July 2025, Release Update 23.9

Included are some notable Oracle AI Vector Search updates with Oracle Database 23ai, Release Update 23.9.

Feature Description

IVF Index Reorganization

The quality of an IVF index may degrade over time if updates are made to the base table after the general vector distribution. IVF index reorganization permits the index to be reorganized while it remains available for DMLs and queries.

Inverted File Flat Vector Indexes Online Rebuild

HNSW Index Snapshots

Incremental snapshots are introduced as an HNSW graph refresh mechanism optimized for minimal, frequent updates, balancing performance and memory use.

HNSW Incremental Snapshots

Partition Maintenance Operation support for global IVF and HNSW Indexes

Partition Maintenance Operations can now be performed on partitioned tables that have global IVF and HNSW Indexes, including operations such as adding, dropping, merging, and splitting partitions.

SPARSE HNSW Indexes

Hierarchical Navigable Small World (HNSW) vector indexes support SPARSE vectors.

SPARSE Vectors

Creating Hybrid Vector Index using DBMS_SEARCH API

Hybrid Vector Indexes (HVI) can now be created using the DBMS_SEARCH PL/SQL package.

Creating Hybrid Vector Indexes using DBMS_SEARCH API

Enhancements for altering Hybrid Vector Indexes

Existing Hybrid Vector Indexes can be altered using the SQL ALTER INDEX syntax to change settings such as replacing the vectorizer preference, the embedding model, or the type of vector index.

ALTER INDEX

Enhancements for searching Hybrid Vector Indexes using the DBMS_HYBRID_VECTOR.SEARCH API

The SEARCH API for Hybrid Vector Indexes adds support for Weighted Reciprocal Rank Fusion (WRRF) as a method to search scores from keyword and semantic search results. It is also possible to define custom score calculations using a new parameter: score_calc.

Additional enhancements include a new query parameter for keyword searches of Hybrid Vector Indexes as well as the ability to filter results using standard relational logic.

SEARCH