Vector similarity indexes can now be created

Why it matters

This feature enables efficient vector similarity search on BFloat16 columns, expanding the data types supported for high-performance approximate nearest neighbor queries. It allows users to leverage smaller, efficient floating-point representations for vector data, optimizing storage and computation.

How to use it

To use this feature, create vector similarity indexes on columns defined with the BFloat16 type as you would with other numeric types. For example, use index creation commands specifying a BFloat16 column for similarity search indexing.