Databricks makes Lakebase Search generally available in Postgres

Databricks has made Lakebase Search generally available. The service adds scalable vector search and BM25 full-text search directly in Postgres for AI agents that need retrieval at large scale.

Databricks says the service sets a new frontier for price-performance, latency, and recall on VectorDBBench, and that it is 4x cheaper than running pgvector for the same workload. It uses pay-per-use pricing, with zero compute cost when idle. Details are in Databricks’ Lakebase Search announcement.