qmd

You can search local notes and transcripts entirely offline by indexing markdown files and text documents into a local knowledge base, then querying them through a hybrid retrieval pipeline combining BM25 keyword matching and vector similarity search.

Hybrid search for a personal knowledge base combines BM25 exact keyword matching, vector semantic similarity, and LLM-based reranking to retrieve the most relevant results from locally indexed notes, documents, and meeting transcripts.

Yes, semantic search runs entirely offline on meeting transcripts. The Skill processes local text files, performs vector similarity search, and applies LLM reranking without uploading any personal data to the cloud.

The best way to rerank local search results is using an LLM-based reranker integrated into a hybrid retrieval pipeline, which refines BM25 and vector search outputs to improve relevance ranking across markdown notes and text files.

Yes, local hybrid search supports optional MCP integration alongside a CLI interface, allowing you to perform keyword and semantic search with LLM reranking across your personal knowledge store within connected environments.

No, you do not need an internet connection. The hybrid retrieval pipeline operates entirely offline, performing vector similarity search, BM25 keyword matching, and LLM reranking directly on your local machine without external API calls.