qmd-search

On-device semantic search for Markdown notes works by indexing your local documents into vector embeddings and retrieving matching content based on meaning rather than exact keywords. This approach keeps personal notes, transcripts, and knowledge bases completely private on your desktop.

Yes, integrating vector search into an Electron app with React components is supported through a secure IPC flow connecting the front-end UI to the local indexing backend. Standardized configurations for collections, embeddings, and multiple search modes streamline the setup process.

Indexing meeting transcripts for secure local retrieval is best handled by on-device vector search that processes files into embeddings without sending data to external services. Set up a dedicated collection, index your transcripts, and run queries to find semantically relevant sections instantly.

Yes, Bun is a required dependency for running this on-device Markdown search Skill, as it executes the underlying scripts and processes vector embeddings. Ensure your environment has Bun configured before setting up your document collections.

Set up a collection and index Markdown documents by installing the search tool globally, configuring collection settings, pointing it at your local files, and running a sample query to verify results. The process applies standardized configs for embeddings and search modes.

Available search modes for querying a local Markdown knowledge base include standard search, vector search, and query modes. These options retrieve documents based on exact matching or semantic similarity depending on your specific retrieval needs.