qmd

What This Skill Does

The QMD skill allows you to efficiently search and retrieve information from your local markdown-based knowledge bases. It leverages the qmd command-line interface to provide powerful search capabilities, including traditional keyword matching (BM25), semantic vector search, and a hybrid approach with LLM re-ranking for the highest quality results. This is ideal for querying indexed notes, documentation, meeting transcripts, code snippets, and any other markdown files you've organized locally. The skill requires the qmd CLI tool to be installed, which can be done via bun.

Installation

To use the QMD skill, you first need to install the qmd command-line interface. Open your terminal and run the following command:

bun install -g https://github.com/tobi/qmd

After installation, you'll need to set up your local knowledge base. This involves adding a collection of markdown files and generating embeddings for semantic search:

# Add a collection, e.g., your local notes directory
qmd collection add ~/notes --name notes --mask "**/*.md"

# Generate embeddings (necessary for vsearch and query commands)
qmd embed

Remember to always use the --json flag when invoking qmd commands with this skill to ensure structured output.

Use Cases

• Developer Documentation: Quickly find specific API details, configuration options, or troubleshooting guides within your project's markdown documentation.
• Personal Notes & Knowledge Management: Retrieve notes, meeting summaries, or research findings based on conceptual meaning rather than exact keywords.
• Technical Writing: Search through drafts, guides, and reference materials to ensure consistency and accuracy.
• Meeting Transcripts: Find action items or decisions made during past meetings.
• Code Snippets: Organize and search through markdown files containing code examples.

Example Prompts

• "Find information on the authentication flow for user logins."
• "What are the best practices for error handling in our documentation?"
• "Show me meeting notes about the Q3 planning session."

Tips & Limitations

• Performance: qmd search is the fastest for keyword-based queries. qmd vsearch offers better semantic understanding but is slower. qmd query provides the best quality through LLM re-ranking but is the slowest and has a potential ~1-minute cold start latency for vector initialization. Prefer search for quick, interactive lookups.
• Indexing: Ensure your markdown files are correctly added to a qmd collection and that embeddings are generated (qmd embed) for semantic searches to function effectively.
• Full Content Retrieval: Use the --full option with search commands if you need the entire content of the matched documents returned in the results, not just snippets.
• Specificity: For best results, be as specific as possible in your queries. For semantic searches (vsearch, query), use natural language that accurately reflects the concept you're looking for.
• Installation Dependency: The skill relies on the qmd CLI being installed and configured on your system. Ensure the bun package manager is available.