KeyError: 'annotation' in fit() when handling /webhook events without annotation payload

[xiaoyao9184]

When receiving a /webhook POST from Label Studio, the ML backend crashes with a KeyError: 'annotation' in examples/bert_classifier/model.py, specifically inside the fit() method:

[2025-05-14 10:33:06,753] [WARNING] [werkzeug::_log::97]  * Debugger is active!
[2025-05-14 10:33:06,754] [INFO] [werkzeug::_log::97]  * Debugger PIN: 485-577-215
[2025-05-14 10:33:18,338] [INFO] [werkzeug::_log::97] 172.16.22.2 - - [14/May/2025 10:33:18] "POST /webhook HTTP/1.1" 500 -
Traceback (most recent call last):
  File "/home/ubuntu/.conda/envs/label-studio/lib/python3.10/site-packages/flask/app.py", line 2213, in __call__
    return self.wsgi_app(environ, start_response)
  File "/home/ubuntu/.conda/envs/label-studio/lib/python3.10/site-packages/flask/app.py", line 2193, in wsgi_app
    response = self.handle_exception(e)
  File "/home/ubuntu/.conda/envs/label-studio/lib/python3.10/site-packages/flask/app.py", line 2190, in wsgi_app
    response = self.full_dispatch_request()
  File "/home/ubuntu/.conda/envs/label-studio/lib/python3.10/site-packages/flask/app.py", line 1486, in full_dispatch_request
    rv = self.handle_user_exception(e)
  File "/home/ubuntu/.conda/envs/label-studio/lib/python3.10/site-packages/flask/app.py", line 1484, in full_dispatch_request
    rv = self.dispatch_request()
  File "/home/ubuntu/.conda/envs/label-studio/lib/python3.10/site-packages/flask/app.py", line 1469, in dispatch_request
    return self.ensure_sync(self.view_functions[rule.endpoint])(**view_args)
  File "/home/ubuntu/.conda/envs/label-studio/lib/python3.10/site-packages/label_studio_ml/api.py", line 126, in webhook
    result = model.fit(event, data)
  File "/home/ubuntu/code/label-studio-ml-backend/label_studio_ml/examples/bert_classifier/model.py", line 125, in fit
    project_id = data['annotation']['project']
KeyError: 'annotation'

Looking at this line

https://github.com/HumanSignal/label-studio-ml-backend/blob/2ba5455cc45377d06f8216ee7d265d95c5222e3a/label_studio_ml/examples/bert_classifier/model.py#L125-L125

However, according to the surrounding code, project_id has already been provided as a constructor argument to the model.

https://github.com/HumanSignal/label-studio-ml-backend/blob/2ba5455cc45377d06f8216ee7d265d95c5222e3a/label_studio_ml/api.py#L117-L125)

I've forked the repository and updated the following example models to use self.project_id instead:

• bert_classifier
• huggingface_ner
• sklearn_text_classifier

However, I have not yet run the full test suite or studied the full contribution guidelines, as I’m not deeply familiar with this codebase yet.

Would it be acceptable to open a PR with this minimal fix and receive further guidance from maintainers? Or would you prefer I do more in-depth validation first?