notionhive-ai/nh-chatbot

🧠 Notionhive AI Chatbot

A FastAPI-based intelligent FAQ chatbot powered by Google Gemini and LangChain, designed to answer questions based on Notionhive's official FAQs and website: notionhive.com.

🚀 Features

• —Semantic search using Sentence Transformers & Chroma DB
• —Gemini AI (gemini-2.0-flash) for accurate response generation
• —CSV-based FAQ management (upload, add, delete, view)
• —RESTful API built with FastAPI
• —Prompt structure enforcing Notionhive's tone and boundaries
• —CORS enabled for frontend integration

📁 Project Structure

.
├── main.py                  # FastAPI app initialization
├── faq_routes.py           # All chatbot + FAQ management API endpoints
├── faq_services.py         # Vector store logic, Gemini model setup
├── chatbot_prompt.py       # Prompt template for Gemini responses
├── faqs.csv                # Stored FAQs (question/answer pairs)
└── README.md

🍞 Requirements

• —Python 3.9+
• —Google API Key for Gemini
• —sentence-transformers
• —langchain
• —chromadb
• —fastapi, uvicorn
• —python-dotenv, pandas

Install dependencies:

pip install -r requirements.txt

🔐 Environment Setup

Create a .env file in your root directory with the following:

GOOGLE_API_KEY=your_google_api_key_here

💠 Usage

Run the API server locally:

uvicorn main:app --reload

Access endpoints:

• —POST /ask — Ask a question
• —POST /add_faq — Add a new FAQ
• —POST /upload_faqs_csv — Upload bulk FAQs from CSV
• —DELETE /delete_faq — Delete FAQ by content
• —DELETE /deleted/{faq_id} — Delete FAQ by ID
• —DELETE /delete/destroyall — Delete all FAQs
• —GET /get_faqs — View all FAQs
• —POST /retrain — Reload vector DB

🤖 Chatbot Behavior

• —Introduces itself as Noah, Notionhive’s AI Assistant.
• —Answers based on FAQs and content from notionhive.com.
• —Uses web search only if FAQ-based answers aren’t available and question is general or critical.
• —Will never generate fabricated or speculative responses.

📌 Notes

• —All FAQs are stored in faqs.csv. Updating this file retrains the vector search index.
• —Currently single-tenant; for multi-tenant expansion, isolate CSV and vector store per tenant ID.

👨‍💼 Credits

Developed by the Notionhive AI team.