eveyuyi/ChatGPT_Prompts

1import os2import logging3 4from llama_index import GPTSimpleVectorIndex5from llama_index import download_loader6from llama_index import (7    Document,8    LLMPredictor,9    PromptHelper,10    QuestionAnswerPrompt,11    RefinePrompt,12)13from langchain.llms import OpenAI14import colorama15 16from modules.presets import *17from modules.utils import *18 19def get_index_name(file_src):20    index_name = []21    for file in file_src:22        index_name.append(os.path.basename(file.name))23    index_name = sorted(index_name)24    index_name = "".join(index_name)25    index_name = sha1sum(index_name)26    return index_name27 28def get_documents(file_src):29    documents = []30    logging.debug("Loading documents...")31    logging.debug(f"file_src: {file_src}")32    for file in file_src:33        logging.info(f"loading file: {file.name}")34        if os.path.splitext(file.name)[1] == ".pdf":35            logging.debug("Loading PDF...")36            CJKPDFReader = download_loader("CJKPDFReader")37            loader = CJKPDFReader()38            text_raw = loader.load_data(file=file.name)[0].text39        elif os.path.splitext(file.name)[1] == ".docx":40            logging.debug("Loading DOCX...")41            DocxReader = download_loader("DocxReader")42            loader = DocxReader()43            text_raw = loader.load_data(file=file.name)[0].text44        elif os.path.splitext(file.name)[1] == ".epub":45            logging.debug("Loading EPUB...")46            EpubReader = download_loader("EpubReader")47            loader = EpubReader()48            text_raw = loader.load_data(file=file.name)[0].text49        else:50            logging.debug("Loading text file...")51            with open(file.name, "r", encoding="utf-8") as f:52                text_raw = f.read()53        text = add_space(text_raw)54        documents += [Document(text)]55    return documents56 57 58def construct_index(59        api_key,60        file_src,61        max_input_size=4096,62        num_outputs=1,63        max_chunk_overlap=20,64        chunk_size_limit=600,65        embedding_limit=None,66        separator=" ",67        num_children=10,68        max_keywords_per_chunk=10,69):70    os.environ["OPENAI_API_KEY"] = api_key71    chunk_size_limit = None if chunk_size_limit == 0 else chunk_size_limit72    embedding_limit = None if embedding_limit == 0 else embedding_limit73    separator = " " if separator == "" else separator74 75    llm_predictor = LLMPredictor(76        llm=OpenAI(model_name="gpt-3.5-turbo-0301", openai_api_key=api_key)77    )78    prompt_helper = PromptHelper(79        max_input_size,80        num_outputs,81        max_chunk_overlap,82        embedding_limit,83        chunk_size_limit,84        separator=separator,85    )86    index_name = get_index_name(file_src)87    if os.path.exists(f"./index/{index_name}.json"):88        logging.info("找到了缓存的索引文件，加载中……")89        return GPTSimpleVectorIndex.load_from_disk(f"./index/{index_name}.json")90    else:91        try:92            documents = get_documents(file_src)93            logging.debug("构建索引中……")94            index = GPTSimpleVectorIndex(95                documents, llm_predictor=llm_predictor, prompt_helper=prompt_helper96            )97            os.makedirs("./index", exist_ok=True)98            index.save_to_disk(f"./index/{index_name}.json")99            return index100        except Exception as e:101            print(e)102            return None103 104 105def chat_ai(106        api_key,107        index,108        question,109        context,110        chatbot,111        reply_language,112):113    os.environ["OPENAI_API_KEY"] = api_key114 115    logging.info(f"Question: {question}")116 117    response, chatbot_display, status_text = ask_ai(118        api_key,119        index,120        question,121        replace_today(PROMPT_TEMPLATE),122        REFINE_TEMPLATE,123        SIM_K,124        INDEX_QUERY_TEMPRATURE,125        context,126        reply_language,127    )128    if response is None:129        status_text = "查询失败，请换个问法试试"130        return context, chatbot131    response = response132 133    context.append({"role": "user", "content": question})134    context.append({"role": "assistant", "content": response})135    chatbot.append((question, chatbot_display))136 137    os.environ["OPENAI_API_KEY"] = ""138    return context, chatbot, status_text139 140 141def ask_ai(142        api_key,143        index,144        question,145        prompt_tmpl,146        refine_tmpl,147        sim_k=1,148        temprature=0,149        prefix_messages=[],150        reply_language="中文",151):152    os.environ["OPENAI_API_KEY"] = api_key153 154    logging.debug("Index file found")155    logging.debug("Querying index...")156    llm_predictor = LLMPredictor(157        llm=OpenAI(158            temperature=temprature,159            model_name="gpt-3.5-turbo-0301",160            prefix_messages=prefix_messages,161        )162    )163 164    response = None  # Initialize response variable to avoid UnboundLocalError165    qa_prompt = QuestionAnswerPrompt(prompt_tmpl.replace("{reply_language}", reply_language))166    rf_prompt = RefinePrompt(refine_tmpl.replace("{reply_language}", reply_language))167    response = index.query(168        question,169        llm_predictor=llm_predictor,170        similarity_top_k=sim_k,171        text_qa_template=qa_prompt,172        refine_template=rf_prompt,173        response_mode="compact",174    )175 176    if response is not None:177        logging.info(f"Response: {response}")178        ret_text = response.response179        nodes = []180        for index, node in enumerate(response.source_nodes):181            brief = node.source_text[:25].replace("\n", "")182            nodes.append(183                f"<details><summary>[{index + 1}]\t{brief}...</summary><p>{node.source_text}</p></details>"184            )185        new_response = ret_text + "\n----------\n" + "\n\n".join(nodes)186        logging.info(187            f"Response: {colorama.Fore.BLUE}{ret_text}{colorama.Style.RESET_ALL}"188        )189        os.environ["OPENAI_API_KEY"] = ""190        return ret_text, new_response, f"查询消耗了{llm_predictor.last_token_usage} tokens"191    else:192        logging.warning("No response found, returning None")193        os.environ["OPENAI_API_KEY"] = ""194        return None195 196 197def add_space(text):198    punctuations = {"，": "， ", "。": "。 ", "？": "？ ", "！": "！ ", "：": "： ", "；": "； "}199    for cn_punc, en_punc in punctuations.items():200        text = text.replace(cn_punc, en_punc)201    return text202