Summary

This project delivers a professional medical domain Retrieval Augmented Generation (RAG) system using LangChain and Milvus. It offers an end-to-end solution for intelligent medical Q&A, featuring hybrid retrieval and an automated RAG agent, targeting researchers and power users needing robust, domain-specific AI applications.

How It Works

The system leverages multi-vector hybrid retrieval (dense + sparse BM25) for enhanced recall and domain-specific optimizations like medical tokenization. A core RAG agent automates retrieval parameters, fact-checking, and web search, simplifying complex RAG operations. Flexible LLM/embedding provider support and an end-to-end data pipeline facilitate integration and deployment.

Quick Start & Requirements

• Installation: Clone repo, git lfs pull, create/activate Conda env (conda env create -f environment.yml, conda activate rag), install project (pip install -e .).
• Prerequisites: Git LFS, Conda, Milvus (start script), Ollama (optional, for local models). Environment variables for LLM/embedding providers, Langchain API key, and Tencent Cloud credentials (for network search) are necessary.
• Configuration: Adjust src/MedicalRag/config/app_config.yaml for Milvus, models, and data.
• Links: GitHub Repository

Highlighted Details

• Medical Domain Optimization: Uses pkuseg for specialized medical tokenization and vocabulary.
• Hybrid Retrieval: Combines dense and sparse vectors for >85% Q&A accuracy, outperforming single methods.
• RAG Agent: Automates retrieval decisions, fact verification, and web search initiation.
• Flexible Providers: Supports OpenAI, Ollama, and others via configuration.
• Performance: Hybrid retrieval < 500ms, handles millions of documents and high concurrency.

Maintenance & Community

The project welcomes Issues and Pull Requests. Specific community links (Discord/Slack) or a roadmap are not detailed in the README.

Licensing & Compatibility

Licensed under the MIT License, which is permissive for commercial use and integration into closed-source applications.

Limitations & Caveats

Intended for research/learning, not professional medical advice. Users must ensure data security and regulatory compliance (HIPAA, GDPR), ideally deploying privately. Network search requires Tencent Cloud credentials. Minimum 16GB RAM recommended.