aiWARE OS 3.77.3.75 release notes

New Viewers

Available for a while now in beta mode, today is the official first release of the new streamlined viewers. 

We set out to increase the amount of space available for viewing data, improve the consistency of data display across categories, and allow for contextual editing in place. We will continue to improve and iterate on this release, but for now these are the primary changes to expect:

• Single tab - no more cognition-specific tabs, just a single “timeline” tab

• Engine filter menu - Replacing the larger filter “pills”, the new more compact menu allows for simple toggling of engines by category and individually

• Engine filter memory - Selected engine filters should (mostly) be remembered across different page loads, so you don’t need to adjust the filters the same way every time you load the page if you have a preference

• Multiple engines per category - Now supporting display of multiple engines per category

• New confidence filter menu - Easily adjust the confidence range across all your categories simultaneously, with both upper and lower limits able to be specified

• Options Menu for auto-scroll, bounding boxes, and show confidence level - Placing all three of these options in a centralized menu so they’re always available but less obtrusive to the average user

• Always-open search - Search bar always starts open, making it easier to jump right in to finding something specific

• In-context cognition editing - Easily trigger edit mode for cognitive categories and make inline changes

  • Note: Currently supported for all cognitive categories that were previously supported in viewers

• Editable speaker separation/diarization labels - Similar functionality that was not previously available in viewers, supports both text entry and insertion of existing labels from the current asset

• Unified facial results - Replaced previous large sections of tiny thumbnails with much simpler pill infrastructure that simplifies complex results. All facial detections are now presented in one area, with detections listed as a separate pill, rather than completely separate from recognitions

• Removed for now:

  • Full transcript edit, including text replace functionality

  • Transcript proofreading controls

  • Consolidated view of all occurrences of a single face/object entity

  • Show raw JSON toggle

New Package UI

A new streamlined package management UI is rolling out. This will be the new consistent UI for all places where the user interacts with packages.

• All tabs removed, now you can see all packages your org has access to, regardless of status, in one list

• Easy toggle switch to enable or disable a package

• New filter and sort options

• Search now works on package ID in addition to package name

aiWare UI Updates

• Fixed organization invitations that could not be completed when the previous invite had expired.

• Fixed the UI display for celebrity face recognition results.

• Improved Resource Center loading by removing redundant read dispatches.

aiWare API Updates

• Fixed cloning a TDO so cloned assets include correct sourceData from transformed metadata and prevented numberOfRecordings from doubling.

• Allowed unknown file types to be uploaded through createTDOWithAsset where permitted by policy.

• Hardened Processing Center fetchJobDetails so optional GraphQL fields prune correctly even when introspection is unavailable, avoiding hangs in job details.

• Stopped core-admin from calling getDefaultOrgAdmin for org-less tokens where the call was inappropriate.

• Removed the deprecated internalLogin mutation and the associated INTERNAL_LOGINBYEMAIL permission; use supported authentication flows instead.

• Improved TypeScript typings across multiple aiWare apps packages to reduce unsafe any usage.

• Various security improvements.

aiWare Processing Updates

• Updated core-eventing cognition completed handling so processing deliverables complete when engines use uploadEngineResult instead of createAsset.

• Implemented bucketScannerTask in aidata discovery to run S3 scanning jobs from the scheduler with coordination via media source state and locking.