qmd

What problem does it solve? Finding information across large collections of markdown notes, meeting transcripts, and documentation is slow with plain grep and inaccurate with naive keyword search. This Skill sets up and operates qmd, a fully local search engine that combines BM25 keyword matching, semantic vector search, and LLM reranking without any cloud dependency. ## Core Features & Use Cases - Three Search Modes: Run instant BM25 keyword search, semantic vector search, or full hybrid search with query expansion and reranking depending on speed and quality needs. - Collection Management: Add directories as named collections, attach context descriptions to improve retrieval, and generate embeddings with a single command. - MCP Integration: Expose search tools to an agent via stdio or a persistent HTTP daemon that keeps models warm for fast repeated queries. - Use Case: A user asks "what decisions were made about the database migration" across months of meeting transcripts; the hybrid query mode expands the question, retrieves candidates in parallel, and reranks them to surface the right note. ## Quick Start Ask the agent to search your indexed notes for a topic, for example: "Search my notes for everything about the API redesign decisions."