HiRAG-MCP

Python server implementing Model Context Protocol (MCP) for HiRAG.

Features

• Naive RAG retrieval
• HiRAG RAG retrieval

API

Tools

• naive_search

  • Perform a naive search over the knowledge base.
  • Inputs:
    • query: The search query text
    • max_tokens (optional): Optional maximum tokens for the response

• hi_search

  • Perform a hybrid search combining both local and global knowledge over the knowledge base. (default to use)
  • Inputs:
    • query: The search query text
    • max_tokens (optional): Optional maximum tokens for the response

Test

uv (recommended)

Make sure uv is installed, if not:

curl -LsSf https://astral.sh/uv/install.sh | sh

Then create the virtual environment and install dependencies using uv as below:

uv venv
source .venv/bin/activate
uv sync

Running the server

Note that we already have a test folder and test.txt file, which are the results after indexing test.txt with hi_indexing_openai.py. You can directly use the following command to run the MCP server and test the tools:

mcp dev server.py

You can also use hi_indexing_openai.py to index your own text files. Remember to adjust the working_dir in config.yaml accordingly.

License

This MCP server is licensed under the MIT License. This means you are free to use, modify, and distribute the software, subject to the terms and conditions of the MIT License. For more details, please see the LICENSE file in the project repository.