databricks Databricks Release Notes ·

Databricks Lakebase Search reaches general availability

aidatabricksgaengineer
feature announcement

Lakebase Search is now generally available, bringing scalable vector and keyword search capabilities directly to Lakebase projects. It enables semantic and exact-term search within a single database via two new Postgres extensions, removing the need for a separate search stack. The `lakebase_vector` extension supports approximate nearest-neighbor search, compatible with pgvector, and scales to over a billion vectors. The `lakebase_text` extension provides BM25 full-text search, adding advanced ranking and top-K pushdown to standard tsvector operations.

  • Lakebase Search generally available for vector and keyword search
  • lakebase_vector extension for approximate nearest-neighbor search
  • lakebase_text extension for BM25 full-text search
Features (3)
  • Lakebase Search generally available for vector and keyword search

    Lakebase Search now offers scalable vector and keyword search directly within Lakebase projects. This functionality allows users to perform semantic and exact-term searches in a single database via Postgres extensions, eliminating the need for an external search stack.

  • lakebase_vector extension for approximate nearest-neighbor search

    The `lakebase_vector` extension introduces the `lakebase_ann` index type, providing approximate nearest-neighbor vector search. It maintains drop-in compatibility with pgvector, using its types, operators, and query syntax, while scaling to over a billion vectors with significantly faster index build times.

  • lakebase_text extension for BM25 full-text search

    The `lakebase_text` extension adds the `lakebase_bm25` index type for enhanced full-text search capabilities. It works with standard tsvector types and operators, providing BM25 ranking and top-K pushdown that surpass PostgreSQL's native GIN index features.

Read the original announcement →

https://docs.databricks.com/aws/en/release-notes/lakebase#lakebase-search-is-generally-available

Related releases