Smart RAG - Intelligent Medical Literature Retrieval and Q&A System

中文版 | English Version

🚀 An intelligent medical literature Q&A system based on advanced RAG technology

An intelligent document Q&A platform that integrates hybrid retrieval, RRF result fusion, and multi-model support, specifically designed for medical literature processing and knowledge retrieval.

✨ Core Features

• 🚀 Advanced RAG Architecture: Adopts industry-leading Retrieval-Augmented Generation (RAG) technology to ensure the accuracy, relevance, and traceability of answers.
• 🧠 Intelligent Document Processing:
  • Multi-modal Parsing: Not only extracts text but also understands complex structures like tables and images in PDFs.
  • Dual-mode Smart Chunking: Prioritizes semantic chunking using HybridTextSplitter and maintains stable "recursive" chunking as a fallback to ensure context integrity.
• 🎯 Small-to-Big Retrieval & 4-Path Hybrid Retrieval:
  • Small-to-Big Architecture: Revolutionary retrieval strategy that uses small chunks for precise matching and parent chunks for complete context, optimizing both cost and quality.
  • Multi-dimensional Recall: Simultaneously performs four retrieval paths—VECTOR (semantic), CONTENT (full-text keyword), SUMMARY (summary keyword), and KEYWORDS (keyword list)—to maximize recall and precision.
  • RRF Fusion: Uses the Reciprocal Rank Fusion (RRF) algorithm to intelligently merge results from multiple paths.
  • Smart Switching: Automatically switches from small chunks (for retrieval) to parent chunks (for generation) to ensure optimal context.
• 🔍 AI Enhancement & Optimization:
  • Query Transformation: Utilizes LLMs to rewrite and expand user queries for better matching with the knowledge base.
  • AI Reranking: After retrieval, a more powerful AI model performs a second-pass "close reading" and reranking to ensure the final answer is based on the most relevant, high-quality content.
• 💬 Enterprise-grade Q&A Experience:
  • Streaming Output: Real-time streaming of answers to enhance user interaction.
  • Precise Source Tracing: All answers provide clear literature sources for easy verification.
  • High Scalability: Asynchronous and modular design makes it easy to integrate new models and features.

🚀 Quick Start

1. Prerequisites

• Python 3.9+
• Node.js 16+
• 16GB+ RAM recommended
• 100GB+ available storage space

2. Install Dependencies

# Install base dependencies
pip install -r requirements.txt

3. Configure Environment Variables

Copy and edit the environment variables file:

cp .env.example .env
# Edit the .env file to configure necessary parameters

4. Start the Backend Service

python run.py
python start_celery_worker.py  // For the queue

The backend API will start at http://localhost:8001

5. Start the Frontend Service

cd frontend
npm install
npm run dev

The frontend interface will start at http://localhost:3001

📚 Detailed Documentation

In-depth RAG Architecture Analysis

To help you understand the internal workings of our system, we provide a series of detailed component analysis documents:

• RAG System Architecture Overview - Recommended to read first for a high-level overview of the system.
  • Component Deep Dive 1: Intelligent Document Processing
  • Component Deep Dive 2: Query Transformation and Optimization
  • Component Deep Dive 3: 4-Path Recall and Fusion Retrieval
  • Component Deep Dive 4: AI-Enhanced Reranking
  • Component Deep Dive 5: Context Construction and Answer Generation

Other Documents

• Usage Guide - Detailed instructions on how to use the web interface and API.
• Installation and Configuration Guide - Detailed environment setup and installation steps.
• API Reference - Complete API documentation and usage examples.
• System Architecture - The original technical architecture, component design, and development guide.

---

🏗️ Project Structure

smart-rag/
├── app/                    # Application backend code (FastAPI)
│   ├── api/                # API routes
│   ├── core/               # Core functionalities (Config, Session Mgmt, etc.)
│   ├── embeddings/         # Embeddings and text chunking module
│   ├── metadata/           # Metadata generation module (summaries, keywords)
│   ├── models/             # Pydantic data models
│   ├── processors/         # Document parsing and processing
│   ├── retrieval/          # Core retrieval module (4-path recall, fusion, reranking)
│   ├── services/           # Business logic services
│   ├── storage/            # Database and vector store
│   ├── tests/              # Test code
│   ├── utils/              # Utility functions
│   └── workflow/           # RAG workflow and LLM clients
├── data/                   # Data directory (vector DB, uploads, etc.)
├── docs/                   # Project documentation (Chinese)
├── docs_en/                # Project documentation (English)
│   ├── rag_components/     # RAG core component deep dives
│   ├── RAG_architecture_overview.md
│   ├── ...
│   └── ...
├── frontend/               # Frontend code (React)
├── logs/                   # Log files
├── scripts/                # Utility scripts
├── .env.example            # Environment variable example
├── docker-compose.yml      # Docker configuration
├── requirements.txt        # Python dependencies
└── run.py                  # Backend startup script

📝 License

This project is licensed under the Apache License 2.0 with additional commercial use restrictions. See the LICENSE file for details.

Commercial Use Restrictions

• ✅ Backend Server Commercial Use: Permitted for direct commercial purposes
• ❌ SaaS Service Restriction: Not permitted without separate commercial license
• ⚠️ Copyright Attribution: Required for all commercial services (unless separately licensed)

For commercial licensing inquiries, please contact: hqzhon@gmail.com

📧 Contact

For questions, suggestions, or commercial licensing:

• Email: hqzhon@gmail.com
• Telegram: @hqzhon