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.
Confidence score
This release brings several improvements relative to the confidence score associated with the Machine Learning results created on Arkindex.
Elements
Arkindex elements now support a confidence score directly set on the element (instead of relying on metadata).
Several endpoints now support this score:
CreateElement
andCreateElements
when creating new elements,ListElements
,ListElementChildren
andListElementParents
when listing existing elements.
Transcriptions
The elements transcriptions already supported a confidence score, but it was not displayed cleanly on the details of an element.
Classification
We have streamlined the confidence display to use a shared visual component amongst all uses (element, transcription and classification).
CLI Documentation
We have worked extensively on documentation for our official Command Line Interface, and published a new website cli.arkindex.org that describes all the available commands through the open-source project arkindex-cli.
This project extends Arkindex features and is really useful for specific data import or exports.
Gitlab Authentication
We have updated our internal Gitlab token management to support expired OAuth token as their expiry is enforced since Gitlab version 15.0.
If you have trouble with your own git repositories in Arkindex, head to the OAuth credentials list (under /user/profile
of your Arkindex instance, then OAuth tab). If you see your Gitlab credentials marked as erroneous, click on Retry.
Elements
The user interface of Arkindex, as well as the ListElements
, ListElementChildren
endpoints now support the random ordering of elements. This can be useful to get a random sample of a list of elements to create training splits.
API Changes
- All users can now use the
CreateTranscriptionEntity
endpoint, which was previously restricted to internal (workers) usage; - It isn't possible anymore to change the worker version ID on a worker result using API endpoints;
- It isn't possible anymore to create duplicate metadata using the
Partial/UpdateMetaData
endpoints.
Worker run
We are also working on extending Machine Learning results API creation endpoints to support the new worker_run_id
. This is a preliminary step to support Machine Learning model training and execution of generated models into Arkindex.
Worker runs are now supported in the following endpoints:
CreateClassification
CreateClassifications
CreateElement
CreateElements
CreateElementTranscriptions
CreateEntity
CreateMetaData
CreateMetaDataBulk
CreateTranscription
CreateTranscriptionEntity
CreateTranscriptions