from deepeval.metrics import FaithfulnessMetric
from deepeval.test_case import TestCase

example = LLMTestCase(
    intput="What is the refund policy?",
    actual_output="You can return items within 30 days.",
    retrieval_context=["Return policy 20 days for AMER", "Return policy 30 days for EMEA"],
    )

# Current behaviour
metric = FaithfulnessMetric(penalize_ambiguous_claims=True)
metric.measure(example)
>> 0.0
metric.reason
>> Awesome job! There are no contradictions ...

# New behaviour
metric = FaithfulnessMetric(penalize_ambiguous_claims=True)
metric.measure(example)
>> 0.0
metric.reason
>> The score is 0.00 because the actual output incorrectly claims ...