Lakebase Search reached general availability on AWS and Azure with native hybrid vector and text search.

Databricks announced the general availability of Lakebase Search on AWS and Azure, embedding semantic, keyword, and hybrid search directly inside Lakebase Postgres [2, 6]. The platform also introduced an autoscaling compute cache that operates alongside standard Postgres shared buffers to reduce storage layer reads and increase throughput [18, 24]. On the developer side, Databricks CLI version 1.19.0 added expiration flags for Postgres branch creation [3], and Replit integrated automated provisioning for Lakebase environments [23]. Databricks also mitigated a memory safety bug in the PostGIS extension across managed environments [39].

Architectural patterns focused on operational data for AI agents. Databricks outlined patterns combining Temporal with Lakebase and Change Data Feed to preserve state across agent executions [27]. Multiple production teams shared architectures using Lakebase alongside Unity Catalog to support agent workflows and operational applications [14, 19]. At Big Data LDN, presenters highlighted LTAP patterns where Postgres storage servers read shared data without dedicated synchronization pipelines [11].

Community members also documented operational limits. Developers reported eight second timeouts on the Lakebase Data API when serverless compute was offline [35], authentication issues with GCP service principal tokens [10], and disabled branch states after auto archive events [31]. In addition, users highlighted that while Lakebase automatically manages failover and patching within a single region, cross region disaster recovery remained in private preview [8, 20].