qmd

QMD is a local retrieval and search skill designed to help users query and explore markdown-based knowledge bases, including notes, documentation, wikis, transcripts, and other structured or unstructured text collections stored locally. It solves the problem of quickly finding relevant information within personal or project-specific markdown repositories, especially in situations where generic web search is unnecessary or less effective because the needed information already exists in a local indexed corpus.

The skill provides multiple search and retrieval capabilities through a command-line interface or MCP server integration. It supports collection discovery and system inspection via commands such as listing available collections and checking system status. QMD offers different search strategies, including fast lexical (BM25-style) search for exact terms, hybrid query search for semantic understanding and reranking, and structured queries that combine intent, lexical anchors, vector semantics, and hypothetical document generation signals. It also includes retrieval commands to fetch full documents or multiple documents, ensuring that responses can be grounded in actual source content rather than partial snippets. Additionally, MCP-based structured query execution enables programmatic multi-strategy search workflows.

QMD is intended for users who maintain large markdown knowledge bases and need efficient ways to retrieve information without manually browsing files. It is especially useful for developers, researchers, writers, and knowledge workers who rely on personal wikis, project documentation, or long-term note archives. It is also suitable for AI-assisted systems that require grounded retrieval before generating answers, helping ensure responses are based on real indexed content rather than hallucinated or external information.