ChatDocument

This is a simple Retrieval Augmented Generation app (RAG) with a full-fledged user interface!

📄 Features

• Document Upload: Users can upload PDF files up to 3145728B (~3MB) containing information for analysis.
• Google Gemini Pro: Leveraging Google's latest LLM with 2 million token contexts to craft responses and embeddings.
• State-of-the-art UI: Can't go wrong with React + Bootstrap
• HTTPS: Deployed on a VPS with a valid SSL certificate

🔧 Tech stack

• Server: Langchain, FastAPI, Redis
• Client: React, Bootstrap

📦 Installation

1. Clone the repository
2. Make sure you are on the branch main
3. cd into it
4. Add GOOGLE_API_KEY and REDIS_URL as variables in the back_end/.env

• Google Gemini: https://ai.google.dev/gemini-api
• Redis: https://redis.io/

5. docker compose up

🔄 System structure (feedback appreciated!)

📋 TODOs:

• Create the app itself :D
  • PDF reader
  • RAG chain with context and chat history
  • REST api endpoints
  • React app client
  • Better PDF parser
  • Scale vectorstore
  • Agents with self-evaluating mechanism instead of current chain
  • Improve processing time + Add support for larger file
• Deploy on Vercel File too large
• Deploy on a VPS with nginx and pm2
• Use WebSocket for faster conversation response rate
• Site reliability engineering (eta soon)

👥 Contributors

• Phong Pham
• Trung Nguyen