A new release is available for Arkindex instances. You can test it on our demo instance: demo.arkindex.org.

You can learn more about Arkindex on its official documentation.


The Arkindex frontend has been migrated from Vue 2 to Vue 3, and now uses vue-router 4 and Vuex 4. This migration allowed us to update a number of dependencies, and to support better tools, such as TypeScript .

The process graph in the process status view is temporary disabled while we build a better one. The workers and chunks that make up a process are however still listed on the left.


It is now possible to create a Machine Learning Training process directly on Arkindex (using annotated images from Arkindex) via the API. The new endpoints are documented in the API documentation.

The next release will make this training feature accessible from the frontend.

File import

The PDF and image imports have been merged, which means that it is now possible to import PDF and image files together in a single file import.

PDF and image files no longer need to be uploaded separately

The build entities option, an obsolete feature that had been introduced for a specific project, has been removed from Transkribus imports.


  • It is now possible to filter the results of ListTranscriptionEntities by the Transcription Entities's worker_version, as well as by the worker_version that created the Entities they reference.
  • An unused element_count field has been removed from all API endpoints that retrieve WorkerVersions.
  • The ListElementEntities endpoint has been deprecated.

Misc cleanup & deprecations

  • Arkindex now supports PostgreSQL 14.
  • Arkindex no longer uses Memcached but Redis for cache.
  • Outdated artifacts have been deleted.
  • The corpus thumbnail feature, which was part of a larger feature that was never implemented, has been removed.


  • A stale read has been fixed on the check_parents signal of WorkerRuns, and some unnecesseary checks have been removed.
  • The corpus_delete task now has retries.