qmd

What This Skill Does

The qmd skill provides a highly efficient and versatile local search engine for your markdown files, notes, and documentation. It goes beyond simple keyword matching by integrating BM25 full-text search, semantic vector search, and LLM-based reranking, all executed locally on your machine. This means you can quickly find information without relying on external services or exposing sensitive data. The skill excels at discovering files, locating specific code snippets, identifying documentation patterns, and gathering relevant context for AI-driven tasks. It is designed as a superior alternative to traditional find commands for searching within large directories, preventing system hangs and providing richer results.

Installation

To install the qmd skill, use the following command:

clawhub install openclaw/skills/skills/bheemreddy181/qmd-local-search

Use Cases

• Rapid File Discovery: Replace slow find commands, especially in extensive directory structures, to quickly locate files based on names or content.
• Content Search: Perform both keyword-based (BM25) and meaning-based (semantic vector) searches across your notes, documentation, and codebases.
• Code Navigation: Efficiently find specific code implementations, configuration files, or recurring patterns within your projects.
• Contextual Information Retrieval: Gather precise snippets of information to provide context for other AI agents or tasks, improving accuracy and relevance.
• Knowledge Management: Organize and search through personal knowledge bases, project documentation, and technical notes.

Example Prompts

1. "Find all markdown files in my ~/Documents/Projects directory that mention 'blockchain consensus algorithms' and show me the file paths and relevance scores."
2. "Semantically search my notes for documentation on how to implement exponential backoff in Python, even if the exact phrasing isn't used."
3. "I need to find the configuration file for the user_service in the backend collection, showing the first 50 lines."

Tips & Limitations

• Collections are Key: Always utilize collections (-c <collection_name>) to narrow down search scope and improve performance. Create collections for distinct projects or data sets.
• Keep Indexes Updated: After adding, modifying, or deleting files within a collection's scope, run qmd update to re-index your data and ensure search accuracy.
• Enable Vector Search: For semantic search capabilities, run qmd embed once. This process might take a few minutes depending on the size of your indexed data.
• Performance: While designed for speed, the initial indexing and embedding processes can consume resources. Subsequent searches are typically very fast.
• Local Operation: All search and embedding operations occur locally, ensuring data privacy. No internet connection is required for core functionality after installation.
• File Masking: When adding collections, use the --mask option to include only relevant file types (e.g., *.md,*.py) to optimize indexing and search.