Releases: marqo-ai/marqo
Releases · marqo-ai/marqo
Release 2.24.8
Release 2.24.7
2.24.7
Bug Fixes and Minor Changes
- Fix an issue where the performance degrades wwhen
collapsingFieldsis used withattributesToRetrieve(#1334).
Release 2.24.6
Release 2.24.5
2.24.5
Bug Fixes and Minor Changes
- Fix an issue where
_relevantCandidatescould change unexpectedly when using relevance cut-off in combination with the collapse field feature (#1332).
Release 2.24.4
Release 2.24.3
2.24.3
Bug Fixes and Minor Changes
- Revert the pagination fix for search using the disjunction retrieval method(#1310).
Release 2.24.2
Release 2.24.1
Release 2.24.0
2.24.0
New features
- Support typeahead search(#1289). You can now get intelligent query suggestions based on partial user input to help users find relevant content faster and reduce zero-result searches. For details, please refer to this API document and this recipe. This is supported only for indexes created with Marqo 2.23.0 or later.
- Support caching embeddings of base64-encoded images(#1285). This feature greatly improves performance by cutting the vectorisation time of commonly reused images from ~10 ms down to ~1 ms.
Bug fixes and minor changes
- Fix an issue where facet counts were incorrect when using the lexical retrieval method with filters and multiple
ORphrases (#1282). - Fix an issue where collasing search stops working when the collapse field is not included in
attributesToRetrievesearch parameter (#1294). - Fix metric emission for the
get-batchanddelete-batchendpoints (#1300). - Ensure access log level now inherits from the root log level (#1301).
- Reduce verbosity of some logs emitted from
HybridSearcher(#1299).
Release 2.23.2
2.23.2
Bug Fixes and Minor Changes
- Fix
totalHitscount for collapsed searches(#1292). When a collapse field is provided in search parameters, thetotalHitsnow reflects the number of unique values for that field (i.e., the number of groups), rather than the raw number of matching documents.