Databricks makes Lakebase text and vector search generally available

Databricks says two search extensions for its Lakebase Postgres database are now generally available on AWS and Azure. One searches for similar items using vectors, numerical representations of data; the other ranks text matches using the BM25 method. Customers can combine both in a database query alongside filters and live operational data. Databricks says its vector extension achieved twice the throughput of the next best system in a 100 million item benchmark. Those performance figures come from company tests. The extensions give Lakebase customers a way to run search inside their operational database instead of maintaining a separate search system.