Predictions from imported tasks cannot be deleted

[Bjoern-Rapp]

Describe the bug
When pre-annotationed tasks are imported to Label Studio, via the cloud storage interface, model_version is not a required field on the JSON files. This make it possible to create predication with no model version information attached, and this then causes a error, if you try to delete those predictions.

To Reproduce
Steps to reproduce the behavior:

1. Import pre-annotationed tasks from a JSON file without a model_version field
2. Go to Predictions tab in the project setting
3. Try to delete the predictions

A runtime error is shown:

Validation error
model_version param is required
Version: 1.23.0

Expected behavior
Either it should be possible to delete predictions without a model version, or it should be impossible to import tasks without a model_version field.

Screenshots

Environment:

• OS: Ubuntu 22.04.5 LTS
• Label Studio Version 1.23

[Snehadas2005]

Hi @Bjoern-Rapp, I would like to work on this issue. I am using Ubuntu and can reproduce the validation error when trying to delete predictions without a model_version.
My approach will be to update the deletion logic/validator to ensure model_version is treated as optional during the delete process, allowing imported tasks without versions to be cleared.

[Snehadas2005]

Hi @Bjoern-Rapp, I have raised a PR. Could you please review #9735 and let me know if any changes are needed?

[heidi-humansignal]

Hello,

Thank you for contacting Label Studio,

Our engineering team will review the Pull Request and will update it accordingly. Thank you for your contribution.

Best Regards!

Comment by Oussama Assili
Workflow Run