RAG System - Modal Deployment Ready

This is a modular RAG (Retrieval-Augmented Generation) system structured for Modal deployment.

Project Structure

Rag_Code/
├── main.py                    # Main Modal application (deploy this to Modal)
├── src/                       # Source package
│   ├── __init__.py
│   ├── config.py             # Configuration constants and models
│   ├── models.py             # Request/Response Pydantic models
│   ├── document_processing.py # Document extraction and processing
│   ├── rag_engine.py         # RAG processing logic
│   └── helpers/              # Helper utilities
│       ├── __init__.py
│       └── utils.py          # Utility classes (logging, timing, email, API client)
├── requirements.txt          # Python dependencies
├── example_usage.py          # Example usage script
└── README.md                # This file

Modal Deployment

Deploy to Modal:

# Install Modal CLI
pip install modal

# Deploy the application
modal deploy main.py

Local Development:

# Install dependencies
pip install -r requirements.txt

# Run example
python example_usage.py

API Usage

Once deployed to Modal, you'll get a URL like https://your-username--hackrx-rag-optimized-main.modal.run

Two-Mode Workflow

1. Upload Document (Mode 1)

First, upload and process your document:

import requests

# Upload document
upload_payload = {
    "document_url": "https://example.com/document.pdf",
    "config": {
        "chunking": {"chunk_size": 1024, "overlap": 250},
        "retrieval": {"semantic_search_top_k": 25, "bm25_search_top_k": 25},
        "generation": {"max_tokens": 400, "temperature": 0.1}
    },
    "email": {
        "enabled": True,
        "to_email": "your-email@example.com"  # Optional: send logs to this email
    }
}

upload_response = requests.post(
    "https://your-username--hackrx-rag-optimized-main.modal.run/hackrx/upload",
    json=upload_payload,
    headers={"Authorization": "Bearer YOUR_SECRET_TOKEN"}
)

session_data = upload_response.json()
session_id = session_data["session_id"]
print(f"Document processed! Session ID: {session_id}")
print(f"Chunks created: {session_data['chunks_count']}")

2. Chat with Document (Mode 2)

Then ask questions using the session ID:

# Ask questions
chat_payload = {
    "session_id": session_id,
    "questions": ["What is the main topic?", "What are the key findings?"],
    "config": {
        "chunking": {"chunk_size": 1024, "overlap": 250},
        "retrieval": {"semantic_search_top_k": 25, "bm25_search_top_k": 25},
        "generation": {"max_tokens": 400, "temperature": 0.1}
    },
    "email": {
        "enabled": True,
        "to_email": "your-email@example.com"  # Optional: send logs to this email
    }
}

chat_response = requests.post(
    "https://your-username--hackrx-rag-optimized-main.modal.run/hackrx/chat",
    json=chat_payload,
    headers={"Authorization": "Bearer YOUR_SECRET_TOKEN"}
)

answers = chat_response.json()["answers"]
print("Answers:", answers)

Legacy Single-Endpoint Mode

You can still use the original single endpoint for backward compatibility:

legacy_payload = {
    "documents": "https://example.com/document.pdf",
    "questions": ["What is the main topic?", "What are the key findings?"],
    "config": {
        "chunking": {"chunk_size": 1024, "overlap": 250},
        "retrieval": {"semantic_search_top_k": 25, "bm25_search_top_k": 25},
        "generation": {"max_tokens": 400, "temperature": 0.1}
    },
    "email": {
        "enabled": True,
        "to_email": "your-email@example.com"  # Optional: send logs to this email
    }
}

response = requests.post(
    "https://your-username--hackrx-rag-optimized-main.modal.run/hackrx/run",
    json=legacy_payload,
    headers={"Authorization": "Bearer YOUR_SECRET_TOKEN"}
)

answers = response.json()["answers"]

Key Features

• Multi-format Document Support: PDF, DOCX, PowerPoint, Excel, images, text, CSV
• Hybrid Search: Combines semantic search with BM25 keyword search
• Configurable Parameters: All components are configurable via Pydantic models
• Email Logging: Comprehensive logs sent via email
• Performance Optimization: Parallel processing and resource management
• Modal Deployment: Ready for cloud deployment with GPU support

Configuration

All configuration is managed through src/config.py:

• ChunkingConfig: Text chunking parameters
• RetrievalConfig: Search and retrieval parameters
• RRFConfig: Reciprocal Rank Fusion weights
• GenerationConfig: Answer generation parameters
• PerformanceConfig: Performance and concurrency settings

Endpoints

New Two-Mode Endpoints

• POST /hackrx/upload - Upload and process document (returns session ID)
• POST /hackrx/chat - Ask questions on uploaded document (uses session ID)

Legacy Endpoints

• POST /hackrx/run - Legacy single-endpoint RAG processing

Utility Endpoints

• GET /hackrx/config/default - Get default configuration
• GET /hackrx/health - Health check

Email Configuration

You can now specify email settings in your requests:

"email": {
    "enabled": True,  # Set to False to disable email logs
    "to_email": "your-email@example.com"  # Optional: specify recipient
}

• If enabled is True and to_email is provided, logs will be sent to that email
• If enabled is True but to_email is not provided, logs will be sent to the default email (if configured)
• If enabled is False or not provided, no emails will be sent