🚀 HackRX RAG API

Enhanced Document Question Answering System with Adaptive Retrieval-Augmented Generation

Built for Bajaj Finserv HackRx Hackathon, this project is a FastAPI-based backend designed for advanced document Q&A. It combines Retrieval-Augmented Generation (RAG), Google Gemini LLM, and FAISS for fast, accurate, and scalable question answering across multiple document formats.

---

✨ Key Features

• 🔹 Adaptive RAG – Dynamically tunes retrieval & generation parameters for optimal answers.
• 🔹 Multi-format Support – Works with PDFs, DOCX, PPTX, images, and more.
• 🔹 Gemini LLM Integration – Uses Google Gemini for high-quality, context-aware responses.
• 🔹 FAISS Vector Store – Efficient, scalable document retrieval.
• 🔹 Batch Q&A – Ask multiple questions per document in a single request.
• 🔹 Caching Layer – Avoids redundant computations and speeds up repeated queries.
• 🔹 Secure Authentication – Endpoints protected with Bearer token.
• 🔹 Dockerized Deployment – Ready for production & hackathon demos.

---

📂 Project Structure

HackRX-RAG-API/
├── App/
│   ├── main.py                # FastAPI entrypoint
│   ├── config.py              # Config & env management
│   ├── RAG/
│   │   └── rag_llm.py         # Core RAG + Gemini logic
│   ├── routers/
│   │   └── merge.py           # API routes
│   └── Utils/                 # Helper utilities
│       ├── cache_utils.py
│       ├── downloader.py
│       ├── metadata_extractor.py
│       ├── mission_handler.py
│       └── parameter_selector.py
├── cache/                     # FAISS index + metadata cache
├── scripts/
│   └── warm_cache.py          # Pre-load docs into cache
├── requirements_f.txt         # Dependencies
├── Dockerfile                 # Container setup
├── docker-compose.yml         # Multi-container orchestration
└── logs/                      # Runtime logs

---

⚡ Installation & Setup

🔧 1. Using Docker (Recommended)

Prerequisites: Docker

# Build the image
docker build -t hackrx-rag-api .

# Run the container
docker run -p 8000:8000 --env-file App/.env hackrx-rag-api

Or with Docker Compose:

docker-compose up --build

---

🔧 2. Local Development

Prerequisites: Python 3.11+

# Clone repo and enter project
git clone https://github.com/your-username/hackrx-rag-api.git
cd hackrx-rag-api

# Create and activate virtual environment
python -m venv venv
source venv/bin/activate   # On Windows: venv\Scripts\activate

# Install dependencies
pip install --upgrade pip
pip install -r requirements_f.txt

# Setup environment
export GEMINI_API_KEY=your_gemini_api_key
export BEARER_TOKEN=your_bearer_token

# Run API
uvicorn App.main:app --reload --host 0.0.0.0 --port 8000

---

🚀 Live Demo

Deployed here:
👉 HackRX RAG API on Render

• API Docs: https://hackrx6-repo-bajaj-finserv.onrender.com/docs
• Main Endpoint:

  POST https://hackrx6-repo-bajaj-finserv.onrender.com/api/v1/hackrx/run
  

---

🔑 Configuration

Set environment variables in App/.env:

• GEMINI_API_KEY → Google Gemini API key
• BEARER_TOKEN → Bearer token for authentication
• UPLOAD_DIR → Directory for uploaded documents (default: /app/uploaded_documents)

---

📌 Usage

• API Docs: http://localhost:8000/docs (Swagger UI)
• Main Endpoint:

POST /api/v1/hackrx/run

Request Example:

{
  "document_url": "https://example.com/sample.pdf",
  "questions": [
    "What is the summary of this document?",
    "List the key financial metrics."
  ]
}

Response Example:

{
  "answers": [
    "This document discusses ...",
    "The financial metrics are ..."
  ]
}

---

🏆 Hackathon Context

This project was built for the Bajaj Finserv HackRx Hackathon to demonstrate how AI-powered document intelligence can:

• 🏆 Achieve 67th rank out of 4800 teams in the hackathon
• 📑 Extract insights from complex documents
• ⚡ Accelerate business decision-making
• 🔒 Ensure security with caching & authentication
• 🚀 Scale with Docker for enterprise-ready deployment

---

🙌 Acknowledgements

• FastAPI
• LangChain
• Google Gemini
• FAISS

---

📜 License

This project is for hackathon and educational purposes. See LICENSE for details.