librarian-library-search

What problem does it solve?

Provides natural-language, semantic search across locally indexed libraries so agents can find relevant research content even when exact keywords differ. It eliminates noisy grep results by ranking document chunks by meaning and returning source metadata for citation.

Core Features & Use Cases

• Vector similarity search against a local Qdrant collection to surface semantically relevant document chunks.
• Source-aware results that include similarity score, file path, chunk index, and source URL for trustworthy citations.
• Scoped filtering and model alignment to restrict searches to a named library and ensure the same FastEmbed embedding model is used for consistent results.
• Use Case: A security researcher asks for "reentrancy vulnerability in pull-payment pattern" and receives ranked excerpts from multiple indexed libraries with direct links to the originating files.

Quick Start

Search the local librarian libraries for "reentrancy vulnerability in pull-payment pattern" and return the top five results with scores and source metadata.