qmd

What This Skill Does

The qmd skill is a high-performance, local-first search engine designed specifically for markdown-based knowledge bases, notes, and documentation. Unlike cloud-reliant search tools, qmd functions entirely on your local machine, ensuring your data remains private while providing lightning-fast retrieval. It indexes your markdown files into collections, supporting both traditional keyword-based BM25 search for instant results and advanced vector-based semantic search for finding contextually relevant information when exact keywords are unknown.

Installation

To install the qmd skill, ensure you have Bun version 1.0.0 or higher installed. On macOS, you will also need to install SQLite extensions via brew install sqlite. Once prerequisites are met, install the global package using bun install -g https://github.com/tobi/qmd. After installation, initialize your collections by running qmd collection add /path/to/notes --name your-collection-name. Finally, run qmd embed to process your files into a searchable vector index. Always ensure your PATH includes the Bun binary directory to keep commands accessible.

Use Cases

This skill is ideal for knowledge workers, developers, and writers who maintain extensive personal markdown wikis or technical documentation. It excels when you need to cross-reference information across hundreds of files, such as finding a specific configuration snippet, retrieving historical project requirements, or discovering connections between disparate notes that share a thematic link but lack explicit cross-links.

Example Prompts

1. "Search my notes for the documentation on setting up the CI/CD pipeline environment variables."
2. "I need to find related notes that discuss distributed systems architecture; please perform a semantic search across my knowledge base."
3. "Retrieve the markdown document about the Q3 retrospective from my notes collection and list the key performance metrics found inside."

Tips & Limitations

Use qmd search as your default command; it is significantly faster than other modes and serves most needs. Use qmd vsearch sparingly, as it triggers a local LLM cold-start that can take up to a minute to process. Avoid qmd query for daily tasks to prevent unnecessary timeouts. Because this tool indexes raw markdown, it is best suited for text-heavy content rather than source code repositories. Keeping the underlying process warm or using a persistent daemon (if available) is recommended for power users requiring frequent, high-performance semantic retrieval.