Postgres Hybrid Text Search

This combines PostgreSQL's BM25 keyword search with pgvector semantic search using Reciprocal Rank Fusion to merge results. You run two queries in parallel (one for exact term matching, one for meaning), then fuse them client-side in Python or TypeScript. It's the right move when your users mix specific terms like product codes with conceptual queries, though it adds complexity over single-method search. The guide uses pg_textsearch for true BM25 ranking, which is currently in prerelease. Includes weighting strategies if you want to favor one method over the other, plus optional ML reranking with cross-encoders for the final candidate set.

Install to Claude Code

npx -y skills add timescale/pg-aiguide --skill postgres-hybrid-text-search --agent claude-code