How do I limit searches to a specific project or folder?

You can use the -c or --collection flag followed by the collection name to restrict your search results to a specific set of indexed documents.

What is the difference between keyword and semantic search in qmd?

Keyword search (BM25) looks for exact terms and is extremely fast, while semantic search (vector) finds conceptually related content even if different words are used.

Can I search source code files with this skill?

No, qmd is specifically optimized for markdown documentation and notes; for code files, traditional tools like grep or ast-grep are recommended.

How does the hybrid search mode work?

The hybrid 'query' mode combines keyword and semantic results and then applies LLM-powered re-ranking to provide the highest quality results for complex questions.

Do I need to run a special command to enable semantic search?

Yes, you must run the 'qmd embed' command to generate vector embeddings for your collections before using the vsearch or query modes.

QMD Markdown Knowledge Base Search

Name: QMD Markdown Knowledge Base Search
Author: pbdeuchler

bypbdeuchler

0•

内容管理

Enables high-performance keyword, semantic, and hybrid search across local markdown documentation and knowledge bases.

The howto-qmd skill empowers Claude with advanced search capabilities for markdown-based knowledge bases, technical notes, and documentation. By integrating the qmd CLI, this skill allows users to move beyond simple string matching with BM25 keyword search, vector-based semantic search for conceptual queries, and LLM-powered hybrid re-ranking for superior context retrieval. It is an essential tool for developers and writers who need to navigate large collections of documentation to gather context for complex tasks or find specific information when exact terms are forgotten.

主要功能

01Flexible output formats including JSON, XML, CSV, and Markdown

02Triple-mode search: Keyword (BM25), Semantic (Vector), and Hybrid (Re-ranked)

030 GitHub stars

04LLM-powered hybrid re-ranking for high-relevance conceptual matching

05Local markdown document indexing and collection management

06Advanced filtering by relevance scores, file size, and collections

使用场景

01Gathering context from scattered markdown notes before starting a development task

02Finding relevant documentation when the exact terminology or phrasing is unknown

03Automating the retrieval of specific document sections for structured LLM processing

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add pbdeuchler/llm-plugins howto-qmd

For use in Claude.ai and ChatGPT

Download Skill