qmd

Index local documents for hybrid BM25 and vector search with reranking.

Updated Mar 16, 2026
One-click install
npx skills add https://github.com/fufuonline/openclaw --skill qmd-fufuonline
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: qmd
Source: https://github.com/fufuonline/openclaw/tree/main/skills/qmd
Command: npx skills add https://github.com/fufuonline/openclaw --skill qmd-fufuonline

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Local searches across documents are slow and fragmented. qmd accelerates this by indexing content and enabling fast BM25 and vector-based ranking, delivering relevant results quickly.

Core Features & Use Cases

  • BM25 + vector hybrid search for local files
  • Multi-collection indexing with simple management
  • Use case: find relevant docs across a codebase, PDFs, and text files with a single query

Quick Start

Index a folder of documents and search with a query using BM25 and vector reranking.

Frequently Asked Questions about qmd

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I perform hybrid search across local documents using BM25 and vector embeddings?

To perform local document search, you index files into multiple collections and query them using a single command. The system applies BM25, vector embeddings, and reranking to retrieve relevant results across diverse file types like codebases and PDFs.

Can I index and search diverse file types like PDFs and text files offline?

Yes, you can index diverse file types like codebases, PDFs, and text files entirely offline. The tool operates in local environments without requiring external network connectivity for indexing or retrieval operations.

What is the best way to manage multiple collections for local file search?

The best way to manage multiple collections is using the tool's configurable multi-collection indexing features. This allows you to organize different document sets separately and query across them with mixed scoring and deterministic retrieval.

How does BM25 mixed scoring work with vector reranking?

BM25 mixed scoring works by combining keyword-based frequency matching with vector-based semantic similarity. The system then applies a reranking step to refine these combined scores, ensuring fast and accurate document retrieval.

Does the local search CLI support deterministic retrieval in MCP mode?

Yes, the local search CLI supports deterministic retrieval and includes an MCP mode. This ensures consistent search results across queries while allowing integration into broader automated workflows.