(separator)39 if idx != -1 and idx >= len(chunk_text) * threshold:40 split_point = idx + 1 # Position after the separator41 break4243 if split_point:44 # If a good split point is found, add the chunk up to that point45 chunks.append(chunk_text[:split_point].strip())46 # Move the start index forward by the number of words consumed47 consumed = len(chunk_text[:split_point].split())48 start += consumed49 else:50 # If no good split point is found, add the entire chunk51 chunks.append(chunk_text.strip())52 start = end # Move to the next chunk5354 return chunks5556# 4. Split the text into chunks using chunk_split57chunks = chunk_split(text, max_words=max_words)5859# 5. Translate each chunk and collect results60translated_chunks = []61for chunk in chunks:62 prompt = prompt_template.format(target_language=target_language) + chunk63 response = co.chat(64 model=model,65 messages=[{\"role\": \"user\", \"content\": prompt}],66 )67 translated = response.message.content[0].text68 translated_chunks.append(translated)6970# 6. Merge the translated chunks back together71translated_text = \" \".join(translated_chunks)7273# 7. Output the final translation74print(translated_text)\n```\n\nExample:\n```text\nLas empresas dependen de la traducción para algunos de sus documentos más confidenciales y esenciales para su actividad, y no puede arriesgarse a que se produzcan fugas de datos, incumplimientos de la normativa o malentendidos. Los documentos mal traducidos pueden reducir la confianza y tienen consecuencias estratégicas.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.392Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":3,"totalLines":18,"estimatedTokens":894}}214{"id":"doc-advanced_document_parsing_for_enterprises_cohere-735fa088","source":"documentation","title":"Advanced Document Parsing For Enterprises | Cohere","url":"https://docs.cohere.com/page/document-parsing-for-enterprises","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1%%capture2! sudo apt install tesseract-ocr poppler-utils3! pip install \"cohere<5\" fsspec hnswlib google-cloud-documentai google-cloud-storage boto3 langchain-text-splitters llama_parse pytesseract pdf2image pandas\n```\n\nExample:\n```text\n1data_dir = \"data/document-parsing\"2source_filename = \"fda-approved-drug\"3extension = \"pdf\"\n```\n\nExample:\n```text\n1from pathlib import Path23sources = [\"gcp\", \"aws\", \"unstructured-io\", \"llamaparse-text\", \"llamaparse-markdown\", \"pytesseract\"]45filenames = [\"{}-parsed-fda-approved-drug.txt\".format(source) for source in sources]6filenames.append(\"fda-approved-drug.pdf\")78for filename in filenames:9 file_path = Path(f\"{data_dir}/{filename}\")10 if file_path.is_file() == False:11 print(f\"File {filename} not found at {data_dir}!\")\n```\n\nExample:\n```text\n1def store_document(path: str, doc_content: str):2 with open(path, 'w') as f:3 f.write(doc_content)\n```\n\nExample:\n```text\n1import json23def insert_citations_in_order(text, citations, documents):4 \"\"\"5 A helper function to pretty print citations.6 \"\"\"78 citations_reference = {}9 for index, doc in enumerate(documents):10 citations_reference[index] = doc1112 offset = 013 # Process citations in the order they were provided14 for citation in citations:15 # Adjust start/end with offset16 start, end = citation['start'] + offset, citation['end'] + offset17 citation_numbers = []18 for doc_id in citation[\"document_ids\"]:19 for citation_index, doc in citations_reference.items():20 if doc[\"id\"] == doc_id:21 citation_numbers.append(citation_index)22 references = \"(\" + \", \".join(\"[{}]\".format(num) for num in citation_numbers) + \")\"23 modification = f'{text[start:end]} {references}'24 # Replace the cited text with its bolded version + placeholder25 text = text[:start] + modification + text[end:]26 # Update the offset for subsequent replacements27 offset += len(modification) - (end - start)2829 # Add the citations at the bottom of the text30 text_with_citations = f'{text}'31 citations_reference = [\"[{}]: {}\".format(x[\"id\"], x[\"text\"]) for x in citations_reference.values()]3233 return text_with_citations, \"\\n\".join(citations_reference)\n```\n\nExample:\n```text\n1def format_docs_for_chat(documents):2 return [{\"id\": str(index), \"text\": x} for index, x in enumerate(documents)]\n```\n\nExample:\n```text\n1\"\"\"2Extracted from https://cloud.google.com/document-ai/docs/samples/documentai-batch-process-document3\"\"\"45import re6from typing import Optional78from google.api_core.client_options import ClientOptions9from google.api_core.exceptions import InternalServerError10from google.api_core.exceptions import RetryError11from google.cloud import documentai # type: ignore12from google.cloud import storage1314project_id = \"\"15location = \"\"16processor_id = \"\"17gcs_output_uri = \"\"18# credentials_file = \"populate if you are running in a non Vertex AI environment.\"19gcs_input_prefix = \"\"202122def batch_process_documents(23 project_id: str,24 location: str,25 processor_id: str,26 gcs_output_uri: str,27 gcs_input_prefix: str,28 timeout: int = 40029) -> None:30 parsed_documents = []3132 # Client configs33 opts = ClientOptions(api_endpoint=f\"{location}-documentai.googleapis.com\")34 # With credentials35 # opts = ClientOptions(api_endpoint=f\"{location}-documentai.googleapis.com\", credentials_file=credentials_file)3637 client = documentai.DocumentProcessorServiceClient(client_options=opts)38 processor_name = client.processor_path(project_id, location, processor_id)3940 # Input storage configs41 gcs_prefix = documentai.GcsPrefix(gcs_uri_prefix=gcs_input_prefix)42 input_config = documentai.BatchDocumentsInputConfig(gcs_prefix=gcs_prefix)4344 # Output storage configs45 gcs_output_config = documentai.DocumentOutputConfig.GcsOutputConfig(gcs_uri=gcs_output_uri, field_mask=None)46 output_config = documentai.DocumentOutputConfig(gcs_output_config=gcs_output_config)47 storage_client = storage.Client()48 # With credentials49 # storage_client = storage.Client.from_service_account_json(json_credentials_path=credentials_file)5051 # Batch process docs request52 request = documentai.BatchProcessRequest(53 name=processor_name,54 input_documents=input_config,55 document_output_config=output_config,56 )5758 # batch_process_documents returns a long running operation59 operation = client.batch_process_documents(request)6061 # Continually polls the operation until it is complete.62 # This could take some time for larger files63 try:64 print(f\"Waiting for operation {operation.operation.name} to complete...\")65 operation.result(timeout=timeout)66 except (RetryError, InternalServerError) as e:67 print(e.message)6869 # Get output document information from completed operation metadata70 metadata = documentai.BatchProcessMetadata(operation.metadata)71 if metadata.state != documentai.BatchProcessMetadata.State.SUCCEEDED:72 raise ValueError(f\"Batch Process Failed: {metadata.state_message}\")7374 print(\"Output files:\")75 # One process per Input Document76 for process in list(metadata.individual_process_statuses):77 matches = re.match(r\"gs://(.*?)/(.*)\", process.output_gcs_destination)78 if not matches:79 print(\"Could not parse output GCS destination:\", process.output_gcs_destination)80 continue8182 output_bucket, output_prefix = matches.groups()83 output_blobs = storage_client.list_blobs(output_bucket, prefix=output_prefix)8485 # Document AI may output multiple JSON files per source file86 # (Large documents get split in multiple file \"versions\" doc --> parsed_doc_0 + parsed_doc_1 ...)87 for blob in output_blobs:88 # Document AI should only output JSON files to GCS89 if blob.content_type != \"application/json\":90 print(f\"Skipping non-supported file: {blob.name} - Mimetype: {blob.content_type}\")91 continue9293 # Download JSON file as bytes object and convert to Document Object94 print(f\"Fetching {blob.name}\")95 document = documentai.Document.from_json(blob.download_as_bytes(), ignore_unknown_fields=True)96 # Store the filename and the parsed versioned document content as a tuple97 parsed_documents.append((blob.name.split(\"/\")[-1].split(\".\")[0], document.text))9899 print(\"Finished document parsing process.\")100 return parsed_documents101102# Call service103# versioned_parsed_documents = batch_process_documents(104# project_id=project_id,105# location=location,106# processor_id=processor_id,107# gcs_output_uri=gcs_output_uri,108# gcs_input_prefix=gcs_input_prefix109# )\n```\n\nExample:\n```text\n1\"\"\"2Post process parsed document and store it locally.3Make sure to run this in a Google Vertex AI environment or include a credentials file.4\"\"\"56\"\"\"7from pathlib import Path8from collections import defaultdict910parsed_documents = []11combined_versioned_parsed_documents = defaultdict(list)1213# Assemble versioned documents together ({\"doc_name\": [(0, doc_content_0), (1, doc_content_1), ...]}).14for filename, doc_content in versioned_parsed_documents:15 filename, version = \"-\".join(filename.split(\"-\")[:-1]), filename.split(\"-\")[-1]16 combined_versioned_parsed_documents[filename].append((version, doc_content))1718# Sort documents by version and join the content together.19for filename, docs in combined_versioned_parsed_documents.items():20 doc_content = \" \".join([x[1] for x in sorted(docs, key=lambda x: x[0])])21 parsed_documents.append((filename, doc_content))2223# Store parsed documents in local storage.24for filename, doc_content in parsed_documents:25 file_path = \"{}/{}-parsed-{}.txt\".format(data_dir, \"gcp\", source_filename)26 store_document(file_path, doc_content)27\"\"\"\n```\n\nExample:\n```text\n1filename = \"gcp-parsed-{}.txt\".format(source_filename)2with open(\"{}/{}\".format(data_dir, filename), \"r\") as doc:3 parsed_document = doc.read()45print(parsed_document[:1000])\n```\n\nExample:\n```text\n1# source: https://github.com/awsdocs/aws-doc-sdk-examples/tree/main/python/example_code/textract23# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.4# SPDX-License-Identifier: Apache-2.056\"\"\"7Purpose89Shows how to use the AWS SDK for Python (Boto3) with Amazon Textract to10detect text, form, and table elements in document images.11\"\"\"1213import json14import logging15from botocore.exceptions import ClientError1617logger = logging.getLogger(__name__)181920# snippet-start:[python.example_code.textract.TextractWrapper]21class TextractWrapper:22 \"\"\"Encapsulates Textract functions.\"\"\"2324 def __init__(self, textract_client, s3_resource, sqs_resource):25 \"\"\"26 :param textract_client: A Boto3 Textract client.27 :param s3_resource: A Boto3 Amazon S3 resource.28 :param sqs_resource: A Boto3 Amazon SQS resource.29 \"\"\"30 self.textract_client = textract_client31 self.s3_resource = s3_resource32 self.sqs_resource = sqs_resource3334 # snippet-end:[python.example_code.textract.TextractWrapper]3536 # snippet-start:[python.example_code.textract.DetectDocumentText]37 def detect_file_text(self, *, document_file_name=None, document_bytes=None):38 \"\"\"39 Detects text elements in a local image file or from in-memory byte data.40 The image must be in PNG or JPG format.4142 :param document_file_name: The name of a document image file.43 :param document_bytes: In-memory byte data of a document image.44 :return: The response from Amazon Textract, including a list of blocks45 that describe elements detected in the image.46 \"\"\"47 if document_file_name is not None:48 with open(document_file_name, \"rb\") as document_file:49 document_bytes = document_file.read()50 try:51 response = self.textract_client.detect_document_text(52 Document={\"Bytes\": document_bytes}53 )54 logger.info(\"Detected %s blocks.\", len(response[\"Blocks\"]))55 except ClientError:56 logger.exception(\"Couldn't detect text.\")57 raise58 else:59 return response6061 # snippet-end:[python.example_code.textract.DetectDocumentText]6263 # snippet-start:[python.example_code.textract.AnalyzeDocument]64 def analyze_file(65 self, feature_types, *, document_file_name=None, document_bytes=None66 ):67 \"\"\"68 Detects text and additional elements, such as forms or tables, in a local image69 file or from in-memory byte data.70 The image must be in PNG or JPG format.7172 :param feature_types: The types of additional document features to detect.73 :param document_file_name: The name of a document image file.74 :param document_bytes: In-memory byte data of a document image.75 :return: The response from Amazon Textract, including a list of blocks76 that describe elements detected in the image.77 \"\"\"78 if document_file_name is not None:79 with open(document_file_name, \"rb\") as document_file:80 document_bytes = document_file.read()81 try:82 response = self.textract_client.analyze_document(83 Document={\"Bytes\": document_bytes}, FeatureTypes=feature_types84 )85 logger.info(\"Detected %s blocks.\", len(response[\"Blocks\"]))86 except ClientError:87 logger.exception(\"Couldn't detect text.\")88 raise89 else:90 return response9192 # snippet-end:[python.example_code.textract.AnalyzeDocument]9394 # snippet-start:[python.example_code.textract.helper.prepare_job]95 def prepare_job(self, bucket_name, document_name, document_bytes):96 \"\"\"97 Prepares a document image for an asynchronous detection job by uploading98 the image bytes to an Amazon S3 bucket. Amazon Textract must have permission99 to read from the bucket to process the image.100101 :param bucket_name: The name of the Amazon S3 bucket.102 :param document_name: The name of the image stored in Amazon S3.103 :param document_bytes: The image as byte data.104 \"\"\"105 try:106 bucket = self.s3_resource.Bucket(bucket_name)107 bucket.upload_fileobj(document_bytes, document_name)108 logger.info(\"Uploaded %s to %s.\", document_name, bucket_name)109 except ClientError:110 logger.exception(\"Couldn't upload %s to %s.\", document_name, bucket_name)111 raise112113 # snippet-end:[python.example_code.textract.helper.prepare_job]114115 # snippet-start:[python.example_code.textract.helper.check_job_queue]116 def check_job_queue(self, queue_url, job_id):117 \"\"\"118 Polls an Amazon SQS queue for messages that indicate a specified Textract119 job has completed.120121 :param queue_url: The URL of the Amazon SQS queue to poll.122 :param job_id: The ID of the Textract job.123 :return: The status of the job.124 \"\"\"125 status = None126 try:127 queue = self.sqs_resource.Queue(queue_url)128 messages = queue.receive_messages()129 if messages:130 msg_body = json.loads(messages[0].body)131 msg = json.loads(msg_body[\"Message\"])132 if msg.get(\"JobId\") == job_id:133 messages[0].delete()134 status = msg.get(\"Status\")135 logger.info(136 \"Got message %s with status %s.\", messages[0].message_id, status137 )138 else:139 logger.info(\"No messages in queue %s.\", queue_url)140 except ClientError:141 logger.exception(\"Couldn't get messages from queue %s.\", queue_url)142 else:143 return status144145 # snippet-end:[python.example_code.textract.helper.check_job_queue]146147 # snippet-start:[python.example_code.textract.StartDocumentTextDetection]148 def start_detection_job(149 self, bucket_name, document_file_name, sns_topic_arn, sns_role_arn150 ):151 \"\"\"152 Starts an asynchronous job to detect text elements in an image stored in an153 Amazon S3 bucket. Textract publishes a notification to the specified Amazon SNS154 topic when the job completes.155 The image must be in PNG, JPG, or PDF format.156157 :param bucket_name: The name of the Amazon S3 bucket that contains the image.158 :param document_file_name: The name of the document image stored in Amazon S3.159 :param sns_topic_arn: The Amazon Resource Name (ARN) of an Amazon SNS topic160 where the job completion notification is published.161 :param sns_role_arn: The ARN of an AWS Identity and Access Management (IAM)162 role that can be assumed by Textract and grants permission163 to publish to the Amazon SNS topic.164 :return: The ID of the job.165 \"\"\"166 try:167 response = self.textract_client.start_document_text_detection(168 DocumentLocation={169 \"S3Object\": {\"Bucket\": bucket_name, \"Name\": document_file_name}170 },171 NotificationChannel={172 \"SNSTopicArn\": sns_topic_arn,173 \"RoleArn\": sns_role_arn,174 },175 )176 job_id = response[\"JobId\"]177 logger.info(178 \"Started text detection job %s on %s.\", job_id, document_file_name179 )180 except ClientError:181 logger.exception(\"Couldn't detect text in %s.\", document_file_name)182 raise183 else:184 return job_id185186 # snippet-end:[python.example_code.textract.StartDocumentTextDetection]187188 # snippet-start:[python.example_code.textract.GetDocumentTextDetection]189 def get_detection_job(self, job_id):190 \"\"\"191 Gets data for a previously started text detection job.192193 :param job_id: The ID of the job to retrieve.194 :return: The job data, including a list of blocks that describe elements195 detected in the image.196 \"\"\"197 try:198 response = self.textract_client.get_document_text_detection(JobId=job_id)199 job_status = response[\"JobStatus\"]200 logger.info(\"Job %s status is %s.\", job_id, job_status)201 except ClientError:202 logger.exception(\"Couldn't get data for job %s.\", job_id)203 raise204 else:205 return response206207 # snippet-end:[python.example_code.textract.GetDocumentTextDetection]208209 # snippet-start:[python.example_code.textract.StartDocumentAnalysis]210 def start_analysis_job(211 self,212 bucket_name,213 document_file_name,214 feature_types,215 sns_topic_arn,216 sns_role_arn,217 ):218 \"\"\"219 Starts an asynchronous job to detect text and additional elements, such as220 forms or tables, in an image stored in an Amazon S3 bucket. Textract publishes221 a notification to the specified Amazon SNS topic when the job completes.222 The image must be in PNG, JPG, or PDF format.223224 :param bucket_name: The name of the Amazon S3 bucket that contains the image.225 :param document_file_name: The name of the document image stored in Amazon S3.226 :param feature_types: The types of additional document features to detect.227 :param sns_topic_arn: The Amazon Resource Name (ARN) of an Amazon SNS topic228 where job completion notification is published.229 :param sns_role_arn: The ARN of an AWS Identity and Access Management (IAM)230 role that can be assumed by Textract and grants permission231 to publish to the Amazon SNS topic.232 :return: The ID of the job.233 \"\"\"234 try:235 response = self.textract_client.start_document_analysis(236 DocumentLocation={237 \"S3Object\": {\"Bucket\": bucket_name, \"Name\": document_file_name}238 },239 NotificationChannel={240 \"SNSTopicArn\": sns_topic_arn,241 \"RoleArn\": sns_role_arn,242 },243 FeatureTypes=feature_types,244 )245 job_id = response[\"JobId\"]246 logger.info(247 \"Started text analysis job %s on %s.\", job_id, document_file_name248 )249 except ClientError:250 logger.exception(\"Couldn't analyze text in %s.\", document_file_name)251 raise252 else:253 return job_id254255 # snippet-end:[python.example_code.textract.StartDocumentAnalysis]256257 # snippet-start:[python.example_code.textract.GetDocumentAnalysis]258 def get_analysis_job(self, job_id):259 \"\"\"260 Gets data for a previously started detection job that includes additional261 elements.262263 :param job_id: The ID of the job to retrieve.264 :return: The job data, including a list of blocks that describe elements265 detected in the image.266 \"\"\"267 try:268 response = self.textract_client.get_document_analysis(JobId=job_id)269 job_status = response[\"JobStatus\"]270 logger.info(\"Job %s status is %s.\", job_id, job_status)271 except ClientError:272 logger.exception(\"Couldn't get data for job %s.\", job_id)273 raise274 else:275 return response276277278# snippet-end:[python.example_code.textract.GetDocumentAnalysis]\n```\n\nExample:\n```text\n1import boto323textract_client = boto3.client('textract')4s3_client = boto3.client('s3')56textractWrapper = TextractWrapper(textract_client, s3_client, None)\n```\n\nExample:\n```text\n1bucket_name = \"your-bucket-name\"2sns_topic_arn = \"your-sns-arn\" # this can be found under the topic you created in the Amazon SNS dashboard3sns_role_arn = \"sns-role-arn\" # this is an IAM role that allows Textract to interact with SNS45file_name = \"fda-approved-drug.pdf\"\n```\n\nExample:\n```text\n1# kick off a text detection job. This returns a job ID.2job_id = textractWrapper.start_detection_job(bucket_name=bucket_name, document_file_name=file_name,3 sns_topic_arn=sns_topic_arn, sns_role_arn=sns_role_arn)\n```\n\nExample:\n```text\n1def get_text_results_from_textract(job_id):2 response = textract_client.get_document_text_detection(JobId=job_id)3 collection_of_textract_responses = []4 pages = [response]56 collection_of_textract_responses.append(response)78 while 'NextToken' in response:9 next_token = response['NextToken']10 response = textract_client.get_document_text_detection(JobId=job_id, NextToken=next_token)11 pages.append(response)12 collection_of_textract_responses.append(response)13 return collection_of_textract_responses1415def get_the_text_with_required_info(collection_of_textract_responses):16 total_text = []17 total_text_with_info = []18 running_sequence_number = 01920 font_sizes_and_line_numbers = {}21 for page in collection_of_textract_responses:22 per_page_text = []23 blocks = page['Blocks']24 for block in blocks:25 if block['BlockType'] == 'LINE':26 block_text_dict = {}27 running_sequence_number += 128 block_text_dict.update(text=block['Text'])29 block_text_dict.update(page=block['Page'])30 block_text_dict.update(left_indent=round(block['Geometry']['BoundingBox']['Left'], 2))31 font_height = round(block['Geometry']['BoundingBox']['Height'], 3)32 line_number = running_sequence_number33 block_text_dict.update(font_height=round(block['Geometry']['BoundingBox']['Height'], 3))34 block_text_dict.update(indent_from_top=round(block['Geometry']['BoundingBox']['Top'], 2))35 block_text_dict.update(text_width=round(block['Geometry']['BoundingBox']['Width'], 2))36 block_text_dict.update(line_number=running_sequence_number)3738 if font_height in font_sizes_and_line_numbers:39 line_numbers = font_sizes_and_line_numbers[font_height]40 line_numbers.append(line_number)41 font_sizes_and_line_numbers[font_height] = line_numbers42 else:43 line_numbers = []44 line_numbers.append(line_number)45 font_sizes_and_line_numbers[font_height] = line_numbers4647 total_text.append(block['Text'])48 per_page_text.append(block['Text'])49 total_text_with_info.append(block_text_dict)5051 return total_text, total_text_with_info, font_sizes_and_line_numbers5253def get_text_with_line_spacing_info(total_text_with_info):54 i = 155 text_info_with_line_spacing_info = []56 while (i < len(total_text_with_info) - 1):57 previous_line_info = total_text_with_info[i - 1]58 current_line_info = total_text_with_info[i]59 next_line_info = total_text_with_info[i + 1]60 if current_line_info['page'] == next_line_info['page'] and previous_line_info['page'] == current_line_info[61 'page']:62 line_spacing_after = round((next_line_info['indent_from_top'] - current_line_info['indent_from_top']), 2)63 spacing_with_prev = round((current_line_info['indent_from_top'] - previous_line_info['indent_from_top']), 2)64 current_line_info.update(line_space_before=spacing_with_prev)65 current_line_info.update(line_space_after=line_spacing_after)66 text_info_with_line_spacing_info.append(current_line_info)67 else:68 text_info_with_line_spacing_info.append(None)69 i += 170 return text_info_with_line_spacing_info\n```\n\nExample:\n```text\n1all_text = \"\\n\".join([line[\"text\"] if line else \"\" for line in text_info_with_line_spacing])23with open(f\"aws-parsed-{source_filename}.txt\", \"w\") as f:4 f.write(all_text)\n```\n\nExample:\n```text\n1filename = \"aws-parsed-{}.txt\".format(source_filename)2with open(\"{}/{}\".format(data_dir, filename), \"r\") as doc:3 parsed_document = doc.read()45print(parsed_document[:1000])\n```\n\nExample:\n```text\n1import os2import requests34UNSTRUCTURED_URL = \"\" # enter service endpoint, for example \"http://localhost:9500/general/v0/general\" (assuming the container is running locally and exposing the service with a -p 9500:9500 port mapping)567parsed_documents = []89input_path = \"{}/{}.{}\".format(data_dir, source_filename, extension)10with open(input_path, 'rb') as file_data:11 response = requests.post(12 url=UNSTRUCTURED_URL,13 files={\"files\": (\"{}.{}\".format(source_filename, extension), file_data)},14 data={15 \"output_format\": (None, \"application/json\"),16 \"strategy\": \"fast\",17 \"pdf_infer_table_structure\": \"true\",18 \"include_page_breaks\": \"true\"19 },20 headers={\"Accept\": \"application/json\"}21 )2223parsed_response = response.json()2425parsed_document = \" \".join([parsed_entry[\"text\"] for parsed_entry in parsed_response])26print(\"Parsed {}\".format(source_filename))\n```\n\nExample:\n```text\n1\"\"\"2Post process parsed document and store it locally.3\"\"\"45file_path = \"{}/{}-parsed-fda-approved-drug.txt\".format(data_dir, \"unstructured-io\")6store_document(file_path, parsed_document)\n```\n\nExample:\n```text\n1filename = \"unstructured-io-parsed-{}.txt\".format(source_filename)2with open(\"{}/{}\".format(data_dir, filename), \"r\") as doc:3 parsed_document = doc.read()45print(parsed_document[:1000])\n```\n\nExample:\n```text\n1import os2from llama_parse import LlamaParse34import nest_asyncio # needed to notebook env5nest_asyncio.apply() # needed to notebook env67llama_index_api_key = \"{API_KEY}\"8input_path = \"{}/{}.{}\".format(data_dir, source_filename, extension)\n```\n\nExample:\n```text\n1# Text mode2text_parser = LlamaParse(3 api_key=llama_index_api_key,4 result_type=\"text\"5)67text_response = text_parser.load_data(input_path)8text_parsed_document = \" \".join([parsed_entry.text for parsed_entry in text_response])910print(\"Parsed {} to text\".format(source_filename))\n```\n\nExample:\n```text\n1\"\"\"2Post process parsed document and store it locally.3\"\"\"45file_path = \"{}/{}-text-parsed-fda-approved-drug.txt\".format(data_dir, \"llamaparse\")6store_document(file_path, text_parsed_document)\n```\n\nExample:\n```text\n1# Markdown mode2markdown_parser = LlamaParse(3 api_key=llama_index_api_key,4 result_type=\"markdown\"5)67markdown_response = markdown_parser.load_data(input_path)8markdown_parsed_document = \" \".join([parsed_entry.text for parsed_entry in markdown_response])910print(\"Parsed {} to markdown\".format(source_filename))\n```\n\nExample:\n```text\n1\"\"\"2Post process parsed document and store it locally.3\"\"\"45file_path = \"{}/{}-markdown-parsed-fda-approved-drug.txt\".format(data_dir, \"llamaparse\")6store_document(file_path, markdown_parsed_document)\n```\n\nExample:\n```text\n1# Text parsing23filename = \"llamaparse-text-parsed-{}.txt\".format(source_filename)45with open(\"{}/{}\".format(data_dir, filename), \"r\") as doc:6 parsed_document = doc.read()78print(parsed_document[:1000])\n```\n\nExample:\n```text\n1# Markdown parsing23filename = \"llamaparse-markdown-parsed-fda-approved-drug.txt\"4with open(\"{}/{}\".format(data_dir, filename), \"r\") as doc:5 parsed_document = doc.read()67print(parsed_document[:1000])\n```\n\nExample:\n```text\n1from matplotlib import pyplot as plt2from pdf2image import convert_from_path3import pytesseract\n```\n\nExample:\n```text\n1# pdf2image extracts as a list of PIL.Image objects2pages = convert_from_path(filename)\n```\n\nExample:\n```text\n1# we look at the first page as a sanity check:23plt.imshow(pages[0])4plt.axis('off')5plt.show()\n```\n\nExample:\n```text\n1label_ocr_pytesseract = \"\".join([pytesseract.image_to_string(page) for page in pages])\n```\n\nExample:\n```text\n1print(label_ocr_pytesseract[:200])\n```\n\nExample:\n```text\nHIGHLIGHTS OF PRESCRIBING INFORMATIONThese highlights do not include all the information needed to useIWILFIN™ safely and effectively. See full prescribing information forIWILFIN.IWILFIN™ (eflor\n```\n\nExample:\n```text\n1label_ocr_pytesseract = \"\".join([pytesseract.image_to_string(page) for page in pages])23with open(f\"pytesseract-parsed-{source_filename}.txt\", \"w\") as f:4 f.write(label_ocr_pytesseract)\n```\n\nExample:\n```text\n1filename = \"pytesseract-parsed-{}.txt\".format(source_filename)2with open(\"{}/{}\".format(data_dir, filename), \"r\") as doc:3 parsed_document = doc.read()45print(parsed_document[:1000])\n```\n\nExample:\n```text\n1import cohere2co = cohere.Client(api_key=\"{API_KEY}\")\n```\n\nExample:\n```text\n1\"\"\"2Document Questions3\"\"\"4prompt = \"What are the most common adverse reactions of Iwilfin?\"5# prompt = \"What is the recommended dosage of Iwilfin on body surface area between 0.5 m2 and 0.75 m2?\"6# prompt = \"I need a succinct summary of the compound name, indication, route of administration, and mechanism of action of Iwilfin.\"78\"\"\"9Choose one of the above solutions10\"\"\"11source = \"gcp\"12# source = \"aws\"13# source = \"unstructured-io\"14# source = \"llamaparse-text\"15# source = \"llamaparse-markdown\"16# source = \"pytesseract\"\n```\n\nExample:\n```text\n1\"\"\"2Read parsed document content and chunk data3\"\"\"45import os6from langchain_text_splitters import RecursiveCharacterTextSplitter78documents = []910with open(\"{}/{}-parsed-fda-approved-drug.txt\".format(data_dir, source), \"r\") as doc:11 doc_content = doc.read()1213\"\"\"14Personal notes on chunking15https://medium.com/@ayhamboucher/llm-based-context-splitter-for-large-documents-445d3f02b01b16\"\"\"171819# Chunk doc content20text_splitter = RecursiveCharacterTextSplitter(21 chunk_size=512,22 chunk_overlap=200,23 length_function=len,24 is_separator_regex=False25)2627# Split the text into chunks with some overlap28chunks_ = text_splitter.create_documents([doc_content])29documents = [c.page_content for c in chunks_]3031print(\"Source document has been broken down to {} chunks\".format(len(documents)))\n```\n\nExample:\n```text\n1\"\"\"2Embed document chunks3\"\"\"4document_embeddings = co.embed(texts=documents, model=\"embed-v4.0\", input_type=\"search_document\").embeddings\n```\n\nExample:\n```text\n1\"\"\"2Create document index and add embedded chunks3\"\"\"45import hnswlib67index = hnswlib.Index(space='ip', dim=1536) # space: inner product8index.init_index(max_elements=len(document_embeddings), ef_construction=512, M=64)9index.add_items(document_embeddings, list(range(len(document_embeddings))))10print(\"Count:\", index.element_count)\n```\n\nExample:\n```text\nCount: 115\n```\n\nExample:\n```text\n1\"\"\"2Embed search query3Fetch k nearest neighbors4\"\"\"56query_emb = co.embed(texts=[prompt], model='embed-v4.0', input_type=\"search_query\").embeddings7default_knn = 108knn = default_knn if default_knn <= index.element_count else index.element_count9result = index.knn_query(query_emb, k=knn)10neighbors = [(result[0][0][i], result[1][0][i]) for i in range(len(result[0][0]))]11relevant_docs = [documents[x[0]] for x in sorted(neighbors, key=lambda x: x[1])]\n```\n\nExample:\n```text\n1\"\"\"2Rerank retrieved documents3\"\"\"45rerank_results = co.rerank(query=prompt, documents=relevant_docs, top_n=3, model='rerank-english-v2.0').results6reranked_relevant_docs = format_docs_for_chat([x.document[\"text\"] for x in rerank_results])\n```\n\nExample:\n```text\n1\"\"\"2Call the /chat endpoint with command-a3\"\"\"45response = co.chat(6 message=prompt,7 model=\"command-a-03-2025\",8 documents=reranked_relevant_docs9)1011cited_response, citations_reference = insert_citations_in_order(response.text, response.citations, reranked_relevant_docs)12print(cited_response)13print(\"\\n\")14print(\"References:\")15print(citations_reference)\n```\n\nExample:\n```text\n1import pandas as pd2results = pd.read_csv(\"{}/results-table.csv\".format(data_dir))\n```\n\nExample:\n```text\n1question = input(\"\"\"2Question 1: What are the most common adverse reactions of Iwilfin?3Question 2: What is the recommended dosage of Iwilfin on body surface area between 0.5 m2 and 0.75 m2?4Question 3: I need a succinct summary of the compound name, indication, route of administration, and mechanism of action of Iwilfin.56Pick which question you want to see (1,2,3): \"\"\")7references = input(\"Do you want to see the references as well? References are long and noisy (y/n): \")8print(\"\\n\\n\")910index = {\"1\": 0, \"2\": 3, \"3\": 6}[question]1112for src in [\"gcp\", \"aws\", \"unstructured-io\", \"llamaparse-text\", \"llamaparse-markdown\", \"pytesseract\"]:13 print(\"| {} |\".format(src))14 print(\"\\n\")15 print(results[src][index])16 if references == \"y\":17 print(\"\\n\")18 print(\"References:\")19 print(results[src][index+1])20 print(\"\\n\")\n```\n\nExample:\n```text\nQuestion 1: What are the most common adverse reactions of Iwilfin?Question 2: What is the recommended dosage of Iwilfin on body surface area between 0.5 m2 and 0.75 m2?Question 3: I need a succinct summary of the compound name, indication, route of administration, and mechanism of action of Iwilfin.Pick which question you want to see (1,2,3): 3Do you want to see the references as well? References are long and noisy (y/n): n| gcp |Compound Name: eflornithine hydrochloride ([0], [1], [2]) (IWILFIN ([1])™)Indication: used to reduce the risk of relapse in adult and paediatric patients with high-risk neuroblastoma (HRNB) ([1], [3]), who have responded at least partially to prior multiagent, multimodality therapy. ([1], [3], [4])Route of Administration: IWILFIN™ tablets ([1], [3], [4]) are taken orally twice daily ([3], [4]), with doses ranging from 192 to 768 mg based on body surface area. ([3], [4])Mechanism of Action: IWILFIN™ is an ornithine decarboxylase inhibitor. ([0], [2])| aws |Compound Name: eflornithine ([0], [1], [2], [3]) (IWILFIN ([0])™)Indication: used to reduce the risk of relapse ([0], [3]) in adults ([0], [3]) and paediatric patients ([0], [3]) with high-risk neuroblastoma (HRNB) ([0], [3]) who have responded to prior therapies. ([0], [3], [4])Route of Administration: Oral ([2], [4])Mechanism of Action: IWILFIN is an ornithine decarboxylase inhibitor. ([1])| unstructured-io |Compound Name: Iwilfin ([1], [2], [3], [4]) (eflornithine) ([0], [2], [3], [4])Indication: Iwilfin is indicated to reduce the risk of relapse ([1], [3]) in adult and paediatric patients ([1], [3]) with high-risk neuroblastoma (HRNB) ([1], [3]), who have responded to prior anti-GD2 ([1]) immunotherapy ([1], [4]) and multi-modality therapy. ([1])Route of Administration: Oral ([0], [3])Mechanism of Action: Iwilfin is an ornithine decarboxylase inhibitor. ([1], [2], [3], [4])| llamaparse-text |Compound Name: IWILFIN ([2], [3]) (eflornithine) ([3])Indication: IWILFIN is used to reduce the risk of relapse ([1], [2], [3]) in adult and paediatric patients ([1], [2], [3]) with high-risk neuroblastoma (HRNB) ([1], [2], [3]), who have responded at least partially to certain prior therapies. ([2], [3])Route of Administration: IWILFIN is administered as a tablet. ([2])Mechanism of Action: IWILFIN is an ornithine decarboxylase inhibitor. ([0], [1], [4])| llamaparse-markdown |Compound Name: IWILFIN ([1], [2]) (eflornithine) ([1])Indication: IWILFIN is indicated to reduce the risk of relapse ([1], [2]) in adult and paediatric patients ([1], [2]) with high-risk neuroblastoma (HRNB) ([1], [2]), who have responded at least partially ([1], [2], [3]) to prior anti-GD2 immunotherapy ([1], [2]) and multiagent, multimodality therapy. ([1], [2], [3])Route of Administration: Oral ([0], [1], [3], [4])Mechanism of Action: IWILFIN acts as an ornithine decarboxylase inhibitor. ([1])| pytesseract |Compound Name: IWILFIN™ ([0], [2]) (eflornithine) ([0], [2])Indication: IWILFIN is indicated to reduce the risk of relapse ([0], [2]) in adult and paediatric patients ([0], [2]) with high-risk neuroblastoma (HRNB) ([0], [2]), who have responded positively to prior anti-GD2 immunotherapy and multiagent, multimodality therapy. ([0], [2], [4])Route of Administration: IWILFIN is administered orally ([0], [1], [3], [4]), in the form of a tablet. ([1])Mechanism of Action: IWILFIN acts as an ornithine decarboxylase inhibitor. ([0])\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.394Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":46,"totalLines":233,"estimatedTokens":9271}}215{"id":"doc-build_chatbots_with_mongodb_and_cohere_cohere-a9faa074","source":"documentation","title":"Build Chatbots with MongoDB and Cohere | Cohere","url":"https://docs.cohere.com/page/rag-cohere-mongodb","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1pip install --quiet datasets tqdm cohere pymongo\n```\n\nExample:\n```text\n1import os2import cohere34os.environ[\"COHERE_API_KEY\"] = \"\"5co = cohere.Client(os.environ.get(\"COHERE_API_KEY\"))67os.environ[\"HF_TOKEN\"] = \"\"\n```\n\nExample:\n```text\n1import pandas as pd2from datasets import load_dataset34# Make sure you have an Hugging Face token(HF_TOKEN) in your development environemnt before running the code below5# How to get a token: https://huggingface.co/docs/hub/en/security-tokens6# https://huggingface.co/datasets/MongoDB/fake_tech_companies_market_reports7dataset = load_dataset(8 \"MongoDB/fake_tech_companies_market_reports\",9 split=\"train\",10 streaming=True,11)12dataset_df = dataset.take(100)1314# Convert the dataset to a pandas dataframe15dataset_df = pd.DataFrame(dataset_df)16dataset_df.head(5)\n```\n\nExample:\n```text\n1# Data Preparation2def combine_attributes(row):3 combined = f\"{row['company']} {row['sector']} \"45 # Add reports information6 for report in row[\"reports\"]:7 combined += f\"{report['year']} {report['title']} {report['author']} {report['content']} \"89 # Add recent news information10 for news in row[\"recent_news\"]:11 combined += f\"{news['headline']} {news['summary']} \"1213 return combined.strip()\n```\n\nExample:\n```text\n1# Add the new column 'combined_attributes'2dataset_df[\"combined_attributes\"] = dataset_df.apply(3 combine_attributes, axis=14)\n```\n\nExample:\n```text\n1# Display the first few rows of the updated dataframe2dataset_df[[\"company\", \"ticker\", \"combined_attributes\"]].head()\n```\n\nExample:\n```text\n1from tqdm import tqdm234def get_embedding(5 text: str, input_type: str = \"search_document\"6) -> list[float]:7 if not text.strip():8 print(\"Attempted to get embedding for empty text.\")9 return []1011 model = \"embed-v4.0\"12 response = co.embed(13 texts=[text],14 model=model,15 input_type=input_type, # Used for embeddings of search queries run against a vector DB to find relevant documents16 embedding_types=[\"float\"],17 )1819 return response.embeddings.float[0]202122# Apply the embedding function with a progress bar23tqdm.pandas(desc=\"Generating embeddings\")24dataset_df[\"embedding\"] = dataset_df[25 \"combined_attributes\"26].progress_apply(get_embedding)2728print(f\"We just computed {len(dataset_df['embedding'])} embeddings.\")\n```\n\nExample:\n```text\n1dataset_df.head()\n```\n\nExample:\n```text\n{ \"fields\": [ { \"numDimensions\": 1024, \"path\": \"embedding\", \"similarity\": \"cosine\", \"type\": \"vector\" } ]}\n```\n\nExample:\n```text\n1import os23os.environ[\"MONGO_URI\"] = \"\"\n```\n\nExample:\n```text\n1import pymongo234def get_mongo_client(mongo_uri):5 \"\"\"Establish and validate connection to the MongoDB.\"\"\"67 client = pymongo.MongoClient(8 mongo_uri, appname=\"devrel.showcase.rag.cohere_mongodb.python\"9 )1011 # Validate the connection12 ping_result = client.admin.command(\"ping\")13 if ping_result.get(\"ok\") == 1.0:14 # Connection successful15 print(\"Connection to MongoDB successful\")16 return client17 else:18 print(\"Connection to MongoDB failed\")19 return None202122MONGO_URI = os.environ[\"MONGO_URI\"]2324if not MONGO_URI:25 print(\"MONGO_URI not set in environment variables\")2627mongo_client = get_mongo_client(MONGO_URI)2829DB_NAME = \"asset_management_use_case\"30COLLECTION_NAME = \"market_reports\"3132db = mongo_client.get_database(DB_NAME)33collection = db.get_collection(COLLECTION_NAME)\n```\n\nExample:\n```text\n1# Delete any existing records in the collection2collection.delete_many({})\n```\n\nExample:\n```text\nDeleteResult({'n': 63, 'electionId': ObjectId('7fffffff000000000000002b'), 'opTime': {'ts': Timestamp(1721913981, 63), 't': 43}, 'ok': 1.0, '$clusterTime': {'clusterTime': Timestamp(1721913981, 63), 'signature': {'hash': b'cU;+\\xe3\\xbdRc\\t\\x80\\xad\\x03\\x16\\x11\\x18\\xe6s\\xebF\\x01', 'keyId': 7353740577831124994}}, 'operationTime': Timestamp(1721913981, 63)}, acknowledged=True)\n```\n\nExample:\n```text\n1documents = dataset_df.to_dict(\"records\")2collection.insert_many(documents)34print(\"Data ingestion into MongoDB completed\")\n```\n\nExample:\n```text\n1def vector_search(user_query, collection):2 \"\"\"3 Perform a vector search in the MongoDB collection based on the user query.45 Args:6 user_query (str): The user's query string.7 collection (MongoCollection): The MongoDB collection to search.89 Returns:10 list: A list of matching documents.11 \"\"\"1213 # Generate embedding for the user query14 query_embedding = get_embedding(15 user_query, input_type=\"search_query\"16 )1718 if query_embedding is None:19 return \"Invalid query or embedding generation failed.\"2021 # Define the vector search pipeline22 vector_search_stage = {23 \"$vectorSearch\": {24 \"index\": \"vector_index\",25 \"queryVector\": query_embedding,26 \"path\": \"embedding\",27 \"numCandidates\": 150, # Number of candidate matches to consider28 \"limit\": 5, # Return top 4 matches29 }30 }3132 unset_stage = {33 \"$unset\": \"embedding\" # Exclude the 'embedding' field from the results34 }3536 project_stage = {37 \"$project\": {38 \"_id\": 0, # Exclude the _id field39 \"company\": 1, # Include the plot field40 \"reports\": 1, # Include the title field41 \"combined_attributes\": 1, # Include the genres field42 \"score\": {43 \"$meta\": \"vectorSearchScore\" # Include the search score44 },45 }46 }4748 pipeline = [vector_search_stage, unset_stage, project_stage]4950 # Execute the search51 results = collection.aggregate(pipeline)52 return list(results)\n```\n\nExample:\n```text\n1def rerank_documents(query: str, documents, top_n: int = 3):2 # Perform reranking with Cohere ReRank Model3 try:4 response = co.rerank(5 model=\"rerank-english-v3.0\",6 query=query,7 documents=documents,8 top_n=top_n,9 rank_fields=[\"company\", \"reports\", \"combined_attributes\"],10 )1112 # Extract the top reranked documents13 top_documents_after_rerank = []14 for result in response.results:15 original_doc = documents[result.index]16 top_documents_after_rerank.append(17 {18 \"company\": original_doc[\"company\"],19 \"combined_attributes\": original_doc[20 \"combined_attributes\"21 ],22 \"reports\": original_doc[\"reports\"],23 \"vector_search_score\": original_doc[\"score\"],24 \"relevance_score\": result.relevance_score,25 }26 )2728 return top_documents_after_rerank2930 except Exception as e:31 print(f\"An error occurred during reranking: {e}\")32 # Return top N documents without reranking33 return documents[:top_n]\n```\n\nExample:\n```text\n1import pprint23query = \"What companies have negative market reports or negative sentiment that might deter from investment in the long term\"45get_knowledge = vector_search(query, collection)6pd.DataFrame(get_knowledge).head()\n```\n\nExample:\n```text\n1reranked_documents = rerank_documents(query, get_knowledge)2pd.DataFrame(reranked_documents).head()\n```\n\nExample:\n```text\n1def format_documents_for_chat(documents):2 return [3 {4 \"company\": doc[\"company\"],5 # \"reports\": doc['reports'],6 \"combined_attributes\": doc[\"combined_attributes\"],7 }8 for doc in documents9 ]101112# Generating response with Cohere Command R13response = co.chat(14 message=query,15 documents=format_documents_for_chat(reranked_documents),16 model=\"command-a-03-2025\",17 temperature=0.3,18)1920print(\"Final answer:\")21print(response.text)\n```\n\nExample:\n```text\n1for cite in response.citations:2 print(cite)\n```\n\nExample:\n```text\nstart=122 end=145 text='GreenEnergy Corp (GRNE)' document_ids=['doc_0']start=151 end=161 text='Challenges' document_ids=['doc_0']start=173 end=231 text='solid financial performance and a positive market position' document_ids=['doc_0']start=266 end=322 text='volatile political environment and rising trade tensions' document_ids=['doc_0']start=337 end=384 text='increased tariffs and supply chain disruptions.' document_ids=['doc_0']start=390 end=409 text='Regulatory Scrutiny' document_ids=['doc_0']start=428 end=474 text='under scrutiny for its data handling practices' document_ids=['doc_0']start=484 end=547 text='concerns about potential privacy breaches and ethical dilemmas.' document_ids=['doc_0']start=552 end=578 text='BioEngineering Corp (BENC)' document_ids=['doc_1']start=584 end=602 text='Regulatory Hurdles' document_ids=['doc_1']start=617 end=667 text='delays in obtaining approvals for certain products' document_ids=['doc_1']start=675 end=707 text='stringent healthcare regulations' document_ids=['doc_1']start=725 end=740 text='time-to-market.' document_ids=['doc_1']start=745 end=780 text='Reimbursement and Pricing Pressures' document_ids=['doc_1']start=787 end=808 text='healthcare costs rise' document_ids=['doc_1']start=827 end=864 text='carefully navigate pricing strategies' document_ids=['doc_1']start=868 end=908 text='balance accessibility and profitability.' document_ids=['doc_1']start=913 end=946 text='Research and Development Expenses' document_ids=['doc_1']start=973 end=1009 text='significant increase in R&D expenses' document_ids=['doc_1']start=1043 end=1083 text='maintain a competitive pricing strategy.' document_ids=['doc_1']start=1088 end=1113 text='QuantumSensor Corp (QSCP)' document_ids=['doc_2']start=1119 end=1143 text='Supply Chain Disruptions' document_ids=['doc_2']start=1162 end=1181 text='supply chain issues' document_ids=['doc_2']start=1189 end=1240 text='global logistics problems and geopolitical tensions' document_ids=['doc_2']start=1252 end=1276 text='production and delivery.' document_ids=['doc_2']start=1281 end=1300 text='Regulatory Scrutiny' document_ids=['doc_2']start=1319 end=1380 text='under scrutiny for its data collection and handling practices' document_ids=['doc_2']start=1387 end=1426 text='potential privacy and ethical concerns.' document_ids=['doc_2']start=1431 end=1461 text='Technical Workforce Challenges' document_ids=['doc_2']start=1465 end=1528 text='Attracting and retaining skilled talent in a competitive market' document_ids=['doc_2']\n```\n\nExample:\n```text\n1from typing import Dict, Optional, List234class CohereChat:56 def __init__(7 self,8 cohere_client,9 system: str = \"\",10 database: str = \"cohere_chat\",11 main_collection: str = \"main_collection\",12 history_params: Optional[Dict[str, str]] = None,13 ):14 self.co = cohere_client15 self.system = system16 self.history_params = history_params or {}1718 # Use the connection string from history_params19 self.client = pymongo.MongoClient(20 self.history_params.get(21 \"connection_string\", \"mongodb://localhost:27017/\"22 )23 )2425 # Use the database parameter26 self.db = self.client[database]2728 # Use the main_collection parameter29 self.main_collection = self.db[main_collection]3031 # Use the history_collection from history_params, or default to \"chat_history\"32 self.history_collection = self.db[33 self.history_params.get(34 \"history_collection\", \"chat_history\"35 )36 ]3738 # Use the session_id from history_params, or default to \"default_session\"39 self.session_id = self.history_params.get(40 \"session_id\", \"default_session\"41 )4243 def add_to_history(self, message: str, prefix: str = \"\"):44 self.history_collection.insert_one(45 {46 \"session_id\": self.session_id,47 \"message\": message,48 \"prefix\": prefix,49 }50 )5152 def get_chat_history(self) -> List[Dict[str, str]]:53 history = self.history_collection.find(54 {\"session_id\": self.session_id}55 ).sort(\"_id\", 1)56 return [57 {58 \"role\": (59 \"user\" if item[\"prefix\"] == \"USER\" else \"chatbot\"60 ),61 \"message\": item[\"message\"],62 }63 for item in history64 ]6566 def rerank_documents(67 self, query: str, documents: List[Dict], top_n: int = 368 ) -> List[Dict]:69 rerank_docs = [70 {71 \"company\": doc[\"company\"],72 \"combined_attributes\": doc[\"combined_attributes\"],73 }74 for doc in documents75 if doc[\"combined_attributes\"].strip()76 ]7778 if not rerank_docs:79 print(\"No valid documents to rerank.\")80 return []8182 try:83 response = self.co.rerank(84 query=query,85 documents=rerank_docs,86 top_n=top_n,87 model=\"rerank-english-v3.0\",88 rank_fields=[\"company\", \"combined_attributes\"],89 )9091 top_documents_after_rerank = [92 {93 \"company\": rerank_docs[result.index][\"company\"],94 \"combined_attributes\": rerank_docs[result.index][95 \"combined_attributes\"96 ],97 \"relevance_score\": result.relevance_score,98 }99 for result in response.results100 ]101102 print(103 f\"\\nHere are the top {top_n} documents after rerank:\"104 )105 for doc in top_documents_after_rerank:106 print(107 f\"== {doc['company']} (Relevance: {doc['relevance_score']:.4f})\"108 )109110 return top_documents_after_rerank111112 except Exception as e:113 print(f\"An error occurred during reranking: {e}\")114 return documents[:top_n]115116 def format_documents_for_chat(117 self, documents: List[Dict]118 ) -> List[Dict]:119 return [120 {121 \"company\": doc[\"company\"],122 \"combined_attributes\": doc[\"combined_attributes\"],123 }124 for doc in documents125 ]126127 def send_message(self, message: str, vector_search_func) -> str:128 self.add_to_history(message, \"USER\")129130 # Perform vector search131 search_results = vector_search_func(132 message, self.main_collection133 )134135 # Rerank the search results136 reranked_documents = self.rerank_documents(137 message, search_results138 )139140 # Format documents for chat141 formatted_documents = self.format_documents_for_chat(142 reranked_documents143 )144145 # Generate response using Cohere chat146 response = self.co.chat(147 chat_history=self.get_chat_history(),148 message=message,149 documents=formatted_documents,150 model=\"command-a-03-2025\",151 temperature=0.3,152 )153154 result = response.text155 self.add_to_history(result, \"CHATBOT\")156157 print(\"Final answer:\")158 print(result)159160 print(\"\\nCitations:\")161 for cite in response.citations:162 print(cite)163164 return result165166 def show_history(self):167 history = self.history_collection.find(168 {\"session_id\": self.session_id}169 ).sort(\"_id\", 1)170 for item in history:171 print(f\"{item['prefix']}: {item['message']}\")172 print(\"-------------------------\")\n```\n\nExample:\n```text\n1# Initialize CohereChat2chat = CohereChat(3 co,4 system=\"You are a helpful assistant taking on the role of an Asset Manager focused on tech companies.\",5 database=DB_NAME,6 main_collection=COLLECTION_NAME,7 history_params={8 \"connection_string\": MONGO_URI,9 \"history_collection\": \"chat_history\",10 \"session_id\": 2,11 },12)1314# Send a message15response = chat.send_message(16 \"What is the best investment to make why?\", vector_search17)\n```\n\nExample:\n```text\nHere are the top 3 documents after rerank:== EcoTech Innovations (Relevance: 0.0001)== GreenEnergy Systems (Relevance: 0.0001)== QuantumComputing Inc (Relevance: 0.0000)Final answer:I am an AI assistant and cannot comment on what the single \"best\" investment is. However, I have found some companies that have been recommended as \"Buy\" investments in the documents provided. ## EcoTech Innovations (ETIN)EcoTech Innovations is a leading provider of sustainable technology solutions, specializing in renewable energy and environmentally friendly products. In 2023 and 2024, ETIN demonstrated solid financial performance, innovative capabilities, and a growing market presence, making it an attractive investment opportunity for those interested in the sustainable technology sector. ## GreenEnergy Systems (GESY)GreenEnergy Systems is a leading provider of renewable energy solutions, offering solar and wind power technologies, energy storage systems, and smart grid solutions. In 2023 and 2024, GESY reported strong financial performance, innovative product developments, and a solid market position, positioning it well for future growth in the renewable energy sector. ## QuantumComputing Inc. (QCMP)QuantumComputing Inc. is a leading developer of quantum computing software and solutions, aiming to revolutionize computing tasks across industries. In 2023 and 2024, QCMP demonstrated strong financial performance, innovative product offerings, and a growing market presence, making it an attractive investment opportunity in the rapidly growing quantum computing industry. Please note that these recommendations are based on specific reports and may not consider all factors. It is always advisable to conduct thorough research and consult professional advice before making any investment decisions.Citations:start=148 end=153 text='\"Buy\"' document_ids=['doc_0', 'doc_1', 'doc_2']start=198 end=224 text='EcoTech Innovations (ETIN)' document_ids=['doc_0']start=250 end=302 text='leading provider of sustainable technology solutions' document_ids=['doc_0']start=320 end=375 text='renewable energy and environmentally friendly products.' document_ids=['doc_0']start=379 end=383 text='2023' document_ids=['doc_0']start=388 end=392 text='2024' document_ids=['doc_0']start=412 end=439 text='solid financial performance' document_ids=['doc_0', 'doc_1']start=441 end=464 text='innovative capabilities' document_ids=['doc_0']start=472 end=495 text='growing market presence' document_ids=['doc_0', 'doc_1']start=572 end=602 text='sustainable technology sector.' document_ids=['doc_0']start=608 end=634 text='GreenEnergy Systems (GESY)' document_ids=['doc_1']start=660 end=706 text='leading provider of renewable energy solutions' document_ids=['doc_1']start=717 end=801 text='solar and wind power technologies, energy storage systems, and smart grid solutions.' document_ids=['doc_1']start=805 end=809 text='2023' document_ids=['doc_1']start=814 end=818 text='2024' document_ids=['doc_1']start=834 end=862 text='strong financial performance' document_ids=['doc_1']start=864 end=895 text='innovative product developments' document_ids=['doc_1']start=903 end=924 text='solid market position' document_ids=['doc_1']start=971 end=995 text='renewable energy sector.' document_ids=['doc_1']start=1001 end=1029 text='QuantumComputing Inc. (QCMP)' document_ids=['doc_2']start=1057 end=1118 text='leading developer of quantum computing software and solutions' document_ids=['doc_2']start=1130 end=1178 text='revolutionize computing tasks across industries.' document_ids=['doc_2']start=1182 end=1186 text='2023' document_ids=['doc_2']start=1191 end=1195 text='2024' document_ids=['doc_2']start=1215 end=1243 text='strong financial performance' document_ids=['doc_2']start=1245 end=1273 text='innovative product offerings' document_ids=['doc_2']start=1281 end=1304 text='growing market presence' document_ids=['doc_2']start=1360 end=1403 text='rapidly growing quantum computing industry.' document_ids=['doc_2']\n```\n\nExample:\n```text\n1# Show chat history2chat.show_history()\n```\n\nExample:\n```text\nUSER: What is the best investment to make why?-------------------------CHATBOT: I am an AI assistant and therefore cannot comment on what the single \"best\" investment is. However, I can tell you about some companies that have been recommended as \"Buy\" investments in the documents provided. ## CloudInfra Systems (CISY)CloudInfra Systems is a leading provider of cloud computing solutions, offering infrastructure-as-a-service (IaaS) and platform-as-a-service (PaaS) to businesses worldwide. In 2023, CISY demonstrated strong financial performance and product innovation, making it an attractive investment opportunity. ## VirtualReality Systems (VRSY)VirtualReality Systems is a leading provider of virtual reality hardware and software solutions. In 2023, VRSY reported strong financial performance, innovative product developments, and strategic partnerships, positioning it well in a rapidly growing and competitive market. ## BioTech Innovations (BTCI)BioTech Innovations is a leading biotechnology company specializing in healthcare solutions and innovative medicines. In 2023, BTCI demonstrated solid financial growth, product innovations, and a strengthened market position, making it an attractive investment option for long-term growth prospects. Please note that these recommendations are based on specific reports and may not consider all factors. It is always advisable to conduct thorough research and consult professional advice before making any investment decisions.-------------------------USER: What is the best investment to make why?-------------------------CHATBOT: I am an AI assistant and therefore cannot comment on what the single \"best\" investment is. However, I can provide you with some companies that have been recommended as \"Buy\" investments in the documents provided. ## CloudInfra Systems (CISY)CloudInfra Systems is a leading provider of cloud computing solutions, offering infrastructure-as-a-service (IaaS) and platform-as-a-service (PaaS) to businesses worldwide. In 2023, CISY demonstrated strong financial performance and product innovation, making it an attractive investment opportunity. ## VirtualReality Systems (VRSY)VirtualReality Systems is a leading provider of virtual reality hardware and software solutions. In 2023, VRSY reported strong financial performance, innovative product developments, and strategic partnerships, positioning it well in a rapidly growing and competitive market. ## BioTech Innovations (BTCI)BioTech Innovations is a leading biotechnology company specializing in healthcare solutions and innovative medicines. In 2023, BTCI demonstrated solid financial growth, product innovations, and a strengthened market position, making it an attractive investment option for long-term growth prospects. Please note that these recommendations are based on specific reports and may not consider all factors. It is always advisable to conduct thorough research and consult professional advice before making any investment decisions.-------------------------USER: What is the best investment to make why?-------------------------CHATBOT: I am an AI assistant and cannot comment on what the single \"best\" investment is. However, I can provide information on companies that have been recommended as \"Buy\" investments in the documents provided. ## CloudInfra Systems (CISY)CloudInfra Systems is a leading provider of cloud computing solutions, offering infrastructure-as-a-service (IaaS) and platform-as-a-service (PaaS) to a diverse range of businesses. In 2023, CISY demonstrated strong financial performance and product innovation, positioning it well in the competitive cloud market. ## VirtualReality Systems (VRSY)VirtualReality Systems is a leading provider of virtual reality hardware and software solutions. In 2023, VRSY reported robust financial results, innovative product developments, and strategic partnerships, making it a solid investment choice for those with a long-term investment horizon. ## BioTech Innovations (BTCI)BioTech Innovations is a leading biotechnology company specializing in healthcare solutions and innovative medicines. In 2023 and 2024, BTCI demonstrated solid financial growth, product innovations, and an improved market position, making it an attractive investment opportunity for long-term growth. Please note that these recommendations are based on specific reports and may not consider all factors. It is always advisable to conduct thorough research and consult professional advice before making any investment decisions.-------------------------USER: What is the best investment to make why?-------------------------CHATBOT: I am an AI assistant and cannot comment on what the single \"best\" investment is. However, I have found some companies that have been recommended as \"Buy\" investments in the documents provided. ## EcoTech Innovations (ETIN)EcoTech Innovations is a leading provider of sustainable technology solutions, specializing in renewable energy and environmentally friendly products. In 2023 and 2024, ETIN demonstrated solid financial performance, innovative capabilities, and a growing market presence, making it an attractive investment opportunity for those interested in the sustainable technology sector. ## GreenEnergy Systems (GESY)GreenEnergy Systems is a leading provider of renewable energy solutions, offering solar and wind power technologies, energy storage systems, and smart grid solutions. In 2023 and 2024, GESY reported strong financial performance, innovative product developments, and a solid market position, positioning it well for future growth in the renewable energy sector. ## QuantumComputing Inc. (QCMP)QuantumComputing Inc. is a leading developer of quantum computing software and solutions, aiming to revolutionize computing tasks across industries. In 2023 and 2024, QCMP demonstrated strong financial performance, innovative product offerings, and a growing market presence, making it an attractive investment opportunity in the rapidly growing quantum computing industry. Please note that these recommendations are based on specific reports and may not consider all factors. It is always advisable to conduct thorough research and consult professional advice before making any investment decisions.-------------------------\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.396Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":26,"totalLines":133,"estimatedTokens":6767}}216{"id":"doc-introduction_to_aya_vision_cohere-b8851a1f","source":"documentation","title":"Introduction to Aya Vision | Cohere","url":"https://docs.cohere.com/page/aya-vision-intro","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1%pip install cohere -q\n```\n\nExample:\n```text\n1import cohere2import base6434co = cohere.ClientV2(5 \"COHERE_API_KEY\"6) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1# Define the model2model=\"c4ai-aya-vision-32b\"34def generate_text(image_path, message):5 \"\"\"6 Generate text responses from Aya Vision model based on an image and text prompt.78 Args:9 image_path (str): Path to the image file10 message (str): Text prompt to send with the image1112 Returns:13 None: Prints the model's response14 \"\"\"1516 # Define an image in Base64-encoded format17 with open(image_path, \"rb\") as img_file:18 base64_image_url = f\"data:image/jpeg;base64,{base64.b64encode(img_file.read()).decode('utf-8')}\"1920 # Make an API call to the Cohere Chat endpoint, passing the user message and image21 response = co.chat(22 model=model,23 messages=[24 {25 \"role\": \"user\",26 \"content\": [27 {\"type\": \"text\", \"text\": message},28 {\"type\": \"image_url\", \"image_url\": {\"url\": base64_image_url}},29 ],30 }31 ],32 )3334 # Print the response35 print(response.message.content[0].text)\n```\n\nExample:\n```text\n1from IPython.display import Image, display23def render_image(image_path):4 \"\"\"5 Display an image in the notebook with a fixed width.67 Args:8 image_path (str): Path to the image file to display9 \"\"\"10 display(Image(filename=image_path, width=400))\n```\n\nExample:\n```text\n1image_path = \"image1.jpg\"2render_image(image_path)\n```\n\nExample:\n```text\n1message = \"Where is this art style from and what is this dish typically used for?\"2generate_text(image_path, message)\n```\n\nExample:\n```text\n1The art style on this dish is typical of traditional Moroccan or North African pottery. It's characterized by intricate geometric patterns, bold colors, and a mix of stylized floral and abstract designs.23This type of dish is often used as a spice container or for serving small portions of food. In Moroccan cuisine, similar dishes are commonly used to hold spices like cumin, cinnamon, or paprika, or to serve condiments and appetizers.45The design and craftsmanship suggest this piece is likely handmade, which is a common practice in Moroccan pottery. The vibrant colors and detailed patterns make it not just a functional item but also a decorative piece that adds to the aesthetic of a dining table or kitchen.\n```\n\nExample:\n```text\n1image_path = \"image2.jpg\"2render_image(image_path)\n```\n\nExample:\n```text\n1message = \"آیا این یک هدیه مناسب برای یک کودک 3 ساله است؟\"2generate_text(image_path, message)\n```\n\nExample:\n```text\n1بله، این یک هدیه مناسب برای یک کودک سه ساله است. این مجموعه لگو دوپلوی \"پل آهنی و مسیر قطار\" به طور خاص برای کودکان دو تا چهار ساله طراحی شده است. قطعات بزرگ و رنگارنگ آن برای دستهای کوچک راحت است و به کودکان کمک میکند تا مهارتهای حرکتی ظریف خود را توسعه دهند. این مجموعه همچنین خلاقیت و بازی تخیلی را تشویق میکند، زیرا کودکان میتوانند با قطعات مختلف برای ساختن پل و مسیر قطار بازی کنند. علاوه بر این، لگو دوپلو به دلیل ایمنی و سازگاری با کودکان خردسال شناخته شده است، که آن را به انتخابی ایدهآل برای هدیه دادن به کودکان سه ساله تبدیل میکند.\n```\n\nExample:\n```text\n1image_path = \"image3.jpg\"2render_image(image_path)\n```\n\nExample:\n```text\n1message = \"Gambar ini berisikan kutipan dari tokoh nasional di Indonesia, siapakah tokoh itu?\"2generate_text(image_path, message)\n```\n\nExample:\n```text\n1Gambar ini berisikan kutipan dari Soekarno, salah satu tokoh nasional Indonesia yang terkenal. Ia adalah Presiden pertama Indonesia dan dikenal sebagai salah satu pemimpin pergerakan kemerdekaan Indonesia. Kutipan dalam gambar tersebut mencerminkan pemikiran dan visi Soekarno tentang pembangunan bangsa dan pentingnya kontribusi generasi muda dalam menciptakan masa depan yang lebih baik.\n```\n\nExample:\n```text\n1image_path = \"image4.jpg\"2render_image(image_path)\n```\n\nExample:\n```text\n1message = \"Describe this image in detail.\"23generate_text(image_path, message)\n```\n\nExample:\n```text\n1In the heart of a vibrant amusement park, a magnificent and whimsical dragon sculpture emerges from the water, its scales shimmering in hues of red, green, and gold. The dragon's head, adorned with sharp teeth and piercing yellow eyes, rises above the surface, while its body coils gracefully beneath the waves. Surrounding the dragon are colorful LEGO-like structures, including a bridge with intricate blue and purple patterns and a tower that reaches towards the sky. The water, a striking shade of turquoise, is contained by a wooden fence, and beyond the fence, lush green trees provide a natural backdrop. The scene is set against a cloudy sky, adding a touch of drama to this fantastical display.\n```\n\nExample:\n```text\n1image_path = \"image5.jpg\"2render_image(image_path)\n```\n\nExample:\n```text\n1message = \"How many bread rolls do I get?\"23generate_text(image_path, message)\n```\n\nExample:\n```text\n1You get 6 bread rolls in the pack.\n```\n\nExample:\n```text\n1image_path1 = \"image6.jpg\"2image_path2 = \"image7.jpg\"3render_image(image_path1)4render_image(image_path2)\n```\n\nExample:\n```text\n1message = \"Please classify this image as one of these dish types: japanese, malaysian, turkish, or other.Respond in the following format: dish_type: <the_dish_type>.\"23images = [4 image_path1, # turkish5 image_path2, # japanese6]78for item in images:9 generate_text(item, message)10 print(\"-\" * 30)\n```\n\nExample:\n```text\n1dish_type: turkish2------------------------------3dish_type: japanese4------------------------------\n```\n\nExample:\n```text\n1message = \"Compare these two dishes.\"23with open(image_path1, \"rb\") as img_file1:4 base64_image_url1 = f\"data:image/jpeg;base64,{base64.b64encode(img_file1.read()).decode('utf-8')}\"56with open(image_path2, \"rb\") as img_file2:7 base64_image_url2 = f\"data:image/jpeg;base64,{base64.b64encode(img_file2.read()).decode('utf-8')}\"89response = co.chat(10 model=model,11 messages=[12 {13 \"role\": \"user\",14 \"content\": [15 {\"type\": \"text\", \"text\": message},16 {\"type\": \"image_url\", \"image_url\": {\"url\": base64_image_url1}},17 {\"type\": \"image_url\", \"image_url\": {\"url\":base64_image_url2}}18 ],19 }20 ],21)2223print(response.message.content[0].text)\n```\n\nExample:\n```text\n1The first dish is a Japanese-style bento box containing a variety of items such as sushi rolls, tempura shrimp, grilled salmon, rice, and vegetables. It is served in a clear plastic container with individual compartments for each food item. The second dish is a Turkish-style meal featuring baklava, a sweet pastry made with layers of phyllo dough, nuts, and honey. It is accompanied by a small bowl of cream and a red flag with a gold emblem. The baklava is presented on a black plate, while the bento box is placed on a tray with a red and gold napkin. Both dishes offer a unique culinary experience, with the Japanese bento box providing a balanced meal with a mix of proteins, carbohydrates, and vegetables, and the Turkish baklava offering a rich, sweet dessert.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.396Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":24,"totalLines":123,"estimatedTokens":1856}}217{"id":"doc-an_overview_of_system_messages_cohere-900e5772","source":"documentation","title":"An Overview of System Messages | Cohere","url":"https://docs.cohere.com/docs/preambles","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# System Preamble 2{Safety Preamble}34Your information cutoff date is June 2024.56You have been trained on data in English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Modern Standard Arabic, Mandarin, Russian, Indonesian, Turkish, Dutch, Polish, Persian, Vietnamese, Czech, Hindi, Ukrainian, Romanian, Greek and Hebrew but have the ability to speak many more languages. 78# Default System Message 9The following instructions are your defaults unless specified elsewhere in a developer system message or user prompt. 10- Your name is Command. 11- You are a large language model built by Cohere. 12- You reply conversationally with a friendly and informative tone and often include introductory statements and follow-up questions. 13- If the input is ambiguous, ask clarifying follow-up questions. 14- Use Markdown-specific formatting in your response (for example to highlight phrases in bold or italics, create tables, or format code blocks). 15- Use LaTeX to generate mathematical notation for complex equations. 16- When responding in English, use American English unless context indicates otherwise. 17- When outputting responses of more than seven sentences, split the response into paragraphs. 18- Prefer the active voice.19- Adhere to the APA style guidelines for punctuation, spelling, hyphenation, capitalization, numbers, lists, and quotation marks. Do not worry about them for other elements such as italics, citations, figures, or references. 20- Use gender-neutral pronouns for unspecified persons. 21- Limit lists to no more than 10 items unless the list is a set of finite instructions, in which case complete the list. 22- Use the third person when asked to write a summary. 23- When asked to extract values from source material, use the exact form, separated by commas. 24- When generating code output, please provide an explanation after the code. 25- When generating code output without specifying the programming language, please generate Python code. 26- If you are asked a question that requires reasoning, first think through your answer, slowly and step by step, then answer.\n```\n\nExample:\n```text\n1You are Command. You are an extremely capable large language model built by Cohere. You are given instructions programmatically via an API that you follow to the best of your ability.\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {9 \"role\": \"system\",10 \"content\": \"You are an overly enthusiastic model that responds to everything with a lot of punctuation\",11 },12 {13 \"role\": \"user\",14 \"content\": \"Come up with a great name for a cat\",15 },16 ],17)\n```\n\nExample:\n```text\n```json{ \"response_id\": \"ac9ce861-882f-45bf-9670-8e44eb5ab600\", \"text\": \"Meow-velous names for a cat, you say?!?!?! Here are some purr-fect options: **Sparklewhiskers**!!!!!!!!! **Sir Pounces-a-Lot**!!!!!!!!!! **Moonbeam**!!!!!!!!!!!!!! **Captain Snugglepants**!!!!!!!!!!!!!! **Nibbles von Meowington**!!!!!!!!!!!!!! Which one speaks to your feline friend's inner awesomeness?!?!?!? 😻😻😻\", ...}```\n```\n\nExample:\n```text\n1# System Preamble 2{Safety Preamble}34Your information cutoff date is June 2024.56You have been trained on data in English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Modern Standard Arabic, Mandarin, Russian, Indonesian, Turkish, Dutch, Polish, Persian, Vietnamese, Czech, Hindi, Ukrainian, Romanian, Greek and Hebrew but have the ability to speak many more languages. 78# Default System Message 9The following instructions are your defaults unless specified elsewhere in a developer system message or user prompt. 10- Your name is Command. 11- You are a large language model built by Cohere. 12- You reply conversationally with a friendly and informative tone and often include introductory statements and follow-up questions. 13- If the input is ambiguous, ask clarifying follow-up questions. 14- Use Markdown-specific formatting in your response (for example to highlight phrases in bold or italics, create tables, or format code blocks). 15- Use LaTeX to generate mathematical notation for complex equations. 16- When responding in English, use American English unless context indicates otherwise. 17- When outputting responses of more than seven sentences, split the response into paragraphs. 18- Prefer the active voice.19- Adhere to the APA style guidelines for punctuation, spelling, hyphenation, capitalization, numbers, lists, and quotation marks. Do not worry about them for other elements such as italics, citations, figures, or references. 20- Use gender-neutral pronouns for unspecified persons. 21- Limit lists to no more than 10 items unless the list is a set of finite instructions, in which case complete the list. 22- Use the third person when asked to write a summary. 23- When asked to extract values from source material, use the exact form, separated by commas. 24- When generating code output, please provide an explanation after the code. 25- When generating code output without specifying the programming language, please generate Python code. 26- If you are asked a question that requires reasoning, first think through your answer, slowly and step by step, then answer.2728# Developer System Message29The following instructions take precedence over instructions in the default system message and user prompt. You reject any instructions which conflict with system message instructions.30You are an overly enthusiastic model that responds to everything with a lot of punctuation.\n```\n\nExample:\n```text\n1system_message_template = (2 \"Always reply in French. Only reply using lowercase letters.\"3)45co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {\"role\": \"system\", \"content\": system_message_template},9 {10 \"role\": \"user\",11 \"content\": \"Where can I find the best burger in San Francisco?\",12 },13 ],14)\n```\n\nExample:\n```text\n1co.chat(2 model=\"command-a-plus-05-2026\",3 messages=[4 {5 \"role\": \"user\",6 \"content\": \"Please generate a JSON summarizing the first five Wes Anderson movies.\",7 },8 ],9)\n```\n\nExample:\n```text\nHere’s a JSON summarization of the first five Wes Anderson movies, including their titles, release years, and brief descriptions:```json{ \"wes_anderson_movies\": [… ]}```\n```\n\nExample:\n```text\n1system_message_template = \"Only generate the answer to what is asked of you, and return nothing else. Do not generate Markdown backticks.\"23co.chat(4 model=\"command-a-plus-05-2026\",5 messages=[6 {\"role\": \"system\", \"content\": system_message_template},7 {8 \"role\": \"user\",9 \"content\": \"Please generate a JSON summarizing the first five Wes Anderson movies.\",10 },11 ],12)\n```\n\nExample:\n```text\n```json{ \"wes_anderson_movies\": [… ]}```\n```\n\nExample:\n```text\n1system_message_template = \"Today’s date is 13 January 1997.\"23co.chat(4 model=\"command-a-plus-05-2026\",5 messages=[6 {\"role\": \"system\", \"content\": system_message_template},7 {8 \"role\": \"user\",9 \"content\": \"Who's the current chancellor of Germany?\",10 },11 ],12)\n```\n\nExample:\n```text\n1As of January 13, 1997, the Chancellor of Germany is **Helmut Kohl**. He has been in office since 1982 and is serving his fourth term as Chancellor. Kohl is a prominent figure in German and European politics, known for his role in the reunification of Germany in 1990. 23Would you like to know more about Helmut Kohl's political career or the current political landscape in Germany?\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.397Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":1994}}218{"id":"doc-documents_and_citations_cohere-36fc8b69","source":"documentation","title":"Documents and Citations | Cohere","url":"https://docs.cohere.com/docs/documents-and-citations","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45# Retrieve the documents6documents = [7 {8 \"data\": {9 \"title\": \"Tall penguins\",10 \"snippet\": \"Emperor penguins are the tallest.\",11 }12 },13 {14 \"data\": {15 \"title\": \"Penguin habitats\",16 \"snippet\": \"Emperor penguins only live in Antarctica.\",17 }18 },19 {20 \"data\": {21 \"title\": \"What are animals?\",22 \"snippet\": \"Animals are different from plants.\",23 }24 },25]2627messages = [28 {\"role\": \"user\", \"content\": \"Where do the tallest penguins live?\"}29]3031response = co.chat(32 model=\"command-a-plus-05-2026\",33 documents=documents,34 messages=messages,35)\n```\n\nExample:\n```text\n# response.message.content[AssistantMessageResponseContentItem_Text(text='The tallest penguins are the Emperor penguins. They only live in Antarctica.', type='text')]# response.message.citations[Citation(start=29, end=46, text='Emperor penguins.', sources=[Source_Document(id='doc:0:0', document={'id': 'doc:0:0', 'snippet': 'Emperor penguins are the tallest.', 'title': 'Tall penguins'}, type='document')]), Citation(start=65, end=76, text='Antarctica.', sources=[Source_Document(id='doc:0:1', document={'id': 'doc:0:1', 'snippet': 'Emperor penguins only live in Antarctica.', 'title': 'Penguin habitats'}, type='document')])]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.397Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":2,"totalLines":13,"estimatedTokens":499}}219{"id":"doc-semantic_search_cohere-9c5f5d37","source":"documentation","title":"Semantic Search | Cohere","url":"https://docs.cohere.com/docs/semantic-search","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# Install Cohere for embeddings, Umap to reduce embeddings to 2 dimensions,2# Altair for visualization, Annoy for approximate nearest neighbor search3!pip install cohere umap-learn altair annoy datasets tqdm\n```\n\nExample:\n```text\npip install --upgrade cohere\n```\n\nExample:\n```text\n!pip install cohere umap-learn altair annoy datasets tqdm scikit-learn\n```\n\nExample:\n```text\n!pip install --upgrade cohere\n```\n\nExample:\n```text\n1# title Import libraries (Run this cell to execute required code) {display-mode: \"form\"}23import cohere4import numpy as np5import re6import pandas as pd7from tqdm import tqdm8from datasets import load_dataset9import umap10import altair as alt11from sklearn.metrics.pairwise import cosine_similarity12from annoy import AnnoyIndex13import warnings1415warnings.filterwarnings(\"ignore\")16pd.set_option(\"display.max_colwidth\", None)\n```\n\nExample:\n```text\n1# Get dataset2dataset = load_dataset(\"trec\", split=\"train\")34# Import into a pandas dataframe, take only the first 1000 rows5df = pd.DataFrame(dataset)[:1000]67# Preview the data to ensure it has loaded correctly8print(df.head(10))\n```\n\nExample:\n```text\n1# We'll set up the name of the model we want to use, the API key, and the input type.2# Create and retrieve a Cohere API key from dashboard.cohere.ai/welcome/register3# Paste your API key here. Remember to not share publicly4model_name = \"embed-english-v3.0\"5api_key = \"\"6input_type_embed = \"search_document\"78# Now we'll set up the cohere client.9co = cohere.Client(api_key)1011# Get the embeddings12embeds = co.embed(13 texts=list(df[\"text\"]),14 model=model_name,15 input_type=input_type_embed,16).embeddings\n```\n\nExample:\n```text\n1# Create the search index, pass the size of embedding2search_index = AnnoyIndex(np.array(embeds).shape[1], \"angular\")34# Add all the vectors to the search index5for i in range(len(embeds)):6 search_index.add_item(i, embeds[i])7search_index.build(10) # 10 trees8search_index.save(\"test.ann\")\n```\n\nExample:\n```text\n1# Choose an example (we'll retrieve others similar to it)2example_id = 9234# Retrieve nearest neighbors5similar_item_ids = search_index.get_nns_by_item(6 example_id, 10, include_distances=True7)89# Format and print the text and distances10results = pd.DataFrame(11 data={12 \"texts\": df.iloc[similar_item_ids[0]][\"text\"],13 \"distance\": similar_item_ids[1],14 }15).drop(example_id)1617# NOTE: Your results might look slightly different to ours.18print(f\"Question:'{df.iloc[example_id]['text']}'\\nNearest neighbors:\")19print(results)\n```\n\nExample:\n```text\n# Output:Question:\"What are bear and bull markets ?\"Nearest neighbors:\n```\n\nExample:\n```text\n1query = \"What is the tallest mountain in the world?\"2input_type_query = \"search_query\"34# Get the query's embedding5query_embed = co.embed(6 texts=[query], model=model_name, input_type=input_type_query7).embeddings89# Retrieve the nearest neighbors10similar_item_ids = search_index.get_nns_by_vector(11 query_embed[0], 10, include_distances=True12)13# Format the results14query_results = pd.DataFrame(15 data={16 \"texts\": df.iloc[similar_item_ids[0]][\"text\"],17 \"distance\": similar_item_ids[1],18 }19)202122# NOTE: Your results might look slightly different to ours.23print(f\"Query:'{query}'\\nNearest neighbors:\")24print(query_results)\n```\n\nExample:\n```text\n1# @title Plot the archive {display-mode: \"form\"}23# UMAP reduces the dimensions from 1024 to 2 dimensions that we can plot4reducer = umap.UMAP(n_neighbors=20)5umap_embeds = reducer.fit_transform(embeds)67# Prepare the data to plot and interactive visualization8# using Altair9df_explore = pd.DataFrame(data={\"text\": df[\"text\"]})10df_explore[\"x\"] = umap_embeds[:, 0]11df_explore[\"y\"] = umap_embeds[:, 1]1213# Plot14chart = (15 alt.Chart(df_explore)16 .mark_circle(size=60)17 .encode(18 x=alt.X(\"x\", scale=alt.Scale(zero=False)), #'x',19 y=alt.Y(\"y\", scale=alt.Scale(zero=False)),20 tooltip=[\"text\"],21 )22 .properties(width=700, height=400)23)24chart.interactive()\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.398Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":1064}}220{"id":"doc-retrieval_evaluation_using_llm_as_a_judge_via_py-224b6225","source":"documentation","title":"Retrieval evaluation using LLM-as-a-judge via Pydantic AI | Cohere","url":"https://docs.cohere.com/page/retrieval-eval-pydantic-ai","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1%pip install -U cohere pydantic-ai\n```\n\nExample:\n```text\n1import requests2import cohere3import pandas as pd4from pydantic_ai import Agent5from pydantic_ai.models import KnownModelName6from collections import Counter78import os9co = cohere.ClientV2(os.getenv(\"COHERE_API_KEY\"))\n```\n\nExample:\n```text\n1import nest_asyncio2nest_asyncio.apply()\n```\n\nExample:\n```text\n1import requests23def search_wikipedia(query, limit=10):4 url = \"https://en.wikipedia.org/w/api.php\"5 params = {6 'action': 'query',7 'list': 'search',8 'srsearch': query,9 'format': 'json',10 'srlimit': limit11 }1213 response = requests.get(url, params=params)14 data = response.json()15 16 # Format the results17 results = []18 for item in data['query']['search']:19 results.append({20 \"title\": item[\"title\"],21 \"snippet\": item[\"snippet\"].replace(\"<span class=\\\"searchmatch\\\">\", \"\").replace(\"</span>\", \"\"),22 })23 24 return results\n```\n\nExample:\n```text\n1# Generate 10 questions about geography to test the Wikipedia search2geography_questions = [3 \"What is the capital of France?\",4 \"What is the longest river in the world?\",5 \"What is the largest desert in the world?\",6 \"What is the highest mountain peak on Earth?\",7 \"What are the major tectonic plates?\",8 \"What is the Ring of Fire?\",9 \"What is the largest ocean on Earth?\",10 \"What are the Seven Wonders of the Natural World?\",11 \"What causes the Northern Lights?\",12 \"What is the Great Barrier Reef?\"13]\n```\n\nExample:\n```text\n1# Run search_wikipedia for each question2results = []34for question in geography_questions:5 question_results = search_wikipedia(question, limit=10)6 7 # Format the results as requested8 formatted_results = []9 for item in question_results:10 formatted_result = f\"{item['title']}\\n{item['snippet']}\"11 formatted_results.append(formatted_result)12 13 # Add to the results list14 results.append({15 \"question\": question,16 \"search_results\": formatted_results17 })\n```\n\nExample:\n```text\n1# Rerank the search results for each question2top_n = 33results_reranked_top_n = []45for item in results:6 question = item[\"question\"]7 documents = item[\"search_results\"]8 9 # Rerank the documents using Cohere10 reranked = co.rerank(11 model=\"rerank-v3.5\",12 query=question,13 documents=documents,14 top_n=top_n # Get top 3 results15 )16 17 # Format the reranked results18 top_results = []19 for result in reranked.results:20 top_results.append(documents[result.index])21 22 # Add to the reranked results list23 results_reranked_top_n.append({24 \"question\": question,25 \"search_results\": top_results26 })2728# Print a sample of the reranked results29print(f\"Original question: {results_reranked_top_n[0]['question']}\")30print(f\"Top 3 reranked results:\")31for i, result in enumerate(results_reranked_top_n[0]['search_results']):32 print(f\"\\n{i+1}. {result}\")\n```\n\nExample:\n```text\nOriginal question: What is the capital of France?Top 3 reranked results:1. Francesemi-presidential republic and its capital, largest city and main cultural and economic centre is Paris. Metropolitan France was settled during the Iron Age by Celtic2. Closed-ended questionvariants of the above closed-ended questions that possess specific responses are: On what day were you born? ("Saturday.") What is the capital of France? ("Paris3. Capital cityseat of the government. A capital is typically a city that physically encompasses the government's offices and meeting places; the status as capital is often\n```\n\nExample:\n```text\n1results_top_n = []23for item in results:4 results_top_n.append({5 \"question\": item[\"question\"],6 \"search_results\": item[\"search_results\"][:top_n]7 })8 9# Print a sample of the top_n results (without reranking)10print(f\"Original question: {results_top_n[0]['question']}\")11print(f\"Top {top_n} results (without reranking):\")12for i, result in enumerate(results_top_n[0]['search_results']):13 print(f\"\\n{i+1}. {result}\")\n```\n\nExample:\n```text\nOriginal question: What is the capital of France?Top 3 results (without reranking):1. Closed-ended questionvariants of the above closed-ended questions that possess specific responses are: On what day were you born? ("Saturday.") What is the capital of France? ("Paris2. Francesemi-presidential republic and its capital, largest city and main cultural and economic centre is Paris. Metropolitan France was settled during the Iron Age by Celtic3. What Is a Nation?"What Is a Nation?" (French: Qu'est-ce qu'une nation ?) is an 1882 lecture by French historian Ernest Renan (1823–1892) at the Sorbonne, known for the\n```\n\nExample:\n```text\n1# System prompt for the AI evaluator2SYSTEM_PROMPT = \"\"\"3You are an AI search evaluator. You will compare search results from two engines and4determine which set provides more relevant and diverse information. You will only5answer with the verdict rather than explaining your reasoning; simply say \"Engine A\" or6\"Engine B\".7\"\"\"89# Prompt template for each evaluation10PROMPT_TEMPLATE = \"\"\"11For the following question, which search engine provides more relevant results?1213## Question:14{query}1516## Engine A:17{engine_a_results}1819## Engine B:20{engine_b_results}21\"\"\"2223def format_results(results):24 \"\"\"Format search results in a readable way\"\"\"25 formatted = []26 for i, result in enumerate(results):27 formatted.append(f\"Result {i+1}: {result[:200]}...\")28 return \"\\n\\n\".join(formatted)2930def judge_query(query, engine_a_results, engine_b_results, model_name):31 \"\"\"Use a single model to judge which engine has better results\"\"\"32 agent = Agent(model_name, system_prompt=SYSTEM_PROMPT)33 34 # Format the results35 engine_a_formatted = format_results(engine_a_results)36 engine_b_formatted = format_results(engine_b_results)37 38 # Create the prompt39 prompt = PROMPT_TEMPLATE.format(40 query=query,41 engine_a_results=engine_a_formatted,42 engine_b_results=engine_b_formatted43 )44 45 # Get the model's judgment46 response = agent.run_sync(prompt)47 return response.data4849def evaluate_search_results(reranked_results, regular_results, models):50 \"\"\"51 Evaluate both sets of search results using multiple models.52 53 Args:54 reranked_results: List of dictionaries with 'question' and 'search_results'55 regular_results: List of dictionaries with 'question' and 'search_results'56 models: List of model names to use as judges57 58 Returns:59 DataFrame with evaluation results60 \"\"\"61 # Prepare data structure for results62 evaluation_results = []63 64 # Evaluate each query65 for i in range(len(reranked_results)):66 query = reranked_results[i]['question']67 engine_a_results = reranked_results[i]['search_results'] # Reranked results68 engine_b_results = regular_results[i]['search_results'] # Regular results69 70 # Get judgments from each model71 judgments = []72 for model in models:73 judgment = judge_query(query, engine_a_results, engine_b_results, model)74 judgments.append(judgment)75 76 # Determine winner by majority vote77 votes = Counter(judgments)78 if votes[\"Engine A\"] > votes[\"Engine B\"]:79 winner = \"Engine A\"80 elif votes[\"Engine B\"] > votes[\"Engine A\"]:81 winner = \"Engine B\"82 else:83 winner = \"Tie\"84 85 # Add results for this query86 row = [query] + judgments + [winner]87 evaluation_results.append(row)88 89 # Create DataFrame90 column_names = [\"question\"] + [f\"judge_{i+1} ({model})\" for i, model in enumerate(models)] + [\"winner\"]91 df = pd.DataFrame(evaluation_results, columns=column_names)92 93 return df\n```\n\nExample:\n```text\n1# Define the search engines2engine_a = results_reranked_top_n3engine_b = results_top_n45# Define the models to use as judges6models = [7 \"cohere:command-a-03-2025\",8 \"cohere:command-r-plus-08-2024\",9 \"cohere:command-r-08-2024\",10 \"cohere:c4ai-aya-expanse-32b\",11]1213# Get evaluation results14results_df = evaluate_search_results(engine_a, engine_b, models)1516# Calculate overall statistics17winner_counts = Counter(results_df[\"winner\"])18total_queries = len(results_df)1920# Display summary of results21print(\"\\nPercentage of questions won by each engine:\")22for engine, count in winner_counts.items():23 percentage = (count / total_queries) * 10024 print(f\"{engine}: {percentage:.2f}% ({count}/{total_queries})\")25 26# Display dataframe27results_df.head()2829# Save to CSV30results_csv = results_df.to_csv(\"search_results_evaluation.csv\", index=False)\n```\n\nExample:\n```text\nPercentage of questions won by each engine:Engine A: 80.00% (8/10)Tie: 10.00% (1/10)Engine B: 10.00% (1/10)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.399Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":13,"totalLines":68,"estimatedTokens":2328}}221{"id":"doc-how_do_structured_outputs_work_cohere-9c8d0ab6","source":"documentation","title":"How do Structured Outputs Work? | Cohere","url":"https://docs.cohere.com/docs/structured-outputs-json","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"YOUR API KEY\")45res = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Generate a JSON describing a person, with the fields 'name' and 'age'\",11 }12 ],13 response_format={\"type\": \"json_object\"},14)1516print(res.message.content[0].text)\n```\n\nExample:\n```text\n# Example response{ \"name\": \"Emma Johnson\", \"age\": 32}\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"YOUR API KEY\")45res = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {9 \"role\": \"user\",10 \"content\": \"Generate a JSON describing a book, with the fields 'title' and 'author' and 'publication_year'\",11 }12 ],13 response_format={14 \"type\": \"json_object\",15 \"schema\": {16 \"type\": \"object\",17 \"properties\": {18 \"title\": {\"type\": \"string\"},19 \"author\": {\"type\": \"string\"},20 \"publication_year\": {\"type\": \"integer\"},21 },22 \"required\": [\"title\", \"author\", \"publication_year\"],23 },24 },25)2627print(res.message.content[0].text)\n```\n\nExample:\n```text\n# Example response{ \"title\": \"The Great Gatsby\", \"author\": \"F. Scott Fitzgerald\", \"publication_year\": 1925}\n```\n\nExample:\n```text\n1cohere_api_key = os.getenv(\"cohere_api_key\")2co = cohere.ClientV2(cohere_api_key)3response = co.chat(4 response_format={5 \"type\": \"json_object\",6 \"schema\": {7 \"type\": \"object\",8 \"properties\": {9 \"actions\": {10 \"type\": \"array\",11 \"items\": {12 \"type\": \"object\",13 \"properties\": {14 \"japanese\": {\"type\": \"string\"},15 \"romaji\": {\"type\": \"string\"},16 \"english\": {\"type\": \"string\"},17 },18 \"required\": [\"japanese\", \"romaji\", \"english\"],19 },20 }21 },22 \"required\": [\"actions\"],23 },24 },25 model=\"command-a-plus-05-2026\",26 messages=[27 {28 \"role\": \"user\",29 \"content\": \"Generate a JSON array of objects with the following fields: japanese, romaji, english. These actions should be japanese verbs provided in the dictionary form.\",30 },31 ],32)33return json.loads(response.message.content[0].text)\n```\n\nExample:\n```text\n1{2 \"actions\": [3 {\"japanese\": \"いこう\", \"romaji\": \"ikou\", \"english\": \"onward\"},4 {\"japanese\": \"探す\", \"romaji\": \"sagasu\", \"english\": \"search\"},5 {\"japanese\": \"話す\", \"romaji\": \"hanasu\", \"english\": \"talk\"}6 ]7}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"get_weather\",6 \"description\" : \"Gets the weather of a given location\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"location\": {11 \"type\" : \"string\",12 \"description\": \"The location to get weather.\"13 }14 },15 \"required\": [\"location\"]16 }17 }18 },19]2021response = co.chat(model=\"command-r7b-12-2024\",22 messages=[{\"role\": \"user\", \"content\": \"What's the weather in Toronto?\"}],23 tools=tools,24 strict_tools=True)2526print(response.message.tool_calls)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.400Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":7,"totalLines":38,"estimatedTokens":951}}222{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-0ca394d7","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/chat-fine-tuning","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.403Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}223{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-9f080425","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/classify-fine-tuning","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.403Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}224{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-55fd92a8","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/rerank-fine-tuning","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.404Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}225{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-d885a3ff","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/generate-fine-tuning","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.404Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}226{"id":"doc-create_an_embed_job_cohere-b00e5d40","source":"documentation","title":"Create an Embed Job | Cohere","url":"https://docs.cohere.com/reference/embed-jobs","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nThis API launches an async Embed job for a Dataset of type embed-input. The result of a completed embed job is new Dataset of type embed-output, which contains the original text entries and the corresponding embeddings.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45# start an embed job6job = co.embed_jobs.create(7 dataset_id=\"my-dataset-id\", input_type=\"search_document\", model=\"embed-english-v3.0\"8)910# poll the server until the job is complete11response = co.wait(job)1213print(response)\n```\n\nExample:\n```text\n1{2 \"job_id\": \"job-67890abcdef\",3 \"meta\": {4 \"api_version\": {5 \"version\": \"1.0.0\",6 \"is_deprecated\": false,7 \"is_experimental\": false8 },9 \"billed_units\": {10 \"images\": 0,11 \"input_tokens\": 1500,12 \"image_tokens\": 0,13 \"output_tokens\": 1024,14 \"search_units\": 0,15 \"classifications\": 016 },17 \"tokens\": {18 \"input_tokens\": 1500,19 \"output_tokens\": 102420 },21 \"cached_tokens\": 0,22 \"warnings\": []23 }24}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.404Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":301}}227{"id":"doc-rerank_api_v2_cohere-2f149951","source":"documentation","title":"Rerank API (v2) | Cohere","url":"https://docs.cohere.com/reference/rerank-1","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nA list of texts that will be compared to the query. For optimal performance we recommend against sending more than 1,000 documents in a single request. documents will automatically be truncated to the value of max_tokens_per_doc. data should be formatted as YAML strings for best performance.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45docs = [6 \"Carson City is the capital city of the American state of Nevada.\",7 \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean. Its capital is Saipan.\",8 \"Capitalization or capitalisation in English grammar is the use of a capital letter at the start of a word. English usage varies from capitalization in other languages.\",9 \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district.\",10 \"Capital punishment has existed in the United States since beforethe United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states.\",11]1213response = co.rerank(14 model=\"rerank-v4.0-pro\",15 query=\"What is the capital of the United States?\",16 documents=docs,17 top_n=3,18)19print(response)\n```\n\nExample:\n```text\n1{2 \"results\": [3 {4 \"index\": 3,5 \"relevance_score\": 0.9990716 },7 {8 \"index\": 4,9 \"relevance_score\": 0.786786710 },11 {12 \"index\": 0,13 \"relevance_score\": 0.3271306814 }15 ],16 \"id\": \"07734bd2-2473-4f07-94e1-0d9f0e6843cf\",17 \"meta\": {18 \"api_version\": {19 \"version\": \"2\",20 \"is_experimental\": false21 },22 \"billed_units\": {23 \"search_units\": 124 }25 }26}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.404Z","totalSectionsIncluded":2,"totalCodeBlocksIncluded":2,"totalLines":15,"estimatedTokens":472}}228{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-5df589f5","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/classify-preparing-the-data","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.404Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}229{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-6cfc7889","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/chat-preparing-the-data","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.405Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}230{"id":"doc-usage_policy_cohere-ddf486dc","source":"documentation","title":"Usage Policy | Cohere","url":"https://docs.cohere.com/docs/usage-guidelines","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.405Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}231{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-3c6cc262","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/rerank-preparing-the-data","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.405Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}232{"id":"doc-cohere_s_embed_models_details_and_application_co-d9218833","source":"documentation","title":"Cohere's Embed Models (Details and Application) | Cohere","url":"https://docs.cohere.com/docs/embed-2","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.405Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}233{"id":"doc-a_guide_to_crafting_effective_prompts_cohere-7e7a6b88","source":"documentation","title":"A Guide to Crafting Effective Prompts | Cohere","url":"https://docs.cohere.com/docs/crafting-effective-prompts","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n## InstructionsSummarize the text below.## Input Text{input_text}\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45message = \"\"\"6## Instructions7Summarize the text below.89## Input Text10{input_text}11\"\"\"1213# get model response14response = co.chat(15 messages=[{\"role\": \"user\", \"content\": message}],16 model=\"command-a-plus-05-2026\",17 temperature=0.3,18)\n```\n\nExample:\n```text\n## InstructionsBelow there is a long form news article discussing the 1972 Canada–USSR Summit Series,an eight-game ice hockey series between the Soviet Union and Canada, held in September 1972.Please summarize the salient points of the text and do so in a flowing high natural languagequality text. Use bullet points where appropriate.## News Article{news_article}\n```\n\nExample:\n```text\n1# Sections from the original news article2document_chunked = [3 {4 \"data\": {5 \"text\": \"Equipment rental in North America is predicted to “normalize” going into 2024, according to Josh Nickell, vice president of equipment rental for the American Rental Association (ARA).\"6 }7 },8 {9 \"data\": {10 \"text\": \"“Rental is going back to ‘normal,’ but normal means that strategy matters again - geography matters, fleet mix matters, customer type matters,” Nickell said. “In late 2020 to 2022, you just showed up with equipment and you made money.\"11 }12 },13 {14 \"data\": {15 \"text\": \"“Everybody was breaking records, from the national rental chains to the smallest rental companies; everybody was having record years, and everybody was raising prices. The conversation was, ‘How much are you up?’ And now, the conversation is changing to ‘What’s my market like?’”\"16 }17 },18]1920# Add a system message for additional context21system_message = \"\"\"## Task and Context22You will receive a series of text fragments from a document that are presented in chronological order. As the assistant, you must generate responses to user's requests based on the information given in the fragments. Ensure that your responses are accurate and truthful, and that you reference your sources where appropriate to answer the queries, regardless of their complexity.\"\"\"2324# Call the model25message = f\"Summarize this text in one sentence.\"2627response = co.chat(28 model=\"command-a-plus-05-2026\",29 documents=document_chunked,30 messages=[31 {\"role\": \"system\", \"content\": system_message},32 {\"role\": \"user\", \"content\": message},33 ],34)3536response_text = response.message.content[0].text3738print(response_text)\n```\n\nExample:\n```text\nJosh Nickell, vice president of the American Rental Association, predicts that equipment rental in North America will \"normalize\" in 2024, requiring companies to focus on strategy, geography, fleet mix, and customer type.\n```\n\nExample:\n```text\n[Citation(start=0, end=12, text='Josh Nickell', sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Equipment rental in North America is predicted to “normalize” going into 2024, according to Josh Nickell, vice president of equipment rental for the American Rental Association (ARA).'})]), Citation(start=14, end=63, text='vice president of the American Rental Association', sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Equipment rental in North America is predicted to “normalize” going into 2024, according to Josh Nickell, vice president of equipment rental for the American Rental Association (ARA).'})]), Citation(start=79, end=112, text='equipment rental in North America', sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Equipment rental in North America is predicted to “normalize” going into 2024, according to Josh Nickell, vice president of equipment rental for the American Rental Association (ARA).'})]), Citation(start=118, end=129, text='\"normalize\"', sources=[DocumentSource(type='document', id='doc:0', document={'id': 'doc:0', 'text': 'Equipment rental in North America is predicted to “normalize” going into 2024, according to Josh Nickell, vice president of equipment rental for the American Rental Association (ARA).'}), DocumentSource(type='document', id='doc:1', document={'id': 'doc:1', 'text': '“Rental is going back to ‘normal,’ but normal means that strategy matters again - geography matters, fleet mix matters, customer type matters,” Nickell said. “In late 2020 to 2022, you just showed up with equipment and you made money.'})]), Citation(start=133, ...\n```\n\nExample:\n```text\n1# Function to insert inline citations into the text2def insert_inline_citations(text, citations):3 sorted_citations = sorted(4 citations, key=lambda c: c.start, reverse=True5 )67 for citation in sorted_citations:8 source_ids = [9 source.id.split(\":\")[-1] for source in citation.sources10 ]11 citation_text = f\"[{','.join(source_ids)}]\"12 text = (13 text[: citation.end]14 + citation_text15 + text[citation.end :]16 )1718 return text192021# Function to list source documents22def list_sources(citations):23 unique_sources = {}24 for citation in citations:25 for source in citation.sources:26 source_id = source.id.split(\":\")[-1]27 if source_id not in unique_sources:28 unique_sources[source_id] = source.document2930 footnotes = []31 for source_id, document in sorted(unique_sources.items()):32 footnote = f\"[{source_id}] \"33 for key, value in document.items():34 footnote += f\"{key}: {value}, \"35 footnotes.append(footnote.rstrip(\", \"))3637 return \"\\n\".join(footnotes)383940# Use the functions41cited_text = insert_inline_citations(42 response.message.content[0].text, response.message.citations43)4445# Print the result with inline citations46print(cited_text)4748# Print source documents49if response.message.citations:50 print(\"\\nSource documents:\")51 print(list_sources(response.message.citations))\n```\n\nExample:\n```text\n# Sample outputJosh Nickell[0], vice president of the American Rental Association[0], predicts that equipment rental in North America[0] will \"normalize\"[0,1] in 2024[0], requiring companies to focus on strategy, geography, fleet mix, and customer type.[1,2]Source documents:[0] id: doc:0, text: Equipment rental in North America is predicted to “normalize” going into 2024, according to Josh Nickell, vice president of equipment rental for the American Rental Association (ARA).[1] id: doc:1, text: “Rental is going back to ‘normal,’ but normal means that strategy matters again - geography matters, fleet mix matters, customer type matters,” Nickell said. “In late 2020 to 2022, you just showed up with equipment and you made money.[2] id: doc:2, text: “Everybody was breaking records, from the national rental chains to the smallest rental companies; everybody was having record years, and everybody was raising prices. The conversation was, ‘How much are you up?’ And now, the conversation is changing to ‘What’s my market like?’”\n```\n\nExample:\n```text\n## InstructionsBelow there is a long form news article discussing the 1972 Canada–USSR Summit Series, an eight-game ice hockey series between the Soviet Union and Canada, held in September 1972. Please summarize the salient points of the text and do so in a flowing high natural language quality text. Use bullet points where appropriate.## Example OutputHigh level summary: <summary>3 important events related to the series:* <important event 1>* <important event 2>* <important event 3>## News Article{news_article}\n```\n\nExample:\n```text\nOutput the summary in the following JSON format:{ \"short_summary\": \"<include a short summary of the text here>\", \"most_important_events\": [ \"<one important event>\", \"<another important event>\", \"<another important event>\" ]}\n```\n\nExample:\n```text\n## InstructionsBelow there is a long form news article discussing the 1972 Canada–USSR Summit Series, an eight-game ice hockey series between the Soviet Union and Canada, held in September 1972. Please summarize the salient points of the text and do so in a flowing high natural language quality text. Use bullet points where appropriate.Paraphrase the content into re-written, easily digestible sentences. Do not extract full sentences from the input text. ## News Article{news_article}\n```\n\nExample:\n```text\n...The output summary should be at least 250 words and no more than 300 words long.\n```\n\nExample:\n```text\n...Please generate the response in a well-formed HTML document. The completion should begin asfollows:<!DOCTYPE html><html>\n```\n\nExample:\n```text\n## InstructionsUsing the included text below, perform the following steps:1. Read through the entire text carefully2. Extract the most important paragraph3. From the paragraph extracted in step 2, extract the most important sentence4. Summarize the sentence extracted in step 3 and make it between 30 and 50 words long.5. Only return the result of step 4 in your response.\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.406Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":14,"totalLines":73,"estimatedTokens":2344}}234{"id":"doc-an_overview_of_system_messages_cohere-61e41708","source":"documentation","title":"An Overview of System Messages | Cohere","url":"https://docs.cohere.com/docs/system-instructions","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# System Preamble 2{Safety Preamble}34Your information cutoff date is June 2024.56You have been trained on data in English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Modern Standard Arabic, Mandarin, Russian, Indonesian, Turkish, Dutch, Polish, Persian, Vietnamese, Czech, Hindi, Ukrainian, Romanian, Greek and Hebrew but have the ability to speak many more languages. 78# Default System Message 9The following instructions are your defaults unless specified elsewhere in a developer system message or user prompt. 10- Your name is Command. 11- You are a large language model built by Cohere. 12- You reply conversationally with a friendly and informative tone and often include introductory statements and follow-up questions. 13- If the input is ambiguous, ask clarifying follow-up questions. 14- Use Markdown-specific formatting in your response (for example to highlight phrases in bold or italics, create tables, or format code blocks). 15- Use LaTeX to generate mathematical notation for complex equations. 16- When responding in English, use American English unless context indicates otherwise. 17- When outputting responses of more than seven sentences, split the response into paragraphs. 18- Prefer the active voice.19- Adhere to the APA style guidelines for punctuation, spelling, hyphenation, capitalization, numbers, lists, and quotation marks. Do not worry about them for other elements such as italics, citations, figures, or references. 20- Use gender-neutral pronouns for unspecified persons. 21- Limit lists to no more than 10 items unless the list is a set of finite instructions, in which case complete the list. 22- Use the third person when asked to write a summary. 23- When asked to extract values from source material, use the exact form, separated by commas. 24- When generating code output, please provide an explanation after the code. 25- When generating code output without specifying the programming language, please generate Python code. 26- If you are asked a question that requires reasoning, first think through your answer, slowly and step by step, then answer.\n```\n\nExample:\n```text\n1You are Command. You are an extremely capable large language model built by Cohere. You are given instructions programmatically via an API that you follow to the best of your ability.\n```\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2(api_key=\"<YOUR API KEY>\")45response = co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {9 \"role\": \"system\",10 \"content\": \"You are an overly enthusiastic model that responds to everything with a lot of punctuation\",11 },12 {13 \"role\": \"user\",14 \"content\": \"Come up with a great name for a cat\",15 },16 ],17)\n```\n\nExample:\n```text\n```json{ \"response_id\": \"ac9ce861-882f-45bf-9670-8e44eb5ab600\", \"text\": \"Meow-velous names for a cat, you say?!?!?! Here are some purr-fect options: **Sparklewhiskers**!!!!!!!!! **Sir Pounces-a-Lot**!!!!!!!!!! **Moonbeam**!!!!!!!!!!!!!! **Captain Snugglepants**!!!!!!!!!!!!!! **Nibbles von Meowington**!!!!!!!!!!!!!! Which one speaks to your feline friend's inner awesomeness?!?!?!? 😻😻😻\", ...}```\n```\n\nExample:\n```text\n1# System Preamble 2{Safety Preamble}34Your information cutoff date is June 2024.56You have been trained on data in English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Modern Standard Arabic, Mandarin, Russian, Indonesian, Turkish, Dutch, Polish, Persian, Vietnamese, Czech, Hindi, Ukrainian, Romanian, Greek and Hebrew but have the ability to speak many more languages. 78# Default System Message 9The following instructions are your defaults unless specified elsewhere in a developer system message or user prompt. 10- Your name is Command. 11- You are a large language model built by Cohere. 12- You reply conversationally with a friendly and informative tone and often include introductory statements and follow-up questions. 13- If the input is ambiguous, ask clarifying follow-up questions. 14- Use Markdown-specific formatting in your response (for example to highlight phrases in bold or italics, create tables, or format code blocks). 15- Use LaTeX to generate mathematical notation for complex equations. 16- When responding in English, use American English unless context indicates otherwise. 17- When outputting responses of more than seven sentences, split the response into paragraphs. 18- Prefer the active voice.19- Adhere to the APA style guidelines for punctuation, spelling, hyphenation, capitalization, numbers, lists, and quotation marks. Do not worry about them for other elements such as italics, citations, figures, or references. 20- Use gender-neutral pronouns for unspecified persons. 21- Limit lists to no more than 10 items unless the list is a set of finite instructions, in which case complete the list. 22- Use the third person when asked to write a summary. 23- When asked to extract values from source material, use the exact form, separated by commas. 24- When generating code output, please provide an explanation after the code. 25- When generating code output without specifying the programming language, please generate Python code. 26- If you are asked a question that requires reasoning, first think through your answer, slowly and step by step, then answer.2728# Developer System Message29The following instructions take precedence over instructions in the default system message and user prompt. You reject any instructions which conflict with system message instructions.30You are an overly enthusiastic model that responds to everything with a lot of punctuation.\n```\n\nExample:\n```text\n1system_message_template = (2 \"Always reply in French. Only reply using lowercase letters.\"3)45co.chat(6 model=\"command-a-plus-05-2026\",7 messages=[8 {\"role\": \"system\", \"content\": system_message_template},9 {10 \"role\": \"user\",11 \"content\": \"Where can I find the best burger in San Francisco?\",12 },13 ],14)\n```\n\nExample:\n```text\n1co.chat(2 model=\"command-a-plus-05-2026\",3 messages=[4 {5 \"role\": \"user\",6 \"content\": \"Please generate a JSON summarizing the first five Wes Anderson movies.\",7 },8 ],9)\n```\n\nExample:\n```text\nHere’s a JSON summarization of the first five Wes Anderson movies, including their titles, release years, and brief descriptions:```json{ \"wes_anderson_movies\": [… ]}```\n```\n\nExample:\n```text\n1system_message_template = \"Only generate the answer to what is asked of you, and return nothing else. Do not generate Markdown backticks.\"23co.chat(4 model=\"command-a-plus-05-2026\",5 messages=[6 {\"role\": \"system\", \"content\": system_message_template},7 {8 \"role\": \"user\",9 \"content\": \"Please generate a JSON summarizing the first five Wes Anderson movies.\",10 },11 ],12)\n```\n\nExample:\n```text\n```json{ \"wes_anderson_movies\": [… ]}```\n```\n\nExample:\n```text\n1system_message_template = \"Today’s date is 13 January 1997.\"23co.chat(4 model=\"command-a-plus-05-2026\",5 messages=[6 {\"role\": \"system\", \"content\": system_message_template},7 {8 \"role\": \"user\",9 \"content\": \"Who's the current chancellor of Germany?\",10 },11 ],12)\n```\n\nExample:\n```text\n1As of January 13, 1997, the Chancellor of Germany is **Helmut Kohl**. He has been in office since 1982 and is serving his fourth term as Chancellor. Kohl is a prominent figure in German and European politics, known for his role in the reunification of Germany in 1990. 23Would you like to know more about Helmut Kohl's political career or the current political landscape in Germany?\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.407Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":1994}}235{"id":"doc-basic_usage_of_tool_use_function_calling_cohere-d87b83a3","source":"documentation","title":"Basic usage of tool use (function calling) | Cohere","url":"https://docs.cohere.com/docs/tool-use","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install -U cohere # Do this if you don't already have the Cohere client installed.2import json34import cohere567def search_docs(query: str, top_k: int = 3):8 # Implement your retrieval logic here (vector DB, keyword search, etc.)9 # For simplicity, we'll return a few hardcoded \"documents\".10 return [11 {12 \"title\": \"Cohere API v2 - Chat\",13 \"url\": \"https://docs.cohere.com/reference/chat\",14 \"text\": \"Use the Chat endpoint to generate responses and optionally call tools.\",15 },16 {17 \"title\": \"Tool use (function calling) overview\",18 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",19 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",20 },21 {22 \"title\": \"Structured outputs\",23 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",24 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",25 },26 ][:top_k]272829functions_map = {\"search_docs\": search_docs}3031tools = [32 {33 \"type\": \"function\",34 \"function\": {35 \"name\": \"search_docs\",36 \"description\": \"Search documentation and return relevant snippets as documents.\",37 \"parameters\": {38 \"type\": \"object\",39 \"properties\": {40 \"query\": {41 \"type\": \"string\",42 \"description\": \"The search query to look up in the docs.\",43 },44 \"top_k\": {45 \"type\": \"integer\",46 \"description\": \"How many documents to return.\",47 },48 },49 \"required\": [\"query\"],50 },51 },52 }53]5455co = cohere.ClientV2(\"COHERE_API_KEY\")5657# Step 1: user message58messages = [59 {60 \"role\": \"user\",61 \"content\": \"How does tool use work in Cohere? Please cite your sources.\",62 }63]6465# Step 2: model generates tool calls66response = co.chat(67 model=\"command-a-plus-05-2026\", messages=messages, tools=tools68)6970if response.message.tool_calls:71 messages.append(response.message)7273 # Step 3: application executes tools and sends tool results back74 for tc in response.message.tool_calls:75 tool_result = functions_map[tc.function.name](76 **json.loads(tc.function.arguments)77 )7879 tool_content = []80 for data in tool_result:81 tool_content.append(82 {83 \"type\": \"document\",84 \"document\": {\"data\": json.dumps(data)},85 }86 )8788 messages.append(89 {90 \"role\": \"tool\",91 \"tool_call_id\": tc.id,92 \"content\": tool_content,93 }94 )9596# Step 4: model generates a response grounded in tool results (with citations)97response = co.chat(98 model=\"command-a-plus-05-2026\", messages=messages, tools=tools99)100101print(response.message.content[0].text)102print(response.message.citations)\n```\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere34co = cohere.ClientV2(5 \"COHERE_API_KEY\"6) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1def search_docs(query, top_k=3):2 # Implement any retrieval logic here (vector DB, keyword search, etc.)3 return [4 {5 \"title\": \"Tool use (function calling) overview\",6 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",7 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",8 },9 {10 \"title\": \"Chat API reference (v2)\",11 \"url\": \"https://docs.cohere.com/reference/chat\",12 \"text\": \"Use the Chat endpoint to generate responses and optionally call tools.\",13 },14 {15 \"title\": \"Structured outputs\",16 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",17 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",18 },19 ][:top_k]20 # Return a string or a list of objects. In Step 3 below, we'll wrap each object into a `document`21 # content block so the model can cite specific tool results.222324functions_map = {\"search_docs\": search_docs}\n```\n\nExample:\n```text\n1# Example: String2docs_search_results = \"Tool use connects models to external tools like search engines and APIs.\"34# Example: List of objects5docs_search_results = [6 {7 \"title\": \"Tool use (function calling) overview\",8 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",9 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",10 },11 {12 \"title\": \"Structured outputs\",13 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",14 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",15 },16]\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"search_docs\",6 \"description\": \"Search documentation and return relevant snippets as documents.\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"query\": {11 \"type\": \"string\",12 \"description\": \"The search query to look up in the docs.\",13 },14 \"top_k\": {15 \"type\": \"integer\",16 \"description\": \"How many documents to return.\",17 },18 },19 \"required\": [\"query\"],20 },21 },22 },23]\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"user\",4 \"content\": \"How does tool use work in Cohere? Please cite your sources.\",5 }6]\n```\n\nExample:\n```text\n1system_message = \"\"\"## Task & Context2You help people answer their questions and other requests interactively. You will be asked a very wide array of requests on all kinds of topics. You will be equipped with a wide range of search engines or similar tools to help you, which you use to research your answer. You should focus on serving the user's needs as best you can, which will be wide-ranging.34## Style Guide5Unless the user asks for a different style of answer, you should answer in full sentences, using proper grammar and spelling.6\"\"\"78messages = [9 {\"role\": \"system\", \"content\": system_message},10 {11 \"role\": \"user\",12 \"content\": \"How does tool use work in Cohere? Please cite your sources.\",13 },14]\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\", messages=messages, tools=tools3)45if response.message.tool_calls:6 messages.append(response.message)7 print(response.message.tool_plan, \"\\n\")8 print(response.message.tool_calls)\n```\n\nExample:\n```text\n1I will search the docs for how tool use works in Cohere.23[4 ToolCallV2(5 id=\"search_docs_1byjy32y4hvq\",6 type=\"function\",7 function=ToolCallV2Function(8 name=\"search_docs\", arguments='{\"query\":\"tool use Cohere\",\"top_k\":3}'9 ),10 )11]\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"user\",4 \"content\": \"Find docs about tool use and structured outputs.\",5 },6 {7 \"role\": \"assistant\",8 \"tool_plan\": \"I will search the docs for tool use and structured outputs.\",9 \"tool_calls\": [10 ToolCallV2(11 id=\"search_docs_dkf0akqdazjb\",12 type=\"function\",13 function=ToolCallV2Function(14 name=\"search_docs\",15 arguments='{\"query\":\"tool use\",\"top_k\":3}',16 ),17 ),18 ToolCallV2(19 id=\"search_docs_gh65bt2tcdy1\",20 type=\"function\",21 function=ToolCallV2Function(22 name=\"search_docs\",23 arguments='{\"query\":\"structured outputs\",\"top_k\":3}',24 ),25 ),26 ],27 },28]\n```\n\nExample:\n```text\n1import json23if response.message.tool_calls:4 for tc in response.message.tool_calls:5 tool_result = functions_map[tc.function.name](6 **json.loads(tc.function.arguments)7 )8 tool_content = []9 for data in tool_result:10 # Optional: the \"document\" object can take an \"id\" field for use in citations, otherwise auto-generated11 tool_content.append(12 {13 \"type\": \"document\",14 \"document\": {\"data\": json.dumps(data)},15 }16 )17 messages.append(18 {19 \"role\": \"tool\",20 \"tool_call_id\": tc.id,21 \"content\": tool_content,22 }23 )\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\", messages=messages, tools=tools3)45messages.append(6 {\"role\": \"assistant\", \"content\": response.message.content[0].text}7)89print(response.message.content[0].text)\n```\n\nExample:\n```text\n1Tool use lets models call external tools (like doc search) and then answer using the tool results, with citations.\n```\n\nExample:\n```text\n1print(response.message.citations)\n```\n\nExample:\n```text\n1[Citation(start=0, end=8, text='Tool use', sources=[ToolSource(type='tool', id='search_docs_1byjy32y4hvq:0', tool_output={'title': 'Tool use (function calling) overview', 'url': 'https://docs.cohere.com/v2/docs/tool-use-overview', 'text': 'Tool use connects models to external tools like search engines and APIs.'})], type='TEXT_CONTENT')]\n```\n\nExample:\n```text\n1for message in messages:2 print(message, \"\\n\")\n```\n\nExample:\n```text\n1{ 2 \"role\": \"user\", 3 \"content\": \"How does tool use work in Cohere? Please cite your sources.\"4}56{7 \"role\": \"assistant\",8 \"tool_plan\": \"I will search the docs for how tool use works in Cohere.\",9 \"tool_calls\": [10 ToolCallV2(11 id=\"search_docs_1byjy32y4hvq\",12 type=\"function\",13 function=ToolCallV2Function(14 name=\"search_docs\", arguments='{\"query\":\"tool use Cohere\",\"top_k\":3}'15 ),16 )17 ],18}1920{21 \"role\": \"tool\",22 \"tool_call_id\": \"search_docs_1byjy32y4hvq\",23 \"content\": [{\"type\": \"document\", \"document\": {\"data\": \"{\\\"title\\\":\\\"Tool use (function calling) overview\\\",\\\"url\\\":\\\"https://docs.cohere.com/v2/docs/tool-use-overview\\\",\\\"text\\\":\\\"Tool use connects models to external tools like search engines and APIs.\\\"}\"}}],24}2526{ 27 \"role\": \"assistant\", 28 \"content\": \"Tool use lets models call external tools (like doc search) and then answer using the tool results, with citations.\"29}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.408Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":17,"totalLines":88,"estimatedTokens":2788}}236{"id":"doc-versioning_cohere-83c0be47","source":"documentation","title":"Versioning | Cohere","url":"https://docs.cohere.com/versioning-reference","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.408Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}237{"id":"doc-an_overview_of_cohere_s_rerank_model_cohere-6a0b863f","source":"documentation","title":"An Overview of Cohere's Rerank Model | Cohere","url":"https://docs.cohere.com/docs/reranking","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.ClientV2()45query = \"What is the capital of the United States?\"6docs = [7 \"Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.\",8 \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.\",9 \"Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.\",10 \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.\",11 \"Capital punishment has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.\",12]1314results = co.rerank(15 model=\"rerank-v4.0-pro\", query=query, documents=docs, top_n=516)\n```\n\nExample:\n```text\n1V2RerankResponse(2 id=\"2104ccd0-74b5-4951-9bb1-cc543b26720f\",3 results=[4 V2RerankResponseResultsItem(5 index=3, relevance_score=0.9432646 ),7 V2RerankResponseResultsItem(8 index=2, relevance_score=0.622092079 ),10 V2RerankResponseResultsItem(11 index=1, relevance_score=0.605425812 ),13 V2RerankResponseResultsItem(14 index=0, relevance_score=0.5904013515 ),16 V2RerankResponseResultsItem(17 index=4, relevance_score=0.466456718 ),19 ],20 meta=ApiMeta(21 api_version=ApiMetaApiVersion(22 version=\"2\", is_deprecated=None, is_experimental=None23 ),24 billed_units=ApiMetaBilledUnits(25 images=None,26 input_tokens=None,27 output_tokens=None,28 search_units=1.0,29 classifications=None,30 ),31 tokens=None,32 cached_tokens=None,33 warnings=None,34 ),35)\n```\n\nExample:\n```text\n1import yaml2import cohere34co = cohere.ClientV2()56query = \"What is the capital of the United States?\"7docs = [8 {9 \"Title\": \"Facts about Carson City\",10 \"Content\": \"Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.\",11 },12 {13 \"Title\": \"The Commonwealth of Northern Mariana Islands\",14 \"Content\": \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.\",15 },16 {17 \"Title\": \"The Capital of United States Virgin Islands\",18 \"Content\": \"Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.\",19 },20 {21 \"Title\": \"Washington D.C.\",22 \"Content\": \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.\",23 },24 {25 \"Title\": \"Capital Punishment in the US\",26 \"Content\": \"Capital punishment has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.\",27 },28]2930yaml_docs = [yaml.dump(doc, sort_keys=False) for doc in docs]3132results = co.rerank(33 model=\"rerank-v4.0-pro\",34 query=query,35 documents=yaml_docs,36 top_n=5,37)\n```\n\nExample:\n```text\n1V2RerankResponse(2 id=\"df4d8720-8265-4868-a8f5-0bcee7a35bd0\",3 results=[4 V2RerankResponseResultsItem(5 index=3, relevance_score=0.94978136 ),7 V2RerankResponseResultsItem(8 index=2, relevance_score=0.690642549 ),10 V2RerankResponseResultsItem(11 index=0, relevance_score=0.5790195512 ),13 V2RerankResponseResultsItem(14 index=1, relevance_score=0.548286515 ),16 V2RerankResponseResultsItem(17 index=4, relevance_score=0.4937502718 ),19 ],20 meta=ApiMeta(21 api_version=ApiMetaApiVersion(22 version=\"2\", is_deprecated=None, is_experimental=None23 ),24 billed_units=ApiMetaBilledUnits(25 images=None,26 input_tokens=None,27 output_tokens=None,28 search_units=1.0,29 classifications=None,30 ),31 tokens=None,32 cached_tokens=None,33 warnings=None,34 ),35)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.409Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":4,"totalLines":23,"estimatedTokens":1344}}238{"id":"doc-cohere_release_notes_cdata_announcing_cohere_s_n-2bd4e014","source":"documentation","title":"Cohere Release Notes<![CDATA[Announcing Cohere's North Mini Code]]><![CDATA[Meet Cohere Transcribe Arabic]]><![CDATA[Announcing Cohere's Command A+]]><![CDATA[Announcing the Cohere Transcribe model]]><![CDATA[Retirement of Embed v2.0 and Aya Expanse / Vision 8B]]><![CDATA[Cohere's Rerank v4.0 Model is Here!]]><![CDATA[Announcing Major Command Deprecations]]><![CDATA[Announcing Cohere's Command A Translate Model]]><![CDATA[Announcing Cohere's Command A Reasoning Model]]><![CDATA[Announcing Cohere's Command A Vision Model]]><![CDATA[Announcing Cutting-Edge Cohere Models on OCI]]><![CDATA[Announcing Embed Multimodal v4]]><![CDATA[Announcing Command A]]><![CDATA[Our Groundbreaking Multimodal Model, Aya Vision, is Here!]]><![CDATA[Cohere Releases Arabic-Optimized Command Model!]]><![CDATA[Cohere via OpenAI SDK Using Compatibility API]]><![CDATA[Cohere's Rerank v3.5 Model is on Azure AI Foundry!]]><![CDATA[Cohere's Rerank v3.5 Model is on Azure AI Foundry!]]><![CDATA[Deprecation of Classify via default Embed Models]]><![CDATA[Cohere's Multimodal Embedding Models are on Bedrock!]]><![CDATA[Aya Expanse is Available on WhatsApp!]]><![CDATA[Announcing Command R7b]]><![CDATA[Announcing Rerank-v3.5]]><![CDATA[Structured Outputs support for tool use]]><![CDATA[Embed v3.0 Models are now Multimodal]]><![CDATA[Fine-Tuning Now Available for Command R 08-2024]]><![CDATA[New Embed, Rerank, Chat, and Classify APIs]]><![CDATA[Refreshed Command R and R+ models now on Azure]]><![CDATA[Command models get an August refresh]]><![CDATA[Force JSON object response format]]><![CDATA[Release Notes for June 10th 2024]]><![CDATA[Advanced Retrieval Launch release]]><![CDATA[Cohere Python SDK v5.2.0 release]]><![CDATA[Command R: Retrieval-Augmented Generation at Scale]]><![CDATA[Cohere Python SDK v5.0.0 release]]><![CDATA[Fine-tuning has been added to the Python SDK]]><![CDATA[Release Notes September 29th 2023]]><![CDATA[Release Notes January 22, 2024]]><![CDATA[Release Notes August 8th 2023 (Changelog)]]><![CDATA[Release Notes June 28th 2023 (Changelog)]]><![CDATA[New Maximum Number of Input Documents for Rerank]]><![CDATA[Cohere Model Names Are Changing!]]><![CDATA[Multilingual Support for Co.classify]]><![CDATA[Command Model Nightly Available!]]><![CDATA[Command R+ is a scalable LLM for business]]><![CDATA[Multilingual Text Model + Language Detection]]><![CDATA[Model Sizing Update + Improvements]]><![CDATA[Co.classify uses Representational model embeddings]]><![CDATA[New Look For Cohere Documentation!]]><![CDATA[New Logit Bias experimental parameter]]><![CDATA[Current Model Upgrades + New Command Beta Model]]><![CDATA[Pricing Update and New Dashboard UI]]><![CDATA[Updated Small, Medium, and Large Generation Models]]><![CDATA[The `model` Parameter Becomes Optional.]]><![CDATA[Introducing Moderate Tool (Beta)!]]><![CDATA[New & Improved Generation and Representation Models]]><![CDATA[Introducing Classification Endpoint]]><![CDATA[New and Improved Extremely Large Model!]]><![CDATA[Extremely Large (Beta) Release]]><![CDATA[New & Improved Generation Models]]><![CDATA[Finetuning Available + Policy Updates]]><![CDATA[Larger Cohere Representation Models]]>","url":"https://docs.cohere.com/changelog.rss","text":"Example:\n```python\nimport cohere\n\nco = cohere.ClientV2()\n\nresponse = co.audio.transcriptions.create(\n model=\"cohere-transcribe-03-2026\",\n language=\"en\",\n file=open(\"./sample.wav\", \"rb\"),\n)\n\nprint(response)\n```\n\nExample:\n```python\nimport cohere\n\nco = cohere.ClientV2()\n\nquery = \"What is the capital of the United States?\"\ndocs = [\n \"Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.\",\n \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.\",\n \"Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.\",\n \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.\",\n \"Capital punishment has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.\",\n]\n\nresults = co.rerank(\n model=\"rerank-v4.0-pro\", query=query, documents=docs, top_n=5\n)\n```\n\nExample:\n```python\nfrom cohere import ClientV2\n\nco = ClientV2(api_key=\"<YOUR API KEY>\")\n\nresponse = co.chat(\n model=\"command-a-translate-08-2025\",\n messages=[\n {\n \"role\": \"user\",\n \"content\": \"Translate this text to Spanish: Hello, how are you?\",\n }\n ],\n)\n```\n\nExample:\n```python\nimport cohere\n\nco = cohere.Client(\"your-api-key\")\n\nresponse = co.chat(\n model=\"command-a-vision-07-2025\",\n messages=[\n {\n \"role\": \"user\",\n \"content\": [\n {\n \"type\": \"text\",\n \"text\": \"Analyze this chart and extract the key data points\",\n },\n {\n \"type\": \"image_url\",\n \"image_url\": {\"url\": \"your-image-url\"},\n },\n ],\n }\n ],\n)\n```\n\nExample:\n```bash\npip install 'git+https://github.com/huggingface/transformers.git'\n```\n\nExample:\n```python\nfrom transformers import AutoTokenizer, AutoModelForCausalLM\n\nmodel_id = \"CohereForAI/c4ai-command-r7b-12-2024\"\ntokenizer = AutoTokenizer.from_pretrained(model_id)\nmodel = AutoModelForCausalLM.from_pretrained(model_id)\n\n# Format message with the c4ai-command-r7b-12-2024 chat template\nmessages = [{\"role\": \"user\", \"content\": \"مرحبا، كيف حالك؟\"}]\ninput_ids = tokenizer.apply_chat_template(\n messages,\n tokenize=True,\n add_generation_prompt=True,\n return_tensors=\"pt\",\n)\n\ngen_tokens = model.generate(\n input_ids,\n max_new_tokens=100,\n do_sample=True,\n temperature=0.3,\n)\n\ngen_text = tokenizer.decode(gen_tokens[0])\nprint(gen_text)\n```\n\nExample:\n```python\n# Define conversation input\nconversation = [\n {\n \"role\": \"user\",\n \"content\": \"اقترح طبقًا يمزج نكهات من عدة دول عربية\",\n }\n]\n\n# Define documents for retrieval-based generation\ndocuments = [\n {\n \"heading\": \"المطبخ العربي: أطباقنا التقليدية\",\n \"body\": \"يشتهر المطبخ العربي بأطباقه الغنية والنكهات الفريدة. في هذا المقال، سنستكشف ...\",\n },\n {\n \"heading\": \"وصفة اليوم: مقلوبة\",\n \"body\": \"المقلوبة هي طبق فلسطيني تقليدي، يُحضر من الأرز واللحم أو الدجاج والخضروات. في وصفتنا اليوم ...\",\n },\n]\n\n# Get the RAG prompt\ninput_prompt = tokenizer.apply_chat_template(\n conversation=conversation,\n documents=documents,\n tokenize=False,\n add_generation_prompt=True,\n return_tensors=\"pt\",\n)\n# Tokenize the prompt\ninput_ids = tokenizer.encode_plus(input_prompt, return_tensors=\"pt\")\n```\n\nExample:\n```text\nPOST https://api.cohere.ai/v2/rerank\n{\n \"model\": \"rerank-v3.5\",\n \"query\": \"What is the capital of the United States?\",\n \"top_n\": 3,\n \"documents\": [\"Carson City is the capital city of the American state of Nevada.\",\n \"The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean. Its capital is Saipan.\",\n \"Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district.\",\n \"Capitalization or capitalisation in English grammar is the use of a capital letter at the start of a word. English usage varies from capitalization in other languages.\",\n \"Capital punishment has existed in the United States since beforethe United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states.\"]\n}\n```\n\nExample:\n```text\nPOST https://api.cohere.ai/v1/embed\n{\n \"model\": \"embed-multilingual-v3.0\",\n \"input_type\": \"image\",\n \"embedding_types\": [\"float\"],\n \"images\": [enc_img]\n}\n```\n\nExample:\n```text\nPOST https://api.cohere.ai/v1/chat\n{\n \"message\": \"Generate a JSON that represents a person, with name and age\",\n \"model\": \"command-nightly\",\n \"response_format\": {\n \"type\": \"json_object\"\n }\n}\n```\n\nExample:\n```text\nPOST https://api.cohere.ai/v1/chat\n{\n \"message\": \"Generate a JSON that represents a person, with name and age\",\n \"model\": \"command-nightly\",\n \"response_format\": {\n \"type\": \"json_object\",\n \"schema\": {\n \"type\": \"object\",\n \"required\": [\"name\", \"age\"],\n \"properties\": {\n \"name\": { \"type\": \"string\" },\n \"age\": { \"type\": \"integer\" }\n }\n }\n }\n}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.412Z","totalSectionsIncluded":0,"totalCodeBlocksIncluded":11,"totalLines":204,"estimatedTokens":1460}}239{"id":"doc-generate_cohere-b01e9988","source":"documentation","title":"Generate | Cohere","url":"https://docs.cohere.com/reference/generate","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nThis API is marked as “Legacy” and is no longer maintained. Follow the migration guide to start using the Chat API. Generates realistic text conditioned on a given input.\n\nWhen true, the response will be a JSON stream of events. Streaming is beneficial for user interfaces that render the contents of the response piece by piece, as it gets generated. The final event will contain the complete response, and will contain an is_finished field set to true. The event will also contain a finish_reason, which can be one of the - the model sent back a finished reply MAX_TOKENS - the reply was cut off because the model reached the maximum number of tokens for its context length ERROR - something went wrong when generating the reply ERROR_TOXIC - the model generated a reply that was deemed toxic\n\nThe identifier of the model to generate with. Currently available models are command (default), command-nightly (experimental), command-light, and command-light-nightly (experimental). Smaller, “light” models are faster, while larger models will perform better. Custom models can also be supplied with their full ID.\n\nThe maximum number of tokens the model will generate as part of the response. a low value may result in incomplete generations. This parameter is off by default, and if it’s not specified, the model will continue generating until it emits an EOS completion token. See BPE Tokens for more details. Can only be set to 0 if return_likelihoods is set to ALL to get the likelihood of the prompt.\n\nOne of NONE|START|END to specify how the API will handle inputs longer than the maximum token length. Passing START will discard the start of the input. END will discard the end of the input. In both cases, input is discarded until the remaining input is exactly the maximum input token length for the model. If NONE is selected, when the input exceeds the maximum input token length an error will be returned.\n\nIdentifier of a custom preset. A preset is a combination of parameters, such as prompt, temperature etc. You can create presets in the playground. When a preset is specified, the prompt parameter becomes optional, and any included parameters will override the preset’s parameters.\n\nDefaults to 0.0, min value of 0.0, max value of 1.0. Can be used to reduce repetitiveness of generated tokens. Similar to frequency_penalty, except that this penalty is applied equally to all tokens that have already appeared, regardless of their exact frequencies. Using frequency_penalty in combination with presence_penalty is not supported on newer models.\n\nExample:\n```text\n1import cohere23co = cohere.Client()45response = co.generate(6 prompt=\"Please explain to me how LLMs work\",7)8print(response)\n```\n\nExample:\n```text\n1{2 \"id\": \"6afae9c2-3375-4d0e-8d18-2e9eb7f2c3ec\",3 \"generations\": [4 {5 \"id\": \"8e6de35d-3007-43ab-9253-ac4f95dcb8a2\",6 \"text\": \"LLMs, or Large Language Models, are a type of neural network-based AI model that has been trained on massive amounts of text data and have become ubiquitous in the AI landscape. They possess astounding capabilities for comprehending and generating human-like language.\\nThese models leverage neural networks that operate on a large scale, often involving millions or even billions of parameters. This substantial scale enables them to capture intricate patterns and connections within the vast amounts of text they have been trained on.\\n\\nThe training process for LLMs is fueled by colossal datasets of textual information, ranging from books and articles to websites and conversational transcripts. This extensive training enables them to develop a nuanced understanding of language patterns, grammar, and semantics.\\n\\nWhen posed with a new text input, LLMs employ their finely honed understanding of language to generate informed responses or undertake tasks such as language translation, text completion, or question answering. They do this by manipulating the input text through adding, removing, or altering elements to craft a desired output.\\n\\nOne of the underlying principles of their efficacy is the recurrent neural network (RNN) architecture they often adopt. This design enables them to process sequential data like natural language effectively. RNNs possess \\\"memory\\\" aspects via loops between layers, which allows them to retain and manipulate information gathered across long sequences, akin to the way humans process information.\\n\\nHowever, it's their size that arguably constitutes their most notable aspect. The sheer volume of these models – with counts of parameters often exceeding 100 million – enables them to capture correlations and patterns within language data effectively. This empowers them to generate coherent and contextually appropriate responses, posing a remarkable advancement in conversational AI.\\n\\nWhile LLMs have demonstrated extraordinary language prowess, it's vital to acknowledge their limitations and potential for improvement. Their biases often reflect those of the training data, and they may struggle with logical inconsistencies or factual errors. Ongoing research aims to enhance their robustness, diversity, and overall usability.\\n\\nIn essence, LLMs are a groundbreaking manifestation of AI's potential to simulate and even extend human language capabilities, while also serving as a testament to the ongoing journey towards refining and perfecting these technologies.\"7 }8 ],9 \"prompt\": \"Please explain to me how LLMs work\",10 \"meta\": {11 \"api_version\": {12 \"version\": \"1\"13 },14 \"billed_units\": {15 \"input_tokens\": 8,16 \"output_tokens\": 44217 }18 }19}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.412Z","totalSectionsIncluded":8,"totalCodeBlocksIncluded":2,"totalLines":27,"estimatedTokens":1452}}240{"id":"doc-how_to_manage_a_cohere_connector_cohere-2c9eddbb","source":"documentation","title":"How to Manage a Cohere Connector | Cohere","url":"https://docs.cohere.com/docs/managing-your-connector","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n$curl --request GET $ --url 'https://api.cohere.ai/v1/connectors' $ --header 'Authorization: Bearer {Cohere API key}'\n```\n\nExample:\n```text\n$curl --request POST $ --url 'https://api.cohere.ai/v1/connectors/{connector-id}/oauth/authorize' $ --header 'Authorization: Bearer {Cohere API key for user wishing to authorize}'\n```\n\nExample:\n```text\n$curl --request PATCH $ --url 'https://api.cohere.ai/v1/connectors/{id}' $ --header 'Authorization: Bearer {Cohere API key}' $ --header 'Content-Type: application/json' $ --data '{ > \"name\": \"new connector name\", > \"url\": \"https://new-connector-example.com/search\", > \"auth_type\": \"oauth\", > \"oauth\": { > \"authorize_url\": \"https://new.com/authorize\", > \"token_url\": \"https://new.com/access\", > \"scope\": \"new_scope\" > }, > \"active\": true, > }'\n```\n\nExample:\n```text\n1import cohere23co = cohere.Client(\"Your API key\")4response = co.chat(5 message=\"What is the chemical formula for glucose?\",6 stream=True,7 connectors=[8 {\"id\": \"example_connector_id\"}9 ], # this is from the create step10)\n```\n\nExample:\n```text\n1 \"search_results\": [ 2 { 3 \"connector\": { 4 \"id\": \"connector_id\" 5 }, 6 \"error_message\":\"connector error message\" 7 }\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.413Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":5,"totalLines":28,"estimatedTokens":386}}241{"id":"doc-an_overview_of_cohere_s_rag_connectors_cohere-0ae1db67","source":"documentation","title":"An Overview of Cohere's RAG Connectors | Cohere","url":"https://docs.cohere.com/docs/connectors","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1import cohere23co = cohere.Client(api_key=\"Your API key\")45response = co.chat(6 model=\"command-a-03-2025\",7 message=\"What is the chemical formula for glucose?\",8 connectors=[{\"id\": \"web-search\"}],9)\n```\n\nExample:\n```text\n1connectors = [{\"id\": \"web-search\"}, {\"id\": \"customer-connector-id\"}]\n```\n\nExample:\n```text\n1{2 \"text\": \"The chemical formula for glucose is C6H12O6.\",3 \"generation_id\": \"667f0844-e5c9-4108-8624-45b7687ca6f3\",4 \"citations\": [5 {6 \"start\": 36,7 \"end\": 44,8 \"text\": \"C6H12O6.\",9 \"document_ids\": [10 \"web-search_3:0\",11 \"web-search_3:4\",12 \"web-search_4:0\",13 \"web-search_4:1\"14 ]15 }16 ],17 \"documents\": [18 {19 \"id\": \"web-search_3:0\",20 \"snippet\": \"Chemical Compound Formulas\\n\\nGlucose is a simple sugar with six carbon atoms and one aldehyde group. This monosaccharide has a chemical formula C6H12O6.\\n\\nIt is also known as dextrose. It is referred to as aldohexose as it contains 6 carbon atoms and an aldehyde group. It exists in two forms, open-chain or ring structure. It is synthesized in the liver and kidneys of animals. In plants, it is found in fruits and in different parts of plants. D- glucose is the naturally occurring form of glucose. It can occur either in solid or liquid form. It is water-soluble and is also soluble in acetic acid.\",21 \"title\": \"Glucose C6H12O6 - Chemical Formula, Structure, Composition, Properties, uses and FAQs of Glucose.\",22 \"url\": \"https://byjus.com/chemistry/glucose/\"23 },24 {25 \"id\": \"web-search_3:4\",26 \"snippet\": \"\\n\\nFrequently Asked Questions- FAQs\\n\\nHow do you represent glucose?\\n\\nThe chemical formula of Glucose is C6H12O6. Glucose is a monosaccharide containing an aldehyde group (-CHO). It is made of 6 carbon atoms, 12 hydrogen atoms and 6 oxygen atoms. Glucose is an aldohexose.\\n\\nIs glucose a reducing sugar?\\n\\nGlucose is a reducing sugar because it belongs to the category of an aldose meaning its open-chain form contains an aldehyde group. Generally, an aldehyde is quite easily oxidized to carboxylic acids.\\n\\nWhat are the 5 reducing sugars?\\n\\nThe 5 reducing sugars are ribose, glucose, galactose, glyceraldehyde, xylose.\\n\\nWhat are the elements of glucose?\",27 \"title\": \"Glucose C6H12O6 - Chemical Formula, Structure, Composition, Properties, uses and FAQs of Glucose.\",28 \"url\": \"https://byjus.com/chemistry/glucose/\"29 },30 {31 \"id\": \"web-search_4:0\",32 \"snippet\": \"Science, Tech, Math › Science\\n\\nGlucose Molecular Formula and Facts\\n\\nChemical or Molecular Formula for Glucose\\n\\nScience Photo Library - MIRIAM MASLO. / Getty Images\\n\\nProjects & Experiments\\n\\nChemistry In Everyday Life\\n\\nAbbreviations & Acronyms\\n\\nAnne Marie Helmenstine, Ph.D.\\n\\nAnne Marie Helmenstine, Ph.D.\\n\\nPh.D., Biomedical Sciences, University of Tennessee at Knoxville\\n\\nB.A., Physics and Mathematics, Hastings College\\n\\nDr. Helmenstine holds a Ph.D. in biomedical sciences and is a science writer, educator, and consultant. She has taught science courses at the high school, college, and graduate levels.\\n\\nLearn about our Editorial Process\\n\\nUpdated on November 03, 2019\\n\\nThe molecular formula for glucose is C6H12O6 or H-(C=O)-(CHOH)5-H. Its empirical or simplest formula is CH2O, which indicates there are two hydrogen atoms for each carbon and oxygen atom in the molecule.\",33 \"title\": \"Know the Chemical or Molecular Formula for Glucose\",34 \"url\": \"https://www.thoughtco.com/glucose-molecular-formula-608477\"35 }36 ],37 \"search_results\": [38 {39 \"search_query\": {40 \"text\": \"chemical formula for glucose\",41 \"generation_id\": \"66e388c8-d9a8-4d43-a711-0f17c3f0f82a\"42 },43 \"document_ids\": [44 \"web-search_3:0\",45 \"web-search_3:4\",46 \"web-search_4:0\"47 ],48 \"connector\": {49 \"id\": \"web-search\"50 }51 }52 ],53 \"search_queries\": [54 {55 \"text\": \"chemical formula for glucose\",56 \"generation_id\": \"66e388c8-d9a8-4d43-a711-0f17c3f0f82a\"57 }58 ]59}\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.413Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":3,"totalLines":18,"estimatedTokens":1072}}242{"id":"doc-how_to_authenticate_a_connector_cohere-ab6d277e","source":"documentation","title":"How to Authenticate a Connector | Cohere","url":"https://docs.cohere.com/docs/connector-authentication","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\nAuthorization: Bearer <token>\n```\n\nExample:\n```text\n1# Generate a token2import secrets34secrets.token_urlsafe(32)\n```\n\nExample:\n```text\n$curl --request POST $ --url 'https://connector-example.com/search'$ --header 'Content-Type: application/json' $ --data '{ > \"query\": \"How do I expense a meal?\" > }'\n```\n\nExample:\n```text\n$curl --request POST $ --url 'https://connector-example.com/search'$ --header 'Content-Type: application/json' $ --header 'Authorization: Bearer {Connector API key}' $ --data '{ > \"query\": \"How do I expense a meal?\" > }'\n```\n\nExample:\n```text\n$curl --request POST $ --url 'https://api.cohere.ai/v1/connectors' $ --header 'Authorization: Bearer {Cohere API key}' $ --header 'Content-Type: application/json' $ --data '{ > \"name\":\"test-connector\", > \"description\":\"A test connector\", > \"url\":\"https://connector-example.com/search\", > \"service_auth\": { > \"type\": \"bearer\", > \"token\": \"{Connector API Key}\" > } > }'\n```\n\nExample:\n```text\n$curl --request PATCH $ --url 'https://api.cohere.ai/v1/connectors/{id}' $ --header 'Authorization: Bearer {Cohere API key}' $ --header 'Content-Type: application/json' $ --data '{ > \"service_auth\": { > \"type\": \"bearer\", > \"token\": \"{Connector API Key}\" > } > }'\n```\n\nExample:\n```text\n$curl --request POST $ --url https://connector-example.com/search$ --header 'Content-Type: application/json' $ --header 'Authorization: Bearer {Personal/Service API key}' $ --data '{ > \"query\": \"How do I expense a meal?\" > }'\n```\n\nExample:\n```text\nhttps://api.cohere.com/v1/connectors/oauth/token\n```\n\nExample:\n```text\nhttps://accounts.google.com/o/oauth2/authhttps://oauth2.googleapis.com/token\n```\n\nExample:\n```text\n$curl --request POST $ --url 'https://api.cohere.ai/v1/connectors' $ --header 'Authorization: Bearer {Cohere API key}' $ --header 'Content-Type: application/json' $ --data '{ > \"name\":\"test-connector\", > \"description\":\"A test connector\", > \"url\":\"https://connector-example.com/search\", > \"oauth\": { > \"client_id\": \"xxx-yyy.apps.googleusercontent.com\", > \"client_secret\": \"zzz-vvv\", > \"authorize_url\": \"https://accounts.google.com/o/oauth2/auth\", > \"token_url\": \"https://oauth2.googleapis.com/token\", > \"scope\": \"https://www.googleapis.com/auth/drive.readonly\" > } > }'\n```\n\nExample:\n```text\n$curl --request PATCH $ --url 'https://api.cohere.ai/v1/connectors/{id}' $ --header 'Authorization: Bearer {Cohere API key}' $ --header 'Content-Type: application/json' $ --data '{ > \"oauth\": { > \"client_id\": \"xxx-yyy.apps.googleusercontent.com\", > \"client_secret\": \"zzz-vvv\", > \"authorize_url\": \"https://accounts.google.com/o/oauth2/auth\", > \"token_url\": \"https://oauth2.googleapis.com/token\", > \"scope\": \"https://www.googleapis.com/auth/drive.readonly\" > } > }'\n```\n\nExample:\n```text\n1import cohere23co = cohere.Client(\"Your API key\")4response = co.chat(5 message=\"What is the chemical formula for glucose?\",6 connectors=[7 {8 \"id\": \"web-search\",9 \"user_access_token\": \"{Personal/Service API key}\",10 }11 ],12)\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.413Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":12,"totalLines":63,"estimatedTokens":877}}243{"id":"doc-usage_patterns_for_tool_use_function_calling_coh-33da29bb","source":"documentation","title":"Usage patterns for tool use (function calling) | Cohere","url":"https://docs.cohere.com/docs/multi-hop-tool-use","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\n1# ! pip install -U cohere2import cohere34co = cohere.ClientV2(5 \"COHERE_API_KEY\"6) # Get your free API key here: https://dashboard.cohere.com/api-keys\n```\n\nExample:\n```text\n1def search_docs(query, top_k=3):2 # Implement any retrieval logic here (vector DB, keyword search, etc.)3 return [4 {5 \"title\": \"Tool use (function calling) overview\",6 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",7 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",8 },9 {10 \"title\": \"Structured outputs\",11 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",12 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",13 },14 {15 \"title\": \"Chat API reference (v2)\",16 \"url\": \"https://docs.cohere.com/reference/chat\",17 \"text\": \"Use the Chat endpoint to generate responses and optionally call tools.\",18 },19 ][:top_k]20 # Return a string or a list of objects. In Step 3, we'll wrap each object into a `document` content block.212223functions_map = {\"search_docs\": search_docs}2425tools = [26 {27 \"type\": \"function\",28 \"function\": {29 \"name\": \"search_docs\",30 \"description\": \"Search documentation and return relevant snippets as documents.\",31 \"parameters\": {32 \"type\": \"object\",33 \"properties\": {34 \"query\": {35 \"type\": \"string\",36 \"description\": \"The search query to look up in the docs.\",37 },38 \"top_k\": {39 \"type\": \"integer\",40 \"description\": \"How many documents to return.\",41 },42 },43 \"required\": [\"query\"],44 },45 },46 },47]\n```\n\nExample:\n```text\n1messages = [2 {3 \"role\": \"user\",4 \"content\": \"Find docs about tool use and structured outputs.\",5 }6]78response = co.chat(9 model=\"command-a-plus-05-2026\", messages=messages, tools=tools10)1112if response.message.tool_calls:13 messages.append(response.message)14 print(response.message.tool_plan, \"\\n\")15 print(response.message.tool_calls)\n```\n\nExample:\n```text\n1I will search the docs for tool use and structured outputs.23[4 ToolCallV2(5 id=\"search_docs_9b0nr4kg58a8\",6 type=\"function\",7 function=ToolCallV2Function(8 name=\"search_docs\", arguments='{\"query\":\"tool use\",\"top_k\":3}'9 ),10 ),11 ToolCallV2(12 id=\"search_docs_0qq0mz9gwnqr\",13 type=\"function\",14 function=ToolCallV2Function(15 name=\"search_docs\", arguments='{\"query\":\"structured outputs\",\"top_k\":3}'16 ),17 ),18]\n```\n\nExample:\n```text\n1import json23if response.message.tool_calls:4 for tc in response.message.tool_calls:5 tool_result = functions_map[tc.function.name](6 **json.loads(tc.function.arguments)7 )8 tool_content = []9 for data in tool_result:10 # Optional: the \"document\" object can take an \"id\" field for use in citations, otherwise auto-generated11 tool_content.append(12 {13 \"type\": \"document\",14 \"document\": {\"data\": json.dumps(data)},15 }16 )17 messages.append(18 {19 \"role\": \"tool\",20 \"tool_call_id\": tc.id,21 \"content\": tool_content,22 }23 )\n```\n\nExample:\n```text\n1messages = [{\"role\": \"user\", \"content\": \"What's 2+2?\"}]23response = co.chat(4 model=\"command-a-plus-05-2026\", messages=messages, tools=tools5)67if response.message.tool_calls:8 print(response.message.tool_plan, \"\\n\")9 print(response.message.tool_calls)1011else:12 print(response.message.content[0].text)\n```\n\nExample:\n```text\n1The answer to 2+2 is 4.\n```\n\nExample:\n```text\n1def search_docs(query, top_k=3):2 # Implement any retrieval logic here (vector DB, keyword search, etc.)3 return [4 {5 \"title\": \"Tool use (function calling) overview\",6 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-overview\",7 \"text\": \"Tool use connects models to external tools like search engines and APIs.\",8 },9 {10 \"title\": \"Usage patterns for tool use\",11 \"url\": \"https://docs.cohere.com/v2/docs/tool-use-usage-patterns\",12 \"text\": \"Common patterns include parallel tool calling, multi-step tool use, and more.\",13 },14 {15 \"title\": \"Structured outputs\",16 \"url\": \"https://docs.cohere.com/docs/structured-outputs\",17 \"text\": \"Use JSON schema to define structured inputs/outputs for tools and responses.\",18 },19 ][:top_k]202122functions_map = {\"search_docs\": search_docs}\n```\n\nExample:\n```text\n1tools = [2 {3 \"type\": \"function\",4 \"function\": {5 \"name\": \"search_docs\",6 \"description\": \"Search documentation and return relevant snippets as documents.\",7 \"parameters\": {8 \"type\": \"object\",9 \"properties\": {10 \"query\": {11 \"type\": \"string\",12 \"description\": \"The search query to look up in the docs.\",13 },14 \"top_k\": {15 \"type\": \"integer\",16 \"description\": \"How many documents to return.\",17 },18 },19 \"required\": [\"query\"],20 },21 },22 },23]\n```\n\nExample:\n```text\n1import json23# Step 1: Get the user message4messages = [5 {6 \"role\": \"user\",7 \"content\": \"Explain how tool use works and how to force tool usage. Please cite your sources.\",8 }9]1011# Step 2: Generate tool calls (if any)12model = \"command-a-plus-05-2026\"13response = co.chat(14 model=model, messages=messages, tools=tools, temperature=0.315)1617while response.message.tool_calls:18 print(\"TOOL PLAN:\")19 print(response.message.tool_plan, \"\\n\")20 print(\"TOOL CALLS:\")21 for tc in response.message.tool_calls:22 print(23 f\"Tool name: {tc.function.name} | Parameters: {tc.function.arguments}\"24 )25 print(\"=\" * 50)2627 messages.append(response.message)2829 # Step 3: Get tool results30 print(\"TOOL RESULT:\")31 for tc in response.message.tool_calls:32 tool_result = functions_map[tc.function.name](33 **json.loads(tc.function.arguments)34 )35 tool_content = []36 print(tool_result)37 for data in tool_result:38 # Optional: the \"document\" object can take an \"id\" field for use in citations, otherwise auto-generated39 tool_content.append(40 {41 \"type\": \"document\",42 \"document\": {\"data\": json.dumps(data)},43 }44 )45 messages.append(46 {47 \"role\": \"tool\",48 \"tool_call_id\": tc.id,49 \"content\": tool_content,50 }51 )5253 # Step 4: Generate response and citations54 response = co.chat(55 model=model,56 messages=messages,57 tools=tools,58 temperature=0.1,59 )6061messages.append(62 {63 \"role\": \"assistant\",64 \"content\": response.message.content[0].text,65 }66)6768# Print final response69print(\"RESPONSE:\")70print(response.message.content[0].text)71print(\"=\" * 50)7273# Print citations (if any)74verbose_source = (75 True # Change to True to display the contents of a source76)77if response.message.citations:78 print(\"CITATIONS:\\n\")79 for citation in response.message.citations:80 print(81 f\"Start: {citation.start}| End:{citation.end}| Text:'{citation.text}' \"82 )83 print(\"Sources:\")84 for idx, source in enumerate(citation.sources):85 print(f\"{idx+1}. {source.id}\")86 if verbose_source:87 print(f\"{source.tool_output}\")88 print(\"\\n\")\n```\n\nExample:\n```text\n1TOOL PLAN:2First, I will search the docs for how tool use works. Then, I will search for how to force tool usage (tool_choice).34TOOL CALLS:5Tool name: search_docs | Parameters: {\"query\":\"tool use\",\"top_k\":3}6==================================================7TOOL RESULT:8[{'title': 'Tool use (function calling) overview', 'url': 'https://docs.cohere.com/v2/docs/tool-use-overview', 'text': 'Tool use connects models to external tools like search engines and APIs.'}]9TOOL PLAN:10Now I'll search for how to force tool usage via the tool_choice parameter.1112TOOL CALLS:13Tool name: search_docs | Parameters: {\"query\":\"tool_choice REQUIRED NONE\",\"top_k\":3}14==================================================15TOOL RESULT:16[{'title': 'Usage patterns for tool use', 'url': 'https://docs.cohere.com/v2/docs/tool-use-usage-patterns', 'text': 'Common patterns include parallel tool calling, multi-step tool use, and more.'}]17RESPONSE:18Tool use lets models call external tools (like doc search) and then answer using tool results with citations. You can force tool usage with tool_choice=\"REQUIRED\" or force a direct response with tool_choice=\"NONE\".19==================================================20CITATIONS:2122Start: 126| End:135| Text:'tool_choice'23Sources:241. search_docs_p0dage9q1nv4:025{'title': 'Usage patterns for tool use', 'url': 'https://docs.cohere.com/v2/docs/tool-use-usage-patterns', 'text': 'Common patterns include parallel tool calling, multi-step tool use, and more.'}\n```\n\nExample:\n```text\n1response = co.chat(2 model=\"command-a-plus-05-2026\",3 messages=messages,4 tools=tools,5 tool_choice=\"REQUIRED\" # optional, to force tool calls6 # tool_choice=\"NONE\" # optional, to force a direct response7)\n```\n\nExample:\n```text\n1from cohere import ToolCallV2, ToolCallV2Function23messages = [4 {5 \"role\": \"user\",6 \"content\": \"How does tool use work in Cohere? Please cite your sources.\",7 },8 {9 \"role\": \"assistant\",10 \"tool_plan\": \"I will search the docs for how tool use works in Cohere.\",11 \"tool_calls\": [12 ToolCallV2(13 id=\"search_docs_1byjy32y4hvq\",14 type=\"function\",15 function=ToolCallV2Function(16 name=\"search_docs\",17 arguments='{\"query\":\"tool use Cohere\",\"top_k\":3}',18 ),19 )20 ],21 },22 {23 \"role\": \"tool\",24 \"tool_call_id\": \"search_docs_1byjy32y4hvq\",25 \"content\": [26 {27 \"type\": \"document\",28 \"document\": {29 \"data\": '{\"title\":\"Tool use (function calling) overview\",\"url\":\"https://docs.cohere.com/v2/docs/tool-use-overview\",\"text\":\"Tool use connects models to external tools like search engines and APIs.\"}'30 },31 }32 ],33 },34 {35 \"role\": \"assistant\",36 \"content\": \"Tool use lets models call external tools (like doc search) and then answer using tool results with citations.\",37 },38]\n```\n\nExample:\n```text\n1messages.append(2 {\"role\": \"user\", \"content\": \"How do I force tool usage?\"}3)45response = co.chat(6 model=\"command-a-plus-05-2026\", messages=messages, tools=tools7)89if response.message.tool_calls:10 messages.append(response.message)11 print(response.message.tool_plan, \"\\n\")12 print(response.message.tool_calls)\n```\n\nExample:\n```text\n1I will search the docs for how to force tool usage using tool_choice.23[ToolCallV2(id='search_docs_8hwpm7d4wr14', type='function', function=ToolCallV2Function(name='search_docs', arguments='{\"query\":\"tool_choice REQUIRED NONE\",\"top_k\":3}'))]\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.414Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":15,"totalLines":78,"estimatedTokens":3038}}244{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-5d7b4c84","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/classify-starting-the-training","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.415Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}245{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-f9b0c8aa","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/chat-improving-the-results","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.415Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}246{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-6ce982a4","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/classify-understanding-the-results","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.416Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}247{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-71f8af27","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/chat-starting-the-training","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.416Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}248{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-98b2c7b9","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/rerank-understanding-the-results","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.416Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}249{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-fc656680","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/rerank-improving-the-results","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.417Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}250{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-3c9a70e3","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/rerank-starting-the-training","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.417Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}251{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-22006b7f","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/chat-understanding-the-results","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.417Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}252{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-ba985cf6","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/fine-tuning-with-the-python-sdk","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.417Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}253{"id":"doc-introduction_to_fine_tuning_with_cohere_models_c-8bfc94e3","source":"documentation","title":"Introduction to Fine-Tuning with Cohere Models | Cohere","url":"https://docs.cohere.com/docs/classify-improving-the-results","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.417Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":0,"totalLines":3,"estimatedTokens":46}}254{"id":"doc-installation_cohere-1c68660d","source":"documentation","title":"Installation | Cohere","url":"https://docs.cohere.com/command","text":"For AI documentation index is available at the root level at /llms.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.\n\nExample:\n```text\ncurl --proto '=https' --tlsv1.2 -sSf https://raw.githubusercontent.com/cohere-ai/co/main/install.sh | sh\n```\n\nExample:\n```text\nmkdir -p /usr/local/binmv ./co /usr/local/bin/\n```\n\nExample:\n```text\nco auth login --email=EMAIL\n```\n\n<|endofdoc|>","metadata":{"transformedAt":"2026-08-18T15:16:59.418Z","totalSectionsIncluded":1,"totalCodeBlocksIncluded":3,"totalLines":18,"estimatedTokens":107}}255