ArchitectureA supervised JVM-class runtime — OLTP on seven engines, OLAP on three. AI-native, MCP-native, observable as plain SQL.Read the architecture
Está viendo la edición Perú. Está viendo la edición Colombia. You're viewing the Pakistan edition. Cambiar a la edición global →Cambiar a la edición global →Switch to the global edition →
Product Release

Vertex AI rerank and multi-query input added to the retrieval pipeline

The retrieval pipeline now accepts a list of query texts in a single findRelevant call and aggregates results across all queries, and Vertex AI is now supported as a rerank model provider.

Two additions improve the retrieval pipeline in this release: a new input form for findRelevant and an additional rerank provider.

Multi-query retrieval

  • Array input for findRelevant. A list of query texts can now be passed in a single call. The pipeline executes all queries against the embedding store and returns a merged result set, removing the need to call findRelevant once per question and assemble results manually.
  • Null and empty safety. The call validates the input array before issuing any queries, so an empty list or a null entry does not produce a partial or misleading result.

Vertex AI rerank

  • Additional rerank provider. Vertex AI can now be configured as a rerank model, joining the existing supported services. Reranking improves retrieval precision by scoring candidate chunks against the query before they reach the language model.
  • Consistent configuration surface. Vertex AI rerank is configured through the same model-registry pattern as other AI providers — no new configuration format.

Both changes are additive and backward-compatible. Existing retrieval configurations and single-query calls continue to work without modification.

See the feature →

← All posts