Google Cloud’s AlloyDB added hybrid-search capabilities combining vector and full-text results through a single Reciprocal Rank Fusion SQL function.

AlloyDB’s update addresses the fusion of semantic and keyword search results. Google explains that vector search supplies semantic context while full-text search (FTS) supplies keyword precision, particularly for AI and retrieval-augmented generation (RAG) applications.

The new approach consolidates result fusion into a single SQL function powered by Reciprocal Rank Fusion (RRF). Previously, developers could need separate queries, score normalization, joins on document identifiers, and application-side re-ranking to merge vector-distance and text-relevance results.

The update also adds text-search indexing capabilities. Google describes a new RUM extension that stores word positions directly in the index for relevance ranking and phrase matching, alongside native BM25 indexing for keyword-based scoring.

External search integration is another component of the release. A new external-search Foreign Data Wrapper (FDW) supports queries against Elasticsearch, OpenSearch, and Solr clusters from within AlloyDB; the supplied excerpt does not quantify latency or relevance improvements.