similarity-search-patterns

This skill provides multi-backend templates for Pinecone, Qdrant, pgvector, and Weaviate to perform vector upserts and nearest-neighbor queries, enabling fast similarity search over large vector collections for semantic retrieval and RAG.

You can perform semantic search by running backend-specific templates that execute batch vector upserts and top-k nearest-neighbor queries, applying best practices for index types, distance metrics, and hybrid search strategies across large vector collections.

Yes, this approach supports pgvector and Weaviate alongside Pinecone and Qdrant, providing core vector operations including upsert, search, and optional cross-encoder reranking for production-grade semantic retrieval.

Yes, you can apply optional cross-encoder reranking to nearest-neighbor search results to improve retrieval accuracy in semantic search, RAG retrieval, and recommendation systems.

You need hybrid search strategies when combining semantic and keyword-based retrieval to improve search accuracy over large vector collections, using provided best practices for index types and distance metrics.