qmd - Local Document Search

Semantic + full-text search across local markdown docs using local LLMs.

When to Use

Use qmd when the user wants to search their personal notes, documentation, or knowledge base. This includes Obsidian vaults, markdown notes directories, and any indexed collections.

Search Commands

qmd search "query" # BM25 full-text search qmd vsearch "query" # Semantic vector search qmd query "query" # Hybrid search with LLM reranking (best quality)

Search Flags

Flag Purpose

-n <num>

Limit results (default: 5)

-c, --collection

Restrict to specific collection

--all

Return all matches

--min-score <num>

Filter by relevance threshold

--full

Display complete document content

--line-numbers

Include line numbers

Output Formats

Flag Format

--files

CSV: docid, score, filepath, context

--json

Structured JSON with snippets

--csv

Comma-separated values

--md

Markdown formatting

--xml

XML structure

Document Retrieval

qmd get "path/to/doc.md" # Get specific document qmd get "#abc123" # Get by document ID qmd multi-get "journals/*.md" # Get multiple docs by glob

Collection Management

qmd collection list # View all collections qmd collection add ~/notes --name notes qmd ls notes # List files in collection qmd status # Index health

Examples

High-quality search with reranking

qmd query -n 10 "API design patterns"

Full document content for LLM context

qmd search --md --full "error handling"

Search specific collection

qmd search "meeting notes" -c work

Workflow

• Use qmd collection list to see available collections

• Use qmd query for best search quality (uses LLM reranking)

• Use --full when you need complete document content

• Use qmd get to retrieve specific documents by path or ID

Quick Start for Common Requests

User says Run this

"search my notes for X" qmd query "X"

"what did I write about X" qmd query --full "X"

"find meeting notes about X" qmd search -c obsidian "meeting X"

"show me my notes on X" qmd query -n 3 --full "X"