Show HN: pg_textsearch — BM25 Relevance-Ranked Full-Text Search for Postgres, Challenging Elasticsearch

GitHub / Hacker News (169pts, 47cmts)·medium signal

Tiger Data (formerly Timescale) open-sourced pg_textsearch, a PostgreSQL extension that brings BM25 relevance-ranked keyword search directly into Postgres without external search infrastructure like Elasticsearch. Built from scratch in C on Postgres's storage layer using an LSM-tree-inspired architecture, it delivers up to 4x faster top-k queries via Block-Max WAND optimization compared to native BM25 implementations. Supports PostgreSQL 17 and 18 under Postgres OSS license. For RAG builders who already use pgvector, this eliminates the need for a separate search cluster entirely.