Can qmd search non-markdown files?

qmd is specifically optimized for Markdown notes, documentation, and knowledge bases to provide the best indexing and retrieval performance for text-heavy files.

What is the difference between search and vsearch?

The 'search' command uses fast BM25 keyword matching for instant results, while 'vsearch' uses semantic vector similarity to find conceptually related content even if keywords don't match exactly.

How does qmd save tokens?

Instead of Claude reading multiple large files to find an answer, qmd searches the index first and directs Claude to only read the specific, relevant sections, reducing context overhead.

Do I need to re-index when files change?

Yes, you can use 'qmd update' to quickly re-index changed files or 'qmd embed' to refresh the vector embeddings for semantic search accuracy.

Is the search performed locally?

Yes, qmd performs local hybrid search on your machine, ensuring fast response times and maintaining privacy for your local documentation and notes.

Quick Markdown Search (qmd)

Name: Quick Markdown Search (qmd)
Author: iagocavalcante

byiagocavalcante

•

学习与文档

Optimizes file exploration and information retrieval in large Markdown codebases to reduce token usage by up to 90%.

qmd is a high-performance local search engine specifically designed for navigating Markdown documentation and knowledge bases within Claude Code. By implementing a hybrid search strategy—combining lightning-fast BM25 keyword matching with semantic vector search—it allows Claude to identify relevant snippets and files before reading them. This prevents the unnecessary consumption of context window tokens, making it an essential tool for developers exploring massive documentation sets or complex project architectures without hitting token limits.

主要功能

01Multi-level search options (search, vsearch, query) for balancing speed and precision

02Efficient index management with fast updates for changed files and collections

03Granular retrieval allows reading specific relevant sections rather than entire files

047 GitHub stars

05Hybrid search engine combining BM25 keyword matching and semantic vector similarity

06Dramatic token optimization, saving up to 90% on codebase exploration tasks

使用场景

01Navigating extensive technical documentation to find specific implementation details

02Locating relevant content across large repositories without reading every file directly

03Efficiently answering questions about project architecture or knowledge base content

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add iagocavalcante/claude-turbo-search qmd

For use in Claude.ai and ChatGPT

Download Skill