qmd

Local semantic retrieval uses embeddings to find conceptually relevant passages in your knowledge base. It combines keyword lookup with semantic search and reranking, allowing you to find information even without exact keyword matches in your notes.

Yes, you can search meeting transcripts and project documentation locally. The Skill queries local qmd collections containing your transcripts and docs, applying hybrid reranking to retrieve the most relevant passages for your research.

Hybrid reranking combines exact keyword lookup with semantic retrieval to rank search results. This mechanism ensures both precise matches and concept-level queries yield highly relevant passages from your local knowledge base.

You need local qmd collections and embeddings to run local knowledge base search. An optional daemon or stdio server setup can be configured to deliver fast offline search results and enable flexible MCP-connected workflows.

MCP supports searching personal notes and transcript archives through connected workflows. It enables the assistant to interface with local qmd collections, retrieving relevant information directly from your terminal or AI workflow.

The best way to find exact matches and concept-level answers is using a hybrid retrieval system. By indexing local documents with embeddings, the search combines keyword precision with semantic understanding for comprehensive recall.