Lakebase Search: vector and BM25 on Neon

Neon launched the beta of Lakebase Search, combining hybrid vector and full-text retrieval through two Postgres extensions. It uses IVF plus RaBitQ indexing instead of HNSW, and BM25 search with top-K pushdown optimization, with indexes stored on object storage so they persist through scale-to-zero and branch instantly.