Arthur Platform 3.14.5

Release notes for Arthur Platform 3.14.5

3.14.5

Enhancements

• Patched vulnerability found in ingestion-service

Bug Fix

• Fixed issue that would cause Kafka Connect Monitor's health check to consistently fail if the was no lag data on existing connectors

3.14.4

Product Enhancements:

• Local importance scores are now sorted by default in descending order.
• Local importance scores now show up to 6 decimals places to improve accuracy of default sorting
• The inference inputs table has increased attribute rows per page from 12 to 50.

Bug Fixes:

• Users are now able to change what_if values and get real time update of local important score as expected
• Fixed an issue where the incorrect interpretation labels were shown under the prediction summary

3.14.3

Enhancements:

• Small updates to deployment files

3.14.2

Enhancements:

• Remediated known critical vulnerabilities in docker images
• Added options to allow for Ceph object storage as a replacement for S3/MinIO

3.14.1

Bug Fix:

• Fixed an issue where null value ingestion caused the UI to show a grey screen when a user tried to expand an inference in the inference deep dive table.

3.14.0

Product Enhancements:

• Infrastructure Upgrades: Arthur now supports Kubernetes versions 1.27 - 1.28.
• Cleaned up the create an alert workflow so that users will only see the relevant options when selecting metrics and attributes based on the alert type they are creating.
• Users can now filter on text attributes in the inference deep dive table more easily by clicking the 3 dots on the top right of the column, selecting filter, and filling out the value to search for in the pre-populated global filter window.
• Users now have the option to select “Show Data Labels” on the performance plot in order to see values labeled for each point on the plot.
• Added an easier way for users to navigate back to the alerts page from the “manage alert rules” page.
• Users will now be notified at the beginning of the edit alert rule workflow if they won’t successfully be able to edit an alert rule due to unresolved alerts.
• Users will no longer see the “generate explanation” button for CV models on an expanded inference if explainability is not enabled.
• Added pagination to the manage alert rules page.
• Adding an additional performance chart does not copy over any configurations applied to the original chart. Updating an additional chart does not overwrite the configurations of the original chart.
• Changed behavior of the global filters time range so that when users select “clear all” filters, this sets the date range to show from all time.

Bug Fixes:

• Multi-org linking now works correctly with SSO when users are not logged in before clicking a link.
• Page state now updates as expected when navigating from the insights “view inferences” window to the inference deep dive table with filters cleared.
• Fixed a query issue that was causing a delay in loading time for the alerts page.
• Fixed an issue with tooltip information for batch models on the performance plot.
• Triggered alerts now show up instantly on the “New Alerts Over Time” plot.
• Fixed an issue that caused attributes of binary classifier models to change properties.
• Feature Importance chart now works as expected.
• Sending alert email notifications from settings now works as expected.
• Removed the option to filter on image as an attribute in global filters.
• Fixed formatting issues with y-axis label on the drift plot.
• Fixed an issue where transitioning alerts between statuses with a large number of alerts was not working consistently.
• Sorting alert rules now works as expected.
• Filtering the schema in the UI now works as expected.
• Clicking on view alert after creating an alert rule now takes the user back to the manage alert rules page
• Disabled the ability to select “Show as a Bar Chart” with multiclass metrics on the performance plot given this view is not supported.
• Fixed a discrepancy between text box year and calendar year in the global filters time range selector.