MCP Rag

🚀 MCP-RAG: Low-Latency RAG Service

A low-latency RAG (Retrieval-Augmented Generation) service architecture based on the MCP (Model Context Protocol).

✨ Features

• Ultra-low latency (<100ms) local knowledge retrieval.
• Dual-mode support: Raw mode (direct retrieval) and Summary mode (retrieval + summarization).
• LLM summarization: Supports intelligent summarization with LLM providers such as Doubao and Ollama.
• Modular architecture: MCP Server serves as a unified knowledge interface layer.
• Asynchronous optimization: Asynchronous calls and model preheating mechanism.
• Scalable design: Reserved interfaces for reranker and cache modules.

📚 Technical Stack

🚀 Quick Start

1. Environment Requirements

• Python >= 3.13
• uv package manager

2. Install Dependencies

uv sync

3. Start the Service

uv run mcp-rag serve

⚠️ Important Note

The service will report an error on the first startup (too lazy to fix). It will work fine after configuring the configuration file.

Web Configuration Page

uv run mcp-rag web

• Access the configuration page: http://localhost:8000/config-page
• Access the document management page: http://localhost:8000/documents-page
• Use the HTTP API: http://localhost:8000/docs (Swagger UI)

4. Configuration Management

MCP-RAG now uses a JSON file for persistent configuration management.

The data\config.json file stores configuration information and supports modification and saving via the Web interface.

Example of default configuration:

{
  "host": "0.0.0.0",
  "port": 8000,
  "http_port": 8000,
  "debug": false,
  "vector_db_type": "chroma",
  "chroma_persist_directory": "./data/chroma",
  "qdrant_url": "http://localhost:6333",
  "embedding_provider": "doubao",
  "embedding_model": "doubao-embedding-text-240715",
  "embedding_device": "cpu",
  "embedding_cache_dir": null,
  "embedding_api_key": "KEY-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
  "embedding_base_url": "https://ark.cn-beijing.volces.com/api/v3",
  "llm_provider": "doubao",
  "llm_model": "doubao-seed-1-6-flash-250828",
  "llm_base_url": "https://ark.cn-beijing.volces.com/api/v3",
  "llm_api_key": "KEY-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
  "enable_llm_summary": false,
  "enable_thinking": false,
  "max_retrieval_results": 5,
  "similarity_threshold": 0.7,
  "enable_reranker": false,
  "enable_cache": false
}

MCP Server Configuration

The Xiaozhi Go Server can interact with MCP-RAG via the MCP protocol. Here is an example configuration:

{
    "mcpServers": {
        "rag": {
            "command": "uv",
            "args": [
                "run",
                "mcp-rag",
                "serve"
            ],
            "env": {
                "PYTHONUNBUFFERED": "1",
                "MODEL_TYPE": "OPENAI",

                "OPENAI_API_KEY": "aa2ae42b-c82b-41ec-bf4e-51c8ab0e4d78",
                "OPENAI_API_BASE": "https://ark.cn-beijing.volces.com/api/v3",
                "OPENAI_MODEL": "doubao-1-5-pro-32k-250115",
                "OPENAI_TEMPERATURE": "0",

                "EMBEDDING_PROVIDER": "OPENAI",
                "OPENAI_EMBEDDING_MODEL": "doubao-embedding-text-240715",

                "COLLECTION_NAME": "default_collection"
            }
        }
    }
}

5. Use the MCP Tool

{
  "name": "rag_ask",
  "arguments": {
    "query": "Query content",
    "mode": "raw",
    "limit": 5
  }
}

📄 License

MIT License

🤝 Contribution

Welcome to submit Issues and Pull Requests!