Product DocumentationAPI and Python SDK ReferenceRelease Notes
Schedule a Demo
Schedule a Demo

Arthur Platform 3.13.3

Release notes for Arthur Platform 3.13.3

3.13.1 (Post-Release)

3.13.1 Patch Release Notes

Bug Fix:

  • Sorting inferences in the UI on models with inferences that shared the same partner_inference_id caused the UI to show duplicated data and broken boxes. This has been remedied in the 3.13.1 Patch.

3.13.0 Release Notes

New Feature:

  • Multi-Org Linking: Users will now have the ability to click a model link and be directed to view a model in a different organization than the one they were signed into given they have the necessary permissions.

Product Enhancements:

  • Added specific permissions related error messages to reduce confusion for users when they attempt to perform an action they don’t have permissions for.
  • Users will now be able to see the option to create alerts on Model Output Metrics in the UI
  • Users now have the ability to select a subset of their predicted classes to be shown on the performance chart
  • Changed the behavior of the time selector in global filters so that when a user clicks on the start or end field and selects a new date, it only affects the time field selected.
  • The alert acknowledge and resolve events now appear in the Arthur Audit log
  • When explainability is not enabled for a model, users no longer see the option to “Select Prediction” on an expanded inference in the inference deep dive, because this action doesn’t work without explainability.
  • In global filters, we’ve added specific error messages for when a user tries to enter an unsupported value type for a given attribute
  • When filtering by attribute in global filters, the condition drop down will now only show the options that are relevant to the attribute type selected

Bug Fixes:

  • Fixed an issue that prevented the app from showing the empty state for the alerts page by improving handling of HTTP requests
  • Fixed a number of small issues with the insights workflow
  • Users will now get an error message indicating a failure when attempting to ingest inferences with bounding-boxes that contain negative coordinates.
  • Users will now see a more informative 400 error message when attempting to send data with null values for models with unstructured text.