We are happy to announce that a new Arkindex release is available. You can explore Arkindex and try out the newest features on our demo instance, demo.arkindex.org.

Technical release notes for developers and instance administrators are available here.

Worker configuration fields

Our worker user configuration system outgrew its intended uses and became hard to maintain and improve upon. In this release, we broke ground on the first step towards a complete reimplementation by introducing a new concept of worker configuration fields and a new file format for .arkindex.yml.

Worker configuration fields

There is currently no way to easily create workers with this new format, or create configuration fields or use them. Most of the complexity of worker configurations comes from the frontend, so we will focus on making them work there over multiple releases before opening them up for creation to avoid confusion.

Thumbnail generation on file imports

In Arkindex 1.6.0, we had removed the thumbnail generation after each file import to simplify processes overall. This feature has been restored using a new system worker as it became simpler to support with the subsequent refactorings we made. When starting or retrying a file import, a new thumbnail generation task will now be added and thumbnails will be regenerated for any folders that were created or updated by the file import.

Hide restarted tasks in processes

In large processes, it can be common to restart a few specific tasks instead of retrying the whole process. This can make the graph more confusing as it does not reflect the latest state of the process instead. To allow to both view the full history of a process and its latest state, a new Show restarted tasks checkbox is now available. When unchecked, tasks that have been restarted will be hidden to simplify the graph.

Toggle restarted tasks in process status view

Misc

This release comes with many bug fixes and smaller improvements.

  • When importing PDF documents that contain text, unsupported Unicode characters are now removed before sending the text to Arkindex as transcriptions.
  • Deleting a corpus no longer fails if it has processes associated with any budget entries.
  • A significant performance improvement was made on the Select all failed elements button of the process worker activities view for Arkindex instances with a large number of worker activities, regardless of how many elements would have been selected.
  • In the process worker activities view, the title of the first worker version is now at the same width as the other worker versions.
  • Display names are now used instead of slugs when displaying worker types.
  • Worker execution costs have been moved to a separate tab in the worker details view and the worker version selection modal.
  • Filters in the filter bar are now shown on multiple lines again when they overflow the page.
  • Viewing any ML result with a worker run assigned to it while logged out no longer causes the frontend no longer makes API calls that always result in a HTTP 403 error.
  • When trying to add a new classification to an element, typing a very long class name that contains more than 10 words separated by spaces or commas no longer causes "Request failed with status code 400" notifications.
  • When an email verification fails, the frontend no longer shows both a success and an error notification.
  • When editing a manual transcription, the textarea HTML input is vertically resizable.