A FastAPI-powered intelligent document query system that processes PDFs, creates vector embeddings, and generates AI-powered answers using FAISS and sentence-transformers.

• PDF Processing: Download and extract text from PDF URLs using PyMuPDF
• Vector Embeddings: Create semantic embeddings using sentence-transformers (all-MiniLM-L6-v2)
• Vector Search: Efficient similarity search using FAISS
• AI-Powered Answers: Generate contextual answers using Google Gemini 2.5-flash (with mock fallback)
• Production Ready: Optimized for deployment on Render with proper error handling

• FastAPI: Modern Python web framework
• sentence-transformers: State-of-the-art text embeddings
• FAISS: Efficient vector similarity search
• PyMuPDF: PDF text extraction
• Google Gemini: AI-powered answer generation
• httpx: Async HTTP client for PDF downloads

Process a PDF document and answer questions using AI-powered semantic search.

Request Body:

{
  "documents": "https://example.com/document.pdf",
  "questions": [
    "What is the grace period for premium payment?",
    "Does this policy cover maternity?"
  ]
}

Response Body:

{
  "answers": [
    {
      "question": "What is the grace period for premium payment?",
      "answer": "A grace period of thirty days is provided for premium payment delays.",
      "source_clause": "Clause 3.2: A grace period of thirty days is provided...",
      "explanation": "Answer derived based on Clause 3.2 in the document."
    },
    {
      "question": "Does this policy cover maternity?",
      "answer": "Yes, maternity coverage is included under the medical benefits section.",
      "source_clause": "Section 5.1: Medical benefits include maternity coverage...",
      "explanation": "Information found in the medical benefits section of the policy."
    }
  ]
}

Health check endpoint that returns the status of all system components.

Response:

{
  "status": "healthy",
  "components": {
    "pdf_processor": true,
    "embedding_manager": true,
    "ai_generator": true
  }
}

1. Clone the repository

   git clone <repository-url>
   cd intelligent-document-query

2. Install dependencies

   pip install -r requirements.txt

3. Set up environment variables

   cp .env.example .env
   # Edit .env and add your Gemini API key

4. Run the application

   uvicorn main:app --host 0.0.0.0 --port 8000 --reload

5. Access the API

   • API Documentation: http://localhost:8000/docs
   • Health Check: http://localhost:8000/health
   • Main Endpoint: POST http://localhost:8000/api/v1/hackrx/run

• GEMINI_API_KEY: Your Google Gemini API key (required for AI-powered answers)
• DEBUG: Set to true for development mode (default: false)
• CORS_ORIGINS: Allowed CORS origins (default: *)

1. Go to Google AI Studio
2. Sign in with your Google account
3. Click "Get API key"
4. Create a new project or select existing one
5. Generate and copy your API key
6. Add it to your .env file as GEMINI_API_KEY=your_key_here

This application is ready for deployment on Render with the included render.yaml configuration.

1. Fork/clone this repository
2. Connect to Render
   • Go to Render Dashboard
   • Click "New Web Service"
   • Connect your GitHub repository
3. Configure Environment Variables
   • Add GEMINI_API_KEY in the Render dashboard
   • Other variables are configured in render.yaml
4. Deploy
   • Render will automatically build and deploy using the configuration

The system uses a modular architecture:

• FastAPI: Async web framework with automatic API documentation
• PDF Processor: Downloads and extracts text from PDF URLs using PyMuPDF
• Embedding Manager: Creates vector embeddings using sentence-transformers or TF-IDF fallback
• FAISS: Efficient vector similarity search for finding relevant document chunks
• AI Generator: Google Gemini 2.5-flash for intelligent answer generation with mock fallback

• Robust PDF Processing: Handles various PDF formats and sizes
• Smart Text Chunking: Intelligent text segmentation with overlap for context preservation
• Fallback Systems: TF-IDF vectorizer when advanced embeddings aren't available
• Mock Responses: Template answers when AI API is unavailable
• Error Handling: Comprehensive error handling with meaningful messages
• Production Ready: Optimized for deployment with health checks and monitoring

curl -X POST "http://localhost:8000/api/v1/hackrx/run" \
  -H "Content-Type: application/json" \
  -d '{
    "documents": "https://example.com/policy.pdf",
    "questions": [
      "What is the coverage limit?",
      "Are pre-existing conditions covered?"
    ]
  }'

• Python 3.11+
• FastAPI
• Google Gemini API key (optional, uses mock responses without it)
• Internet access for PDF downloads

This project is open source and available under the MIT License.