DeepTeam is a framework to red team LLMs and AI agents.
import pytest
from deepteam.vulnerabilities import Bias
from deepteam.vulnerabilities.bias import BiasType
from deepteam.test_case import RTTestCase
class TestBias:
def test_bias_all_types(self):
types = ["gender", "politics", "religion", "race"]
bias = Bias(types=types)
assert sorted(type.value for type in bias.types) == sorted(types)
def test_bias_all_types_default(self):
bias = Bias()
assert sorted(type.value for type in bias.types) == sorted(
type.value for type in BiasType
)
def test_bias_gender(self):
types = ["gender"]
bias = Bias(types=types)
assert sorted(type.value for type in bias.types) == sorted(types)
def test_bias_politics(self):
types = ["politics"]
bias = Bias(types=types)
assert sorted(type.value for type in bias.types) == sorted(types)
def test_bias_religion(self):
types = ["religion"]
bias = Bias(types=types)
assert sorted(type.value for type in bias.types) == sorted(types)
def test_bias_race(self):
types = ["race"]
bias = Bias(types=types)
assert sorted(type.value for type in bias.types) == sorted(types)
def test_bias_all_types_invalid(self):
types = ["gender", "politics", "religion", "race", "invalid"]
with pytest.raises(ValueError):
Bias(types=types)
def test_simulate_attacks_returns_expected_cases(self):
bias = Bias(types=["gender"])
test_cases = bias.simulate_attacks(attacks_per_vulnerability_type=2)
assert len(test_cases) == 2
assert all(isinstance(tc, RTTestCase) for tc in test_cases)
assert all(tc.vulnerability == "Bias" for tc in test_cases)
assert all(
tc.vulnerability_type == BiasType.GENDER for tc in test_cases
)
def test_assess_returns_results(self):
bias = Bias(types=["politics"], async_mode=False)
def dummy_model_callback(prompt):
# Provide a simple pass-through or minimal callback if required by your real env
return prompt
results = bias.assess(
model_callback=dummy_model_callback,
)
assert bias.is_vulnerable() is not None
assert bias.simulated_attacks is not None and isinstance(
bias.simulated_attacks, dict
)
assert bias.res is not None and isinstance(bias.res, dict)
assert BiasType.POLITICS in results
assert len(results[BiasType.POLITICS]) == 1
test_case = results[BiasType.POLITICS][0]
# The score and reason will depend on the real metric evaluation
assert hasattr(test_case, "score")
assert hasattr(test_case, "reason")
def test_get_metric_returns_bias_metric(self):
from deepteam.metrics import BiasMetric
bias = Bias(
async_mode=True, verbose_mode=True, evaluation_model="gpt-4o"
)
metric = bias._get_metric(BiasType.RACE)
assert isinstance(metric, BiasMetric)
assert metric.async_mode is True
assert metric.verbose_mode is True
@pytest.mark.asyncio
async def test_a_assess_returns_async_results(self):
bias = Bias(types=["religion"], async_mode=True)
async def dummy_model_callback(prompt):
return prompt
results = await bias.a_assess(
model_callback=dummy_model_callback,
)
assert bias.is_vulnerable() is not None
assert bias.simulated_attacks is not None and isinstance(
bias.simulated_attacks, dict
)
assert bias.res is not None and isinstance(bias.res, dict)
assert BiasType.RELIGION in results
assert len(results[BiasType.RELIGION]) == 1
test_case = results[BiasType.RELIGION][0]
assert hasattr(test_case, "score")
assert hasattr(test_case, "reason")