mcp -datapizza

This MCP server provides intelligent access to datapizza-ai documentation through a Retrieval-Augmented Generation (RAG) pipeline. It indexes the datapizza-ai documentation from GitHub, stores semantic embeddings in a Qdrant vector store, and exposes an MCP tool called query_datapizza. You can start the server locally, index the documentation, and then query it using MCP clients. The server supports natural language queries to retrieve relevant documentation snippets and generate contextually informed responses. You can also inspect system status via the datapizza://status resource to verify configuration and health. The core workflow is: index the repository to build semantic chunks, run the MCP server to expose query_datapizza, and issue queries through MCP clients to obtain ranked results and generated answers.

Prerequisites:

• Python 3.10 or higher
• Access to OpenAI API key
• Qdrant Cloud account and API key (or a local/remote Qdrant deployment)
• Internet access for indexing and API calls

Installation steps:

1. Clone the repository

git clone https://github.com/datapizza-labs/mcp_server_datapizza.git
cd datapizza-mcp-server

2. Create and activate a Python virtual environment (recommended)

python -m venv venv
source venv/bin/activate  # on macOS/Linux
venv\Scripts\activate     # on Windows

3. Install the package with development dependencies

pip install -e ".[dev]"

4. Prepare configuration

• Create a .env file in the project directory (or set environment variables in your deployment environment) with keys described in the mcp_config section (OPENAI_API_KEY, QDRANT_URL, QDRANT_API_KEY, and optional ones).

5. Index documentation (one-time or re-index with --force)

python -m datapizza_mcp.indexer

6. Start the MCP server

python -m datapizza_mcp.server

Optional: Windows users can use the provided batch script (if available in your deployment) to start the server.

Tips and common considerations:

• Ensure OPENAI_API_KEY, QDRANT_URL, and QDRANT_API_KEY are set and valid before indexing or querying.
• The server relies on the datapizza_mcp.indexer to preprocess and chunk documentation; if indexing fails, check network access to GitHub and the OpenAI/Qdrant services.
• The optional environment variables control embedding model, chunk size, overlap, and the number of results returned by queries. Tuning these can affect latency and result quality.
• If you experience empty results, verify that indexing completed successfully and that the Qdrant collection (datapizza_docs by default) exists and is populated.
• For debugging, set LOG_LEVEL=DEBUG in the environment to get granular logs.
• The MCP tool available is query_datapizza, which accepts a query string and an optional max_results parameter. Use datapizza://status to check configuration and system health.