Document QA System with RAG

Overview

An intelligent document-based question answering API built with FastAPI, Google Gemini 2.0, and MongoDB. The system uses Retrieval-Augmented Generation (RAG) to provide AI-powered answers from uploaded documents with real-time WebSocket streaming.

Key Capabilities

• Multi-User Support: User authentication with MongoDB and JWT-based sessions
• Document Processing: Upload and process PDF, TXT, Markdown, and JSON documents
• Semantic Search: FAISS vector similarity search with per-user data isolation
• AI-Powered Q&A: Google Gemini for answer generation with source attribution
• Real-time Streaming: WebSocket support for streaming responses
• Dual Intelligence Modes: RAG mode (document-based) and general chat mode

Core Functionality

• User Management: Secure registration, login, and JWT authentication with MongoDB storage
• Document Upload: Support for PDF, TXT, Markdown, and JSON with automatic text extraction and chunking
• Vector Search: FAISS for efficient similarity search using Gemini text embedding model
• Question Answering: Context-aware responses using Google Gemini with source citations
• Per-User Isolation: Complete data separation with user-scoped vector stores and documents
• WebSocket Streaming: Real-time bidirectional communication with authentication
• Document Management: Full CRUD operations for document lifecycle

Setup and Installation

Prerequisites

• Python 3.11 or 3.12
• MongoDB Atlas account
• Google Gemini API key (Get one here)

Installation Steps

1. Clone and Setup Environment

# Clone repository
git clone <your-repo-url>
cd document-qa-chatbot

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

# Install dependencies
pip install -r requirements.txt

2. Configure MongoDB Atlas

1. Create a free account at MongoDB Atlas
2. Create a new cluster
3. Configure network access (add your IP address or allow from anywhere: 0.0.0.0/0)
4. Create a database user with credentials
5. Get your connection string (Format: mongodb+srv://<username>:<password>@cluster.mongodb.net)

Configure Environment Variables

Create a .env file in the project root:

# Application Settings
APP_NAME=Document QA API
APP_VERSION=3.0.0
API_PREFIX=/api/v1
DEBUG=true

# Gemini API Configuration (Required)
GEMINI_API_KEY=your_gemini_api_key_here

# MongoDB Atlas Configuration (Required)
MONGODB_URL=mongodb+srv://username:password@cluster.mongodb.net
MONGODB_DB_NAME=document_qa_db

# Authentication Configuration (Required)
# Generate SECRET_KEY with: openssl rand -hex 32
SECRET_KEY=your_secure_random_secret_key_here
ALGORITHM=HS256
ACCESS_TOKEN_EXPIRE_MINUTES=1440

# Document Processing Configuration
CHUNK_SIZE=500
CHUNK_OVERLAP=50
MAX_FILE_SIZE=10485760

# Storage Paths
VECTOR_STORE_PATH=./data/vector_store
DOCUMENTS_DATA_PATH=./data/documents
UPLOADS_PATH=./data/uploads

# CORS Settings
ALLOWED_ORIGINS=*

Important: Generate a secure SECRET_KEY using:

openssl rand -hex 32

3. Run the Application

# Development mode
python main.py

# Or using uvicorn directly
uvicorn src.main:app --reload --host 0.0.0.0 --port 8000

Access the application:

• Web UI: http://localhost:8000
• API Documentation: http://localhost:8000/docs
• Health Check: http://localhost:8000/health

Project Structure

document-qa-chatbot/
├── main.py                          # Application entry point
├── requirements.txt                 # Python dependencies
├── .env                            # Environment configuration
│
├── src/                            # Application source code
│   ├── main.py                     # FastAPI app instance
│   ├── core/                       # Core functionality
│   │   ├── config.py               # Configuration management
│   │   └── security.py             # Authentication & JWT
│   ├── models/                     # Data models
│   │   └── schemas.py              # Pydantic schemas
│   ├── services/                   # Business logic
│   │   ├── database.py             # MongoDB service
│   │   ├── vector_store.py         # FAISS vector operations
│   │   ├── llm_service.py          # Gemini LLM integration
│   │   └── file_processor.py       # File processing
│   └── api/routes/                 # API endpoints
│       ├── auth.py                 # Authentication
│       ├── documents.py            # Document management
│       ├── query.py                # Query endpoint
│       ├── websocket.py            # WebSocket streaming
│       └── health.py               # Health check
│
├── static/                         # Frontend assets
│   └── index.html                 # Web interface
│
└── sample_documents/               # Test data
    ├── ai_basics.txt
    ├── machine_learning.md
    ├── python_guide.txt
    └── test_sample.pdf