pg_textsearch

BM25 extension for PostgreSQL by Tiger Data (formerly Timescale). Implements BM25 as a native index access method — same architectural approach as psql_bm25s, but with a pgvector-style operator interface and Block-Max WAND top-k optimization.

• GitHub: https://github.com/timescale/pg_textsearch
• License: PostgreSQL license — most permissive in the Postgres BM25 ecosystem; use anywhere, for anything.

Key features

• CREATE INDEX … USING bm25 — native index access method
• Query via distance-style operator: ORDER BY content <@> 'search terms' LIMIT k
• Block-Max WAND top-k pruning for fast retrieval
• Parallel index builds
• Partitioned-table support and expression/partial indexes
• Hybrid pairing with pgvector and pgvectorscale

Maturity

Open-sourced late 2025. As of June 2026: past v1.0, self-declared production-ready (v1.4.0-dev), ~1,800 GitHub stars within weeks of release. Young relative to ParadeDB but no longer a preview.

Operational caveat: requires shared_preload_libraries, which limits availability on managed Postgres (RDS, Cloud SQL, Supabase) until providers add support.

Related

• PostgreSQL · BM25 · Block-Max WAND · pgvector
• ParadeDB · psql_bm25s · vchord_bm25 — other Postgres BM25 extensions; see Search using PostgreSQL
• Tiger Data — author/maintainer