Muxue Rag

Provides local RAG indexing and MCP-driven search for text documents

python

0

GitHub Stars

python

Language

3 months ago

First Indexed

3 weeks ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "musnows-muxue_rag_mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/rag_mcp",
        "run",
        "mcp_rag_tool",
        "--serve"
      ],
      "env": {
        "RAG_MCP_CONFIG": "/path/to/rag_mcp/config.yaml"
      }
    }
  }
}

This MCP server enables local retrieval-augmented generation by indexing text files in a directory and exposing search and raw file retrieval through the MCP protocol. It helps you quickly locate relevant documents and read their original contents via an MCP client.

How to use

You can use an MCP client to connect to this tool and perform two core actions: search the indexed documents with keywords and read the raw content of a specific file. First, start the MCP server in a directory where your documents reside, then configure your MCP client to point at that server. Keep the indexing directory consistent so all queries reference the same RAG data.

How to install

Prerequisites you need before installation:

Python 3.13 or later is required.

A local package manager is recommended. Use uv to manage packages and run the tool.

Run the following commands to set up the tool locally.

Install steps

# Clone the project
git clone <repository_url>
cd rag_mcp

# Install dependencies via the preferred local runner
uv sync

Configure the tool

llm:
  service_type: "local"  # or "openai" and so on
  base_url: "http://localhost:1234/v1"  # LLM service API address
  api_key: "your-api-key"  # if required
  timeout: 60

model:
  name: "text-embedding-qwen3-embedding-4b"
  context_window: 4096
  temperature: 0.7

processing:
  chunk_count: 5

Run the server and build the index

Build the index for your documents using the following command.

uv run mcp_rag_tool --dir /path/to/your/documents

Start the MCP server to serve MCP clients.

uv run mcp_rag_tool --serve

If you want to serve for a specific directory, lock the query directory

uv run mcp_rag_tool --dir /path/to/your/documents --serve

Additional usage notes

The indexing is incremental by design. It reindexes only files that are new or modified since the last run.
Hidden directories (such as those starting with a dot) are auto-filtered out to avoid indexing system or sensitive files.
The tool supports common plain text formats like .txt, .md, .json, .py, .js, and more.

Available tools

search_rag

Search the index for documents matching a keyword and return the matched content along with the source file path.

read_raw_file

Read and return the raw contents of a specified file for analysis.