RAG API with Dynamic Document Processing

🚨 Important Notice 🚨

🔔 New Feature Added! This project now includes a fully functional User Interface (UI) that complements the API. The UI allows you to interact with the API directly without needing to write any code.

🚀 Explore the UI Now: UI Link 🚀

🔗 Quick Access: Installation | Usage | API Endpoints

Please make sure to review the updated documentation to take full advantage of the UI.

Project Description

This project implements a Retrieval-Augmented Generation (RAG) API using LlamaIndex and FastAPI. It's designed for efficient text-based question answering, integrating LlamaIndex for indexing and LlamaParse for flexible text preprocessing. The service is containerized using Docker for scalability, and a web UI is being developed with Streamlit for seamless user interaction.

Features

• RAG API using FastAPI for efficient question answering
• Document processing with LlamaParse
• Indexing and retrieval using LlamaIndex
• Containerized service with Docker
• Web UI (in development) using Streamlit

Technologies Used

• Python
• FastAPI
• LlamaIndex
• LlamaParse
• Docker
• Streamlit (for UI)
• Groq (LLM)
• Cohere (Embeddings)

Installation

1. Clone the repository:

git https://github.com/artmiss-gns/RAG_API 
cd RAG_API

2. Set up environment variables: Create a .env file in the root directory and add the following:

GROQ_API_KEY=your_groq_api_key
COHERE_API_KEY=your_cohere_api_key
LLAMA_CLOUD_API_KEY=your_llama_cloud_api_key
PORT=your_desired_port

Usage

Running Locally

• Set the Python path:

export PYTHONPATH="/$(pwd):$PYTHONPATH"

• Install dependencies:

pip install -r requirements.txt

• Start the FastAPI server:

  python src/main.py

  or

  uvicorn src.main:app --reload --port=8003

note: you can set the port to any port you want for this method
The API will be available at http://localhost:8003

Using Docker

1. Build the Docker image:

docker-compose up --build

After that, you can call the api with the following command:

http -f POST \
    http://localhost:8003\
    context@YOUR_FILE_TO_UPLOAD.pdf\
    query="YOUR QUERY"

API Endpoints

• GET /: Welcome message
• POST /: Main RAG endpoint
• Parameters:
  • context: File upload (document for context)
  • query: String (question to ask)
  • rebuild_index: Boolean (optional)
  • save_index: Boolean (optional)

Deployment

The API is deployed at: https://growing-bessy-hossein-golmohammadi-03788de4.koyeb.app/

License

MIT License