Collect errors from automatic suggestion backends

[nijel]

Describe the problem

Currently, automatic suggestion backends silent themselves for errors which are most likely caused by rate limiting:

weblate/weblate/machinery/base.py

Lines 246 to 259 in 8b514cb

	def is_rate_limit_error(self, exc):
	if isinstance(exc, MachineryRateLimit):
	return True
	if not isinstance(exc, HTTPError):
	return False
	# Apply rate limiting for following status codes:
	# HTTP 456 Client Error: Quota Exceeded (DeepL)
	# HTTP 429 Too Many Requests
	# HTTP 401 Unauthorized
	# HTTP 403 Forbidden
	# HTTP 503 Service Unavailable
	if exc.response.status_code in (456, 429, 401, 403, 503):
	return True
	return False

This is nice for user viewed suggestions as it doesn't show an error message on each page.

But this makes such errors invisible to admins.

Describe the solution you'd like

Collect such errors and display them in the management interface.

Describe alternatives you've considered

No response

Screenshots

No response

Additional context

No response

[github-actions]

This issue seems to be a good fit for newbie contributors. You are welcome to contribute to Weblate! Don't hesitate to ask any questions you would have while implementing this.

You can learn about how to get started in our contributors documentation.

[AbdullahHabib-github]

I've over 3 years of experience in Python development and would love to contribute. @nijel Can you please assign this issue to me?

[nijel]

@AbdullahHabib-github you're welcome!

[AbdullahHabib-github]

@nijel should I create a new view or am I supposed to integrate it in any of the existing views?

[nijel]

There is already a view for managing automatic suggestions in https://github.com/WeblateOrg/weblate/blob/main/weblate/machinery/views.py, I think it should display such errors next to the configuration.

Something similar can be seen on GitHub webhooks - one tab is settings, another logs:

[nishitiwari22]

Is this issue still open? I am looking to contribute to it.

[nijel]

Yes, this is open. There is a similar change happening for add-ons in #10053 , it should at least share the UI.

[nishitiwari22]

@nijel

Okay, so I am trying to understand the particular Python file and I came up with a temporary code:

To resolve the issue of rate-limiting errors being silenced for admins, I made changes in the is_rate_limit_error method of the MachineTranslation class in the weblate/weblate/machinery/base.py file.

def is_rate_limit_error(self, exc):
if isinstance(exc, MachineryRateLimit):
return True
if not isinstance(exc, HTTPError):
return False
# Apply rate limiting for following status codes:
# HTTP 456 Client Error: Quota Exceeded (DeepL)
# HTTP 429 Too Many Requests
# HTTP 401 Unauthorized
# HTTP 403 Forbidden
# HTTP 503 Service Unavailable
if exc.response.status_code in (456, 429, 401, 403, 503):
return True
return False

I am trying to locally run on my machine as well. I'll add more as a move forward if the maintainer allows. Actually, this is my first PR and I want to take some time making it a valuable one. Therefore before committing any change I thought to let the maintainer know about it.

[nijel]

The implementation should be:

• Add model to store the errors to weblate/machinery/models.py, it needs include engine name, nullable foreign key to Project (for project scoped installations), timestamp and field to store the error

• Extend MachineTranslation.report_error method to take exception paramater and save the error to the database:

  weblate/weblate/machinery/base.py

  Lines 203 to 206 in 8b514cb

  	def report_error(self, message):
  	"""Wrapper for handling error situations."""
  	report_error(cause="Machinery error")
  	LOGGER.error(message, self.name)

• Add UI to list these errors

• Add a background task to prune these errors after some time (30 days?)