aiWare UI Updates
-
Added an autoplay option to the viewers so media starts automatically when opening data detail or clicking a cognition result; autoplay and autoscroll work concurrently.
-
Fixed the persistent engine filter settings in the data detail viewers -- engine on/off choices are now remembered between sessions instead of resetting each time.
-
Improved data detail timeline performance by deferring rendering of off-screen bucket sections, reducing scripting and layout cost on long or complex assets.
-
Added visibility of Pending Node-RED pods in the Container Dashboard, closing an observability gap that previously hid stuck pods during cluster resource exhaustion.
-
Fixed confidence filtering not applying consistently to OCR/Text Recognition and Vehicle/LPR engine results in the viewers.
-
Fixed the Package Detail panel in Resource Center capping the grant list at 30 entries when more grants exist.
aiWare API Updates
-
Fixed a regression in the media-streamer where passing a mediaId parameter incorrectly filtered the response to only that TDO instead of using it as a starting anchor when combined with time ranges.
-
Fixed an issue in updateEntityIdentifierType where the data access layer returned an array but the resolver read a scalar, causing the update to fail silently.
-
Fixed S3 ingest not populating the fileData file size field on TDO assets; assets created via the ingestor-service now correctly report their file size.
-
Fixed watchlist mention-generation failing silently, which prevented watchlists from refreshing for affected organizations.
-
Fixed deleteTDO on Azure throwing an error when an asset with a signed URI was stored in the database.
-
Fixed intermittent WebSocket errors and 502 responses from the tasks GraphQL query in high-frequency polling scenarios.
-
Fixed the OIDC vanity domain redirect flow breaking when the OpenID context was already in an authenticated state.
-
Improved virtual asset download decision logic in the media-streamer to correctly handle edge cases identified post-release.
-
Various security improvements.
aiWare Processing Updates
-
Fixed Engine Toolkit not terminating when an engine running in an Automate flow reported failure via heartbeat, leaving the process webhook handler and chunk pinger blocked indefinitely.
-
Updated edge-controller to include Pending pods in the active pod count check before launching new Node-RED containers, preventing unbounded pod creation when cluster resources are exhausted.
-
Fixed engine-agent assigning pods to SchedulingDisabled (cordoned) nodes, causing pods to remain stuck in a Pending state even after the node became ready.
-
Fixed ai package import writing the source registry URL into the dockerImage field on imported engines instead of the configured target registry.
-
Improved ai package import to skip image extraction when the image already exists in the target registry, reducing import time for partially-migrated environments.
-
Updated engine-agent build inactivation conditions to handle manual-scaling VMSS configurations where cluster autoscaler is not enabled.
-
Added support for configuring the browser Content-Security-Policy header through aiWARE helm chart values, allowing operators to customize CSP without rebuilding frontend images.
-
Updated the aiWARE processing-only package for air-gapped environments with recent platform fixes.
