Testing

Testing applications that use LLMs can be tricky and expensive if you hit the real API every time. django-ai-validator provides tools to make testing easier.

Using the Mock Provider

The easiest way to test is to use the mock provider. This provider returns deterministic responses without making network requests.

Configuration

In your tests.py or conftest.py, register the mock provider:

from django.test import TestCase, override_settings
from django_ai_validator.llm.factory import LLMFactory
from django_ai_validator.llm.mock_factory import MockFactory

# Register the mock factory (do this once, e.g., in AppConfig.ready or setUpClass)
LLMFactory.register('mock', MockFactory)

@override_settings(AI_CLEANER_DEFAULT_PROVIDER='mock')
class MyModelTests(TestCase):
    def test_validation(self):
        # This will use the MockAdapter
        ...

Mock Adapter Behavior

The default MockAdapter: - Validation: Returns False if the input contains "bad", otherwise True. - Cleaning: Replaces "dirty" with "clean".

You can customize this behavior by creating your own mock adapter and factory if needed.

Mocking with unittest.mock

For more granular control, you can mock the AICleaningFacade or the underlying LLM client.

from unittest.mock import patch
from django.test import TestCase
from .models import MyModel

class MyTests(TestCase):
    @patch('django_ai_validator.validators.AICleaningFacade.validate')
    def test_custom_validation(self, mock_validate):
        mock_validate.return_value = (True, None)

        # Your test code here
        ...

Testing Asynchronous Tasks

If you are using use_async=True, you can test the task execution by using celery.contrib.testing or by mocking the task.

from unittest.mock import patch

@patch('django_ai_validator.tasks.ai_clean_model_instance.delay')
def test_async_trigger(self, mock_delay):
    # Save model
    instance.save()

    # Verify task was called
    mock_delay.assert_called_once()