What are HyDE queries and when should I use them?

HyDE (Hypothetical Document Embeddings) creates a hypothetical answer to your question before searching. Use it for complex topics where you don't know the specific vocabulary used in the source documents.

How do I set up the QMD skill for Claude Code?

You need to install the QMD CLI globally via 'npm install -g @tobilu/qmd', add your markdown directories as collections using 'qmd collection add', and run 'qmd embed' to index them.

Can I filter my search to specific folders?

Yes, QMD supports collection filtering, allowing you to narrow down searches to specific groups of markdown files like 'docs', 'notes', or 'journals'.

What is hybrid search in QMD?

QMD combines BM25 lexical search for exact keyword and code identifier matches with vector semantic search for conceptual understanding, providing the best possible recall and precision.

Does QMD work with code blocks inside markdown files?

Yes, the lexical search is optimized for code identifiers and technical terms, making it highly effective for searching developer-centric markdown documentation.

QMD Markdown Knowledge Search

Name: QMD Markdown Knowledge Search
Author: ehc-io

byehc-io

•

Content Management

Searches and retrieves information from local markdown knowledge bases using advanced hybrid lexical and vector search techniques.

QMD (Query Markdown) is a specialized search engine and MCP server designed to help Claude navigate large local markdown repositories, personal notes, and technical documentation. It provides a robust retrieval-augmented generation (RAG) layer by combining lexical (BM25) keyword matching with semantic vector search. By supporting advanced techniques like Hypothetical Document Embeddings (HyDE) and query expansion, QMD ensures that Claude can find precise information even when the exact terminology isn't known, making it an ideal companion for developers managing extensive documentation or 'second brain' notes.

Key Features

013 GitHub stars

02Hybrid search combining BM25 lexical keywords and semantic vector models

03MCP server integration for seamless context retrieval within Claude Code

04HyDE support for generating hypothetical answers to improve retrieval accuracy

05Multi-collection filtering to isolate searches to specific documentation sets

06Bulk document retrieval via glob patterns and unique document IDs

Use Cases

01Querying internal engineering documentation for specific architectural patterns or code snippets

02Searching through a personal knowledge base or research notes to find related concepts

03Retrieving relevant technical specifications while performing codebase migrations or refactors

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add ehc-io/qmd qmd

For use in Claude.ai and ChatGPT

Download Skill