2.26.0 New features Custom score rerankers in global score modifiers for hybrid search (#1370). You can now rerank hybrid search results using BM25 or vector closeness scores (per-field or aggregate via sum/max/avg) by specifying marqo__score_* fields in scoreModifiers. Scores/aggregates are divide-by-max normalized. This is supported only for semi-structured indexes created with Marqo 2.26.0 or l…
2.25.3 Bug Fixes and Minor Changes Adjust minSortCandidates to offset + limit instead of raising an error when the user-provided value is too small. (#1386)
2.25.2 Bug Fixes and Minor Changes Add configurable MARQO_SEARCH_RANDOM_CONNECTION_CLOSE_RATE environment variable to randomly send Connection: close headers on search queries, forcing periodic connection recycling to Vespa to mitigate imbalanced connection issues in long-running deployments (#1383)
2.25.1 New Features Add center and applyToSubqueries parameters to recency scoring (#1376) center — A fixed Unix epoch timestamp (seconds) to use as the reference point instead of now(), enabling reproducible recency scores across queries applyToSubqueries — Control which hybrid subqueries ("tensor", "lexical", or both) receive recency boosting in RRF hybrid search
2.25.0 Changes Triton-based Inference Architecture Marqo's inference layer has been restructured from a monolithic design into three dedicated components: Inference Orchestrator — A FastAPI service that coordinates inference requests. (#1315) Model Management Container — A FastAPI service for managing ML model lifecycles with Triton Inference Server, including model loading/unloading, health check…
2.24.14 Bug Fixes and Minor Changes Add Vespa convergence checks in add_documents flow to avoid adding partial documents (#1360)
2.24.15 Bug Fixes and Minor Changes Back-port update_index_settings API to 2.24 release branch to support modifying modelProperties of an existing index (#1369)
2.24.13 Bug Fixes and Minor Changes Use orjson in get_document(s) endpoints (#1349) Support picking the representative document within each collapsed group based on a numeric field sorting result (#1350)
2.24.12 Bug Fixes and Minor Changes Add support for the dual-stack endpoint in S3 model downloads (#1345)
2.24.11 Bug Fixes and Minor Changes Support weankAnd lexical retrieval and second-phase score modifiers (#1344).