lance-mcp
LanceDB MCP Server for LLMS is designed to enable large language models to dynamically search and interact with locally stored documents through LanceDB. It offers efficient token usage and ensures data security by storing the index locally.
What is the primary function of the LanceDB MCP Server?
The primary function is to enable LLMs to interact with on-disk documents through agentic RAG and hybrid search, using a serverless vector index and document summary catalog powered by LanceDB.
How does the server ensure data security?
The server ensures data security by storing the index locally, which prevents any data from being transferred to the cloud when using a local LLM.
What are the prerequisites for using the LanceDB MCP Server?
The prerequisites include Node.js 18+, npx, an MCP Client like Claude Desktop App, and installed summarization and embedding models.
How can I run the seed script to create tables in LanceDB?
Use the command npm run seed -- --dbpath <PATH_TO_LOCAL_INDEX_DIR> --filesdir <PATH_TO_DOCS> to create tables for document summaries and vectorized document chunks.
Can I adjust the default summarization and embedding models?
Yes, you can adjust the models in the config.ts file and recreate the index by rerunning the seed script with the --overwrite option.