...
Only the source names of the items that will be rerun, and not all the sources from which items were considered, are now displayed in the rerun message.
Fixed language reset on the items when rerunning a step of a pipeline.
Items from Rerun from Index will contain all their facets in the keywords field, irrespective of their data type. Before they were found in kw_<data_type> keys, which required the pipelet (or any other step) to have logic to differentiate between items that came from a normal ingestion or a rerun.
Fixed an issue with the Item detail side panel loading indefinitely.
Now, the ocr_confidence facet is a float type (instead of a string type) when the pdf-ocr pipeline step is created or updated and the confidence option equals True. This corrects a validation error that previously appeared due ocr_confidence presenting as a string type.
Now, term-sequence rescoring on index-scrolling isn’t addedis no longer added. This avoids potentially invalid Elasticsearch query clauses.
Fixed an issue where dashboard searches weren’t cleared when the user switch dashboards if it was applied by the persistentLinks feature.
Removed unnecessary POST parameters in the bulk labeling GET request that would lead
to the frontend API breaking.Added an empty state image to the Community List widget.
Fixed an issue with the carousel indicator not showing up in global search.
...