Processes
Two new Model details and Model Version details views are available, at /model/$model-id
and /model-version/$model-version-id
respectively. The model version view is also accessible from the Worker Run details modal.
Workers
- An invalid worker configuration now stops the build for its worker version, not allowing the creation of an invalid/unusable worker version.
- The worker versions in the worker results deletion modal are now sorted alphabetically.
Worker configurations
- It is now possible to edit an existing (non-archived) worker configuration's name.
- In order to fix a display bug when using list of booleans worker configuration parameters, booleans in the worker configuration form are now displayed as checkboxes instead of round toggles.
Worker Activities
- The average minimum, maximum, average and medium processing time per element is now displayed in the worker activities.
- Running a process on elements that have more than one parent no longer generates duplicated worker activities.
Search
- It is now possible to filter Search results by source (Elements, Transcriptions, Entities or Metadata).
- Switching between projects using the select no longer triggers a search on the previously selected project.
- It is now possible to re-run the same search by clicking the Search button again.
Training
- Two new API endpoints have been created, to allow the logging of metrics during a Machine Learning training process: CreateMetricValue and CreateMetricValues.
- It is not longer possible to create a training process using a worker version that is not of state
available
.
Metadata
- It is now possible to save/create a metadata byt hitting the
Enter
key. - Detailed errors are now displayed within the metadata creation form.
- A cog icon now allows you to see the Worker Run details on a MetaData.
Entities
- It is now possible to filter entities by Worker Run, and the Worker Run summary is displayed and clickable to access the Worker Run details.
- The Worker Run summary and details are also accessible from the Entity details page.
API
- The bulk classification creation endpoint
CreateClassifications
(API doc) now uses the ML Class UUIDs instead of their name. - A new
RetrieveModel
endpoint has been added. - The Worker Run object returned by endpoints such as
ListWorkerRun
orCreateWorkerRun
no longer contains the separate IDs of the objects it contains (such as aworker_run_id
when it also containsworker_run
which already has the ID). - The
CreateElements
andCreateMetaDataBulk
endpoints now require either a Worker Version ID or a Worker Run ID to create objects, as these endpoints are designed to be used by Machine Learning Workers, not humans.
Housekeeping
A bug which prevented some expired workflows from being deleted was corrected, allowing us to free up a lot of storage space.