librarian-library-search

Semantic vector search over local library content uses a local Qdrant collection to rank indexed document chunks by meaning. This surfaces relevant excerpts for natural language queries, returning JSON-ranked results with similarity scores and source metadata.

Semantic search returns better results than exact text grep because it ranks document chunks by meaning rather than exact keyword matches. This eliminates noisy grep results by finding relevant research content even when the exact phrasing differs.

Yes, querying a local Qdrant knowledge base requires the same FastEmbed embedding model used during indexing. This model alignment ensures consistent vector similarity scores and accurate semantic retrieval of document chunks.

Yes, you can restrict semantic search results to a specific named library using scoped filtering. This limits the vector similarity search to a subset of the local Qdrant collection, returning source-aware results only from the specified library.

Ranked document chunks from a vector similarity search include similarity scores, file paths, chunk indices, and source URLs. This source-aware metadata provides trustworthy citations for natural language queries against indexed libraries.