A MCP server for Godot RAG
@weekitmo
This MCP server is used to provide Godot documentation to the Godot RAG model.
Overview
What is Godot RAG?
Godot RAG is an MCP server that provides Godot documentation to a Godot RAG (Retrieval-Augmented Generation) model. It is intended for developers using AI assistants to access Godot documentation.
How to use Godot RAG?
Configure the MCP client with the server command python and arguments pointing to main.py, the chroma_db path (-d), and the collection name (-c). Run the setup steps: clone Godot docs, convert RST to Markdown, chunk the files, create a vector database, then start the server.
Key features of Godot RAG
- Retrieves Godot documentation for RAG models.
- Uses ChromaDB as vector store.
- Supports multiple sentence transformer models.
- Provides setup scripts for document ingestion.
Use cases of Godot RAG
- Enhance AI coding assistants with Godot documentation.
- Build a custom Q&A bot for Godot engine queries.
- Provide context-aware answers during game development.
FAQ from Godot RAG
What models does Godot RAG support?
It supports all-MiniLM-L6-v2, bge-m3, and bge-large-zh-v1.5 sentence transformer models.
How do I set up the vector database?
Run python download_godot_docs.py, then convert_rst2md.py, then chunker.py, then vectorizer.py with the appropriate arguments.
What are the runtime requirements?
The server requires Python 3.12, uv for virtual environment, and dependencies from uv sync.
Where is the vector data stored?
The chroma_db is stored at a path specified by the -d argument when starting the server.
How do I debug the server?
Use npx @modelcontextprotocol/inspector with the uv run main.py command and the chromadb path and collection name arguments.
