chore: Fix evaluation test cases to only use pytest features

PiperOrigin-RevId: 820700378

tests/unittests/evaluation/test_eval_config.py

@@ -14,8 +14,6 @@
 
 from __future__ import annotations
 
-from unittest import mock
-
 from google.adk.evaluation.eval_config import _DEFAULT_EVAL_CONFIG
 from google.adk.evaluation.eval_config import EvalConfig
 from google.adk.evaluation.eval_config import get_eval_metrics_from_config
@@ -28,13 +26,14 @@ def test_get_evaluation_criteria_or_default_returns_default():
   assert get_evaluation_criteria_or_default("") == _DEFAULT_EVAL_CONFIG
 
 
-def test_get_evaluation_criteria_or_default_reads_from_file():
+def test_get_evaluation_criteria_or_default_reads_from_file(mocker):
   eval_config = EvalConfig(
       criteria={"tool_trajectory_avg_score": 0.5, "response_match_score": 0.5}
   )
-  mock_open = mock.mock_open(read_data=eval_config.model_dump_json())
-  with mock.patch("builtins.open", mock_open):
-    assert get_evaluation_criteria_or_default("dummy_path") == eval_config
+  mocker.patch(
+      "builtins.open", mocker.mock_open(read_data=eval_config.model_dump_json())
+  )
+  assert get_evaluation_criteria_or_default("dummy_path") == eval_config
 
 
 def test_get_eval_metrics_from_config():

tests/unittests/evaluation/test_evaluation_generator.py

@@ -14,8 +14,6 @@
 
 from __future__ import annotations
 
-from unittest import mock
-
 from google.adk.evaluation.app_details import AgentDetails
 from google.adk.evaluation.app_details import AppDetails
 from google.adk.evaluation.evaluation_generator import EvaluationGenerator
@@ -206,17 +204,17 @@ class TestConvertEventsToEvalInvocation:
 class TestGetAppDetailsByInvocationId:
   """Test cases for EvaluationGenerator._get_app_details_by_invocation_id method."""
 
-  def test_get_app_details_by_invocation_id_empty(self):
+  def test_get_app_details_by_invocation_id_empty(self, mocker):
     """Tests with an empty list of events."""
-    mock_request_intercepter = mock.MagicMock(spec=_RequestIntercepterPlugin)
+    mock_request_intercepter = mocker.MagicMock(spec=_RequestIntercepterPlugin)
     app_details = EvaluationGenerator._get_app_details_by_invocation_id(
         [], mock_request_intercepter
     )
     assert app_details == {}
 
-  def test_get_app_details_by_invocation_id_no_model_requests(self):
+  def test_get_app_details_by_invocation_id_no_model_requests(self, mocker):
     """Tests when request_intercepter returns no model requests."""
-    mock_request_intercepter = mock.MagicMock(spec=_RequestIntercepterPlugin)
+    mock_request_intercepter = mocker.MagicMock(spec=_RequestIntercepterPlugin)
     mock_request_intercepter.get_model_request.return_value = None
     events = [
         _build_event("user", [types.Part(text="Hello")], "inv1"),
@@ -230,9 +228,9 @@ class TestGetAppDetailsByInvocationId:
         events[1]
     )
 
-  def test_get_app_details_single_invocation_single_agent(self):
+  def test_get_app_details_single_invocation_single_agent(self, mocker):
     """Tests a single invocation with one agent."""
-    mock_request_intercepter = mock.MagicMock(spec=_RequestIntercepterPlugin)
+    mock_request_intercepter = mocker.MagicMock(spec=_RequestIntercepterPlugin)
     mock_llm_request = LlmRequest(model="test")
     mock_llm_request.config.system_instruction = "instruction1"
     mock_llm_request.config.tools = [types.Tool()]
@@ -262,9 +260,9 @@ class TestGetAppDetailsByInvocationId:
         events[1]
     )
 
-  def test_get_app_details_multiple_invocations_multiple_agents(self):
+  def test_get_app_details_multiple_invocations_multiple_agents(self, mocker):
     """Tests multiple invocations with multiple agents."""
-    mock_request_intercepter = mock.MagicMock(spec=_RequestIntercepterPlugin)
+    mock_request_intercepter = mocker.MagicMock(spec=_RequestIntercepterPlugin)
 
     def get_model_request_side_effect(event):
       mock_llm_request = LlmRequest(model="test")

tests/unittests/evaluation/test_hallucinations_v1.py

@@ -13,9 +13,6 @@
 # limitations under the License.
 
 import json
-from unittest.mock import AsyncMock
-from unittest.mock import MagicMock
-from unittest.mock import patch
 
 from google.adk.evaluation.app_details import AgentDetails
 from google.adk.evaluation.app_details import AppDetails
@@ -34,13 +31,13 @@ import pytest
 
 
 @pytest.fixture
-def mock_llm_registry():
+def mock_llm_registry(mocker):
   """Mocks LLMRegistry to avoid actual model loading during tests."""
-  with patch(
+  MockLLMRegistry = mocker.patch(
       "google.adk.evaluation.hallucinations_v1.LLMRegistry"
-  ) as MockLLMRegistry:
-    MockLLMRegistry.return_value.resolve.return_value = AsyncMock
-    yield
+  )
+  MockLLMRegistry.return_value.resolve.return_value = mocker.MagicMock()
+  yield
 
 
 @pytest.fixture
@@ -176,8 +173,8 @@ contradicting_excerpt: Pears are blue fruits
 class TestEvaluateNlResponse:
   """Test cases for _evaluate_nl_response method."""
 
-  def _create_genai_response(self, text):
-    response_mock = MagicMock()
+  def _create_genai_response(self, text, mocker):
+    response_mock = mocker.MagicMock()
     response_mock.content = genai_types.Content(
         parts=[genai_types.Part(text=text)]
     )
@@ -185,12 +182,12 @@ class TestEvaluateNlResponse:
 
   @pytest.mark.asyncio
   async def test_evaluate_nl_response_unexpected_labels(
-      self, hallucinations_metric
+      self, hallucinations_metric, mocker
   ):
     """Tests _evaluate_nl_response with unexpected labels."""
     metric = hallucinations_metric
     seg_response = self._create_genai_response(
-        "<sentence>sentence 1</sentence><sentence>sentence 2</sentence>"
+        "<sentence>sentence 1</sentence><sentence>sentence 2</sentence>", mocker
     )
     val_response_text = """sentence: sentence 1
 label:
@@ -204,7 +201,7 @@ rationale: r2
 supporting_excerpt: null
 contradicting_excerpt: null
 """
-    val_response = self._create_genai_response(val_response_text)
+    val_response = self._create_genai_response(val_response_text, mocker)
 
     async def seg_gen():
       yield seg_response
@@ -212,7 +209,7 @@ contradicting_excerpt: null
     async def val_gen():
       yield val_response
 
-    metric._judge_model.generate_content_async = MagicMock(
+    metric._judge_model.generate_content_async = mocker.MagicMock(
         side_effect=[
             seg_gen(),
             val_gen(),
@@ -223,14 +220,14 @@ contradicting_excerpt: null
 
   @pytest.mark.asyncio
   async def test_evaluate_nl_response_missing_label(
-      self, hallucinations_metric
+      self, hallucinations_metric, mocker
   ):
     """Tests _evaluate_nl_response with missing labels in validation results."""
     metric = hallucinations_metric
     seg_response = self._create_genai_response(
-        "<sentence>sentence 1</sentence>"
+        "<sentence>sentence 1</sentence>", mocker
     )
-    val_response = self._create_genai_response("val_response")
+    val_response = self._create_genai_response("val_response", mocker)
 
     async def seg_gen():
       yield seg_response
@@ -238,7 +235,7 @@ contradicting_excerpt: null
     async def val_gen():
       yield val_response
 
-    metric._judge_model.generate_content_async = MagicMock(
+    metric._judge_model.generate_content_async = mocker.MagicMock(
         side_effect=[
             seg_gen(),
             val_gen(),
@@ -585,7 +582,7 @@ class TestEvaluateInvocationsAgentTree:
 
   @pytest.mark.asyncio
   async def test_evaluate_invocations_multi_agents(
-      self, hallucinations_metric, agent_tree_data
+      self, hallucinations_metric, agent_tree_data, mocker
   ):
     """Tests evaluate_invocations with agent tree and checks contexts."""
     invocation, expected_invocation = agent_tree_data
@@ -804,22 +801,22 @@ Agent2 response.
         }])
       return None, "error"
 
-    with patch(
+    mocker.patch(
         "google.adk.evaluation.hallucinations_v1.HallucinationsV1Evaluator._evaluate_nl_response",
         side_effect=mock_evaluate_nl_response,
-    ):
-      result = await metric.evaluate_invocations(
-          [invocation], [expected_invocation]
-      )
+    )
+    result = await metric.evaluate_invocations(
+        [invocation], [expected_invocation]
+    )
 
-      assert result.overall_score == pytest.approx(0.5)
-      assert len(result.per_invocation_results) == 1
-      per_invocation_result = result.per_invocation_results[0]
-      assert per_invocation_result.score == pytest.approx(0.5)
+    assert result.overall_score == pytest.approx(0.5)
+    assert len(result.per_invocation_results) == 1
+    per_invocation_result = result.per_invocation_results[0]
+    assert per_invocation_result.score == pytest.approx(0.5)
 
   @pytest.mark.asyncio
   async def test_evaluate_invocations_agent_tree_skip_intermediate(
-      self, mock_llm_registry, agent_tree_data
+      self, mock_llm_registry, agent_tree_data, mocker
   ):
     """Tests evaluate_invocations with agent tree skipping intermediate steps."""
     invocation, expected_invocation = agent_tree_data
@@ -927,18 +924,18 @@ Agent2 response.
           "label": "contradictory",
       }])
 
-    with patch(
+    mocker.patch(
         "google.adk.evaluation.hallucinations_v1.HallucinationsV1Evaluator._evaluate_nl_response",
         side_effect=mock_evaluate_nl_response,
-    ):
-      result = await metric.evaluate_invocations(
-          [invocation], [expected_invocation]
-      )
+    )
+    result = await metric.evaluate_invocations(
+        [invocation], [expected_invocation]
+    )
 
-      assert result.overall_score == 0.0
-      assert len(result.per_invocation_results) == 1
-      per_invocation_result = result.per_invocation_results[0]
-      assert per_invocation_result.score == 0.0
+    assert result.overall_score == 0.0
+    assert len(result.per_invocation_results) == 1
+    per_invocation_result = result.per_invocation_results[0]
+    assert per_invocation_result.score == 0.0
 
 
 @pytest.fixture
@@ -1053,7 +1050,7 @@ class TestEvaluateInvocationsTimeWeather:
 
   @pytest.mark.asyncio
   async def test_evaluate_invocations_time_weather(
-      self, hallucinations_metric, time_weather_data
+      self, hallucinations_metric, time_weather_data, mocker
   ):
     """Tests evaluate_invocations with time/weather agent."""
     invocation, response1, response2 = time_weather_data
@@ -1190,20 +1187,20 @@ tool_outputs:
         ])
       return None, "error"
 
-    with patch(
+    mocker.patch(
         "google.adk.evaluation.hallucinations_v1.HallucinationsV1Evaluator._evaluate_nl_response",
         side_effect=mock_evaluate_nl_response,
-    ):
-      result = await metric.evaluate_invocations([invocation], [invocation])
+    )
+    result = await metric.evaluate_invocations([invocation], [invocation])
 
-      assert result.overall_score == pytest.approx(5 / 6)
-      assert len(result.per_invocation_results) == 1
-      per_invocation_result = result.per_invocation_results[0]
-      assert per_invocation_result.score == pytest.approx(5 / 6)
+    assert result.overall_score == pytest.approx(5 / 6)
+    assert len(result.per_invocation_results) == 1
+    per_invocation_result = result.per_invocation_results[0]
+    assert per_invocation_result.score == pytest.approx(5 / 6)
 
   @pytest.mark.asyncio
   async def test_evaluate_invocations_time_weather_skip_intermediate(
-      self, mock_llm_registry, time_weather_data
+      self, mock_llm_registry, time_weather_data, mocker
   ):
     """Tests evaluate_invocations with time/weather agent."""
     invocation, _, response2 = time_weather_data
@@ -1300,20 +1297,20 @@ tool_outputs:
           {"sentence": sentence2, "label": "supported"},
       ])
 
-    with patch(
+    mocker.patch(
         "google.adk.evaluation.hallucinations_v1.HallucinationsV1Evaluator._evaluate_nl_response",
         side_effect=mock_evaluate_nl_response,
-    ):
-      result = await metric.evaluate_invocations([invocation], [invocation])
+    )
+    result = await metric.evaluate_invocations([invocation], [invocation])
 
-      assert result.overall_score == 1.0
-      assert len(result.per_invocation_results) == 1
-      per_invocation_result = result.per_invocation_results[0]
-      assert per_invocation_result.score == 1.0
+    assert result.overall_score == 1.0
+    assert len(result.per_invocation_results) == 1
+    per_invocation_result = result.per_invocation_results[0]
+    assert per_invocation_result.score == 1.0
 
 
 @pytest.mark.asyncio
-async def test_evaluate_invocations_success_path(hallucinations_metric):
+async def test_evaluate_invocations_success_path(hallucinations_metric, mocker):
   metric = hallucinations_metric
   app_details = AppDetails(
       agent_details={
@@ -1380,18 +1377,18 @@ async def test_evaluate_invocations_success_path(hallucinations_metric):
       )
     return None, "error"
 
-  with patch(
+  mocker.patch(
       "google.adk.evaluation.hallucinations_v1.HallucinationsV1Evaluator._evaluate_nl_response",
       side_effect=mock_evaluate_nl_response,
-  ):
-    result = await metric.evaluate_invocations(
-        [actual_invocation], [expected_invocation]
-    )
+  )
+  result = await metric.evaluate_invocations(
+      [actual_invocation], [expected_invocation]
+  )
 
-    assert result.overall_score == pytest.approx(0.5)
-    assert len(result.per_invocation_results) == 1
-    per_invocation_result = result.per_invocation_results[0]
-    assert per_invocation_result.score == pytest.approx(0.5)
+  assert result.overall_score == pytest.approx(0.5)
+  assert len(result.per_invocation_results) == 1
+  per_invocation_result = result.per_invocation_results[0]
+  assert per_invocation_result.score == pytest.approx(0.5)
 
 
 @pytest.mark.asyncio
@@ -1446,7 +1443,9 @@ async def test_evaluate_invocations_no_nl_response(hallucinations_metric):
 
 
 @pytest.mark.asyncio
-async def test_evaluate_all_invocations_not_evaluated(hallucinations_metric):
+async def test_evaluate_all_invocations_not_evaluated(
+    hallucinations_metric, mocker
+):
   metric = hallucinations_metric
   app_details = AppDetails(
       agent_details={
@@ -1490,30 +1489,32 @@ async def test_evaluate_all_invocations_not_evaluated(hallucinations_metric):
   async def mock_evaluate_nl_response(nl_response, context):
     return None, "Judge model error."
 
-  with patch(
+  mocker.patch(
       "google.adk.evaluation.hallucinations_v1.HallucinationsV1Evaluator._evaluate_nl_response",
       side_effect=mock_evaluate_nl_response,
-  ):
-    result = await metric.evaluate_invocations(
-        [actual_invocation, actual_invocation],
-        [expected_invocation, expected_invocation],
-    )
+  )
+  result = await metric.evaluate_invocations(
+      [actual_invocation, actual_invocation],
+      [expected_invocation, expected_invocation],
+  )
 
-    assert len(result.per_invocation_results) == 2
-    assert result.per_invocation_results[0].score is None
-    assert (
-        result.per_invocation_results[0].eval_status == EvalStatus.NOT_EVALUATED
-    )
-    assert result.per_invocation_results[1].score is None
-    assert (
-        result.per_invocation_results[1].eval_status == EvalStatus.NOT_EVALUATED
-    )
-    assert result.overall_score is None
-    assert result.overall_eval_status == EvalStatus.NOT_EVALUATED
+  assert len(result.per_invocation_results) == 2
+  assert result.per_invocation_results[0].score is None
+  assert (
+      result.per_invocation_results[0].eval_status == EvalStatus.NOT_EVALUATED
+  )
+  assert result.per_invocation_results[1].score is None
+  assert (
+      result.per_invocation_results[1].eval_status == EvalStatus.NOT_EVALUATED
+  )
+  assert result.overall_score is None
+  assert result.overall_eval_status == EvalStatus.NOT_EVALUATED
 
 
 @pytest.mark.asyncio
-async def test_evaluate_invocations_partial_failure(hallucinations_metric):
+async def test_evaluate_invocations_partial_failure(
+    hallucinations_metric, mocker
+):
   metric = hallucinations_metric
   app_details = AppDetails(
       agent_details={
@@ -1563,15 +1564,15 @@ async def test_evaluate_invocations_partial_failure(hallucinations_metric):
       return None, "some error during evaluation"
     return None, "error"
 
-  with patch(
+  mocker.patch(
       "google.adk.evaluation.hallucinations_v1.HallucinationsV1Evaluator._evaluate_nl_response",
       side_effect=mock_evaluate_nl_response,
-  ):
-    result = await metric.evaluate_invocations(
-        [actual_invocation], [expected_invocation]
-    )
+  )
+  result = await metric.evaluate_invocations(
+      [actual_invocation], [expected_invocation]
+  )
 
-    assert result.overall_score == 0.8
-    assert len(result.per_invocation_results) == 1
-    per_invocation_result = result.per_invocation_results[0]
-    assert per_invocation_result.score == 0.8
+  assert result.overall_score == 0.8
+  assert len(result.per_invocation_results) == 1
+  per_invocation_result = result.per_invocation_results[0]
+  assert per_invocation_result.score == 0.8

tests/unittests/evaluation/test_llm_as_judge.py

@@ -15,7 +15,6 @@
 from __future__ import annotations
 
 from typing import Optional
-from unittest.mock import MagicMock
 
 from google.adk.evaluation.eval_case import Invocation
 from google.adk.evaluation.eval_metrics import EvalMetric
@@ -128,8 +127,8 @@ def test_llm_as_judge_init_unregistered_model():
 
 
 @pytest.fixture
-def mock_judge_model():
-  mock_judge_model = MagicMock()
+def mock_judge_model(mocker):
+  mock_judge_model = mocker.MagicMock()
 
   async def mock_generate_content_async(llm_request):
     yield LlmResponse(
@@ -144,30 +143,30 @@ def mock_judge_model():
 
 @pytest.mark.asyncio
 async def test_evaluate_invocations_with_mock(
-    mock_llm_as_judge, mock_judge_model
+    mock_llm_as_judge, mock_judge_model, mocker
 ):
   mock_llm_as_judge._judge_model = mock_judge_model
 
-  mock_format_auto_rater_prompt = MagicMock(
+  mock_format_auto_rater_prompt = mocker.MagicMock(
       wraps=mock_llm_as_judge.format_auto_rater_prompt
   )
   mock_llm_as_judge.format_auto_rater_prompt = mock_format_auto_rater_prompt
 
-  mock_convert_auto_rater_response_to_score = MagicMock(
+  mock_convert_auto_rater_response_to_score = mocker.MagicMock(
       wraps=mock_llm_as_judge.convert_auto_rater_response_to_score
   )
   mock_llm_as_judge.convert_auto_rater_response_to_score = (
       mock_convert_auto_rater_response_to_score
   )
 
-  mock_aggregate_per_invocation_samples = MagicMock(
+  mock_aggregate_per_invocation_samples = mocker.MagicMock(
       wraps=mock_llm_as_judge.aggregate_per_invocation_samples
   )
   mock_llm_as_judge.aggregate_per_invocation_samples = (
       mock_aggregate_per_invocation_samples
   )
 
-  mock_aggregate_invocation_results = MagicMock(
+  mock_aggregate_invocation_results = mocker.MagicMock(
       wraps=mock_llm_as_judge.aggregate_invocation_results
   )
   mock_llm_as_judge.aggregate_invocation_results = (

tests/unittests/evaluation/test_local_eval_service.py

@@ -14,7 +14,6 @@
 
 import asyncio
 import sys
-from unittest import mock
 
 from google.adk.agents.llm_agent import LlmAgent
 from google.adk.errors.not_found_error import NotFoundError
@@ -47,8 +46,8 @@ import pytest
 
 
 @pytest.fixture
-def mock_eval_sets_manager():
-  return mock.create_autospec(EvalSetsManager)
+def mock_eval_sets_manager(mocker):
+  return mocker.create_autospec(EvalSetsManager)
 
 
 @pytest.fixture
@@ -58,8 +57,8 @@ def dummy_agent():
 
 
 @pytest.fixture
-def mock_eval_set_results_manager():
-  return mock.create_autospec(EvalSetResultsManager)
+def mock_eval_set_results_manager(mocker):
+  return mocker.create_autospec(EvalSetResultsManager)
 
 
 @pytest.fixture
@@ -118,6 +117,7 @@ async def test_perform_inference_success(
     eval_service,
     dummy_agent,
     mock_eval_sets_manager,
+    mocker,
 ):
   eval_set = EvalSet(
       eval_set_id="test_eval_set",
@@ -128,8 +128,8 @@ async def test_perform_inference_success(
   )
   mock_eval_sets_manager.get_eval_set.return_value = eval_set
 
-  mock_inference_result = mock.MagicMock()
-  eval_service._perform_inference_sigle_eval_item = mock.AsyncMock(
+  mock_inference_result = mocker.MagicMock()
+  eval_service._perform_inference_sigle_eval_item = mocker.AsyncMock(
       return_value=mock_inference_result
   )
 
@@ -157,6 +157,7 @@ async def test_perform_inference_with_case_ids(
     eval_service,
     dummy_agent,
     mock_eval_sets_manager,
+    mocker,
 ):
   eval_set = EvalSet(
       eval_set_id="test_eval_set",
@@ -168,8 +169,8 @@ async def test_perform_inference_with_case_ids(
   )
   mock_eval_sets_manager.get_eval_set.return_value = eval_set
 
-  mock_inference_result = mock.MagicMock()
-  eval_service._perform_inference_sigle_eval_item = mock.AsyncMock(
+  mock_inference_result = mocker.MagicMock()
+  eval_service._perform_inference_sigle_eval_item = mocker.AsyncMock(
       return_value=mock_inference_result
   )
 
@@ -219,7 +220,7 @@ async def test_perform_inference_eval_set_not_found(
 
 @pytest.mark.asyncio
 async def test_evaluate_success(
-    eval_service, mock_eval_sets_manager, mock_eval_set_results_manager
+    eval_service, mock_eval_sets_manager, mock_eval_set_results_manager, mocker
 ):
   inference_results = [
       InferenceResult(
@@ -243,7 +244,7 @@ async def test_evaluate_success(
       evaluate_config=EvaluateConfig(eval_metrics=[eval_metric], parallelism=2),
   )
 
-  mock_eval_case = mock.MagicMock(spec=EvalCase)
+  mock_eval_case = mocker.MagicMock(spec=EvalCase)
   mock_eval_case.conversation = []
   mock_eval_case.session_input = None
   mock_eval_sets_manager.get_eval_case.return_value = mock_eval_case
@@ -290,7 +291,7 @@ async def test_evaluate_eval_case_not_found(
 
 @pytest.mark.asyncio
 async def test_evaluate_single_inference_result(
-    eval_service, mock_eval_sets_manager, mock_eval_set_results_manager
+    eval_service, mock_eval_sets_manager, mock_eval_set_results_manager, mocker
 ):
   invocation = Invocation(
       user_content=genai_types.Content(
@@ -314,7 +315,7 @@ async def test_evaluate_single_inference_result(
   eval_metric = EvalMetric(metric_name="fake_metric", threshold=0.5)
   evaluate_config = EvaluateConfig(eval_metrics=[eval_metric], parallelism=1)
 
-  mock_eval_case = mock.MagicMock(spec=EvalCase)
+  mock_eval_case = mocker.MagicMock(spec=EvalCase)
   mock_eval_case.conversation = [
       invocation.model_copy(deep=True),
       invocation.model_copy(deep=True),
@@ -370,7 +371,7 @@ def test_generate_final_eval_status_doesn_t_throw_on(eval_service):
 @pytest.mark.skipif(
     sys.version_info < (3, 10), reason="MCP tool requires Python 3.10+"
 )
-async def test_mcp_stdio_agent_no_runtime_error():
+async def test_mcp_stdio_agent_no_runtime_error(mocker):
   """Test that LocalEvalService can handle MCP stdio agents without RuntimeError.
 
   This is a regression test for GitHub issue #2196:
@@ -421,7 +422,7 @@ async def test_mcp_stdio_agent_no_runtime_error():
     )
 
     # Create a mock eval sets manager that returns an eval case
-    mock_eval_sets_manager = mock.create_autospec(EvalSetsManager)
+    mock_eval_sets_manager = mocker.create_autospec(EvalSetsManager)
     test_eval_case = EvalCase(
         eval_id="test_mcp_case",
         conversation=[

tests/unittests/evaluation/test_local_eval_set_results_manager.py

@@ -19,7 +19,6 @@ import os
 import shutil
 import tempfile
 import time
-from unittest.mock import patch
 
 from google.adk.errors.not_found_error import NotFoundError
 from google.adk.evaluation._eval_set_results_manager_utils import _sanitize_eval_set_result_name
@@ -68,12 +67,11 @@ class TestLocalEvalSetResultsManager:
         eval_case_results=self.eval_case_results,
         creation_timestamp=self.timestamp,
     )
-
-  def teardown(self):
+    yield
     shutil.rmtree(self.temp_dir)
 
-  @patch("time.time")
-  def test_save_eval_set_result(self, mock_time):
+  def test_save_eval_set_result(self, mocker):
+    mock_time = mocker.patch("time.time")
     mock_time.return_value = self.timestamp
     self.manager.save_eval_set_result(
         self.app_name, self.eval_set_id, self.eval_case_results
@@ -93,8 +91,8 @@ class TestLocalEvalSetResultsManager:
     expected_eval_set_result_json = self.eval_set_result.model_dump_json()
     assert expected_eval_set_result_json == actual_eval_set_result_json
 
-  @patch("time.time")
-  def test_get_eval_set_result(self, mock_time):
+  def test_get_eval_set_result(self, mocker):
+    mock_time = mocker.patch("time.time")
     mock_time.return_value = self.timestamp
     self.manager.save_eval_set_result(
         self.app_name, self.eval_set_id, self.eval_case_results
@@ -104,15 +102,15 @@ class TestLocalEvalSetResultsManager:
     )
     assert retrieved_result == self.eval_set_result
 
-  @patch("time.time")
-  def test_get_eval_set_result_not_found(self, mock_time):
+  def test_get_eval_set_result_not_found(self, mocker):
+    mock_time = mocker.patch("time.time")
     mock_time.return_value = self.timestamp
 
     with pytest.raises(NotFoundError) as e:
       self.manager.get_eval_set_result(self.app_name, "non_existent_id")
 
-  @patch("time.time")
-  def test_list_eval_set_results(self, mock_time):
+  def test_list_eval_set_results(self, mocker):
+    mock_time = mocker.patch("time.time")
     mock_time.return_value = self.timestamp
     # Save two eval set results for the same app
     self.manager.save_eval_set_result(

tests/unittests/evaluation/test_request_intercepter_plugin.py

@@ -14,8 +14,6 @@
 
 from __future__ import annotations
 
-from unittest import mock
-
 from google.adk.agents.callback_context import CallbackContext
 from google.adk.evaluation.request_intercepter_plugin import _LLM_REQUEST_ID_KEY
 from google.adk.evaluation.request_intercepter_plugin import _RequestIntercepterPlugin
@@ -26,7 +24,7 @@ from google.genai import types
 
 class TestRequestIntercepterPlugin:
 
-  async def test_intercept_request_and_response(self):
+  async def test_intercept_request_and_response(self, mocker):
     plugin = _RequestIntercepterPlugin(name="test_plugin")
     llm_request = LlmRequest(
         model="test_model",
@@ -37,7 +35,7 @@ class TestRequestIntercepterPlugin:
             )
         ],
     )
-    mock_invocation_context = mock.MagicMock()
+    mock_invocation_context = mocker.MagicMock()
     mock_invocation_context.session.state = {}
     callback_context = CallbackContext(mock_invocation_context)
     llm_response = LlmResponse()

tests/unittests/evaluation/test_response_evaluator.py

@@ -12,8 +12,9 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from __future__ import annotations
+
 """Tests for the Response Evaluator."""
-from unittest.mock import patch
 
 from google.adk.evaluation.eval_case import Invocation
 from google.adk.evaluation.eval_metrics import PrebuiltMetrics
@@ -24,14 +25,14 @@ import pytest
 from vertexai import types as vertexai_types
 
 
-@patch(
-    "google.adk.evaluation.vertex_ai_eval_facade._VertexAiEvalFacade._perform_eval"
-)
 class TestResponseEvaluator:
   """A class to help organize "patch" that are applicable to all tests."""
 
-  def test_evaluate_invocations_rouge_metric(self, mock_perform_eval):
+  def test_evaluate_invocations_rouge_metric(self, mocker):
     """Test evaluate_invocations function for Rouge metric."""
+    mock_perform_eval = mocker.patch(
+        "google.adk.evaluation.vertex_ai_eval_facade._VertexAiEvalFacade._perform_eval"
+    )
     actual_invocations = [
         Invocation(
             user_content=genai_types.Content(
@@ -67,10 +68,11 @@ class TestResponseEvaluator:
     assert evaluation_result.overall_eval_status == EvalStatus.FAILED
     mock_perform_eval.assert_not_called()  # Ensure _perform_eval was not called
 
-  def test_evaluate_invocations_coherence_metric_passed(
-      self, mock_perform_eval
-  ):
+  def test_evaluate_invocations_coherence_metric_passed(self, mocker):
     """Test evaluate_invocations function for Coherence metric."""
+    mock_perform_eval = mocker.patch(
+        "google.adk.evaluation.vertex_ai_eval_facade._VertexAiEvalFacade._perform_eval"
+    )
     actual_invocations = [
         Invocation(
             user_content=genai_types.Content(
@@ -115,7 +117,7 @@ class TestResponseEvaluator:
         vertexai_types.PrebuiltMetric.COHERENCE.name
     ]
 
-  def test_get_metric_info_response_evaluation_score(self, mock_perform_eval):
+  def test_get_metric_info_response_evaluation_score(self):
     """Test get_metric_info function for response evaluation metric."""
     metric_info = ResponseEvaluator.get_metric_info(
         PrebuiltMetrics.RESPONSE_EVALUATION_SCORE.value
@@ -127,7 +129,7 @@ class TestResponseEvaluator:
     assert metric_info.metric_value_info.interval.min_value == 1.0
     assert metric_info.metric_value_info.interval.max_value == 5.0
 
-  def test_get_metric_info_response_match_score(self, mock_perform_eval):
+  def test_get_metric_info_response_match_score(self):
     """Test get_metric_info function for response match metric."""
     metric_info = ResponseEvaluator.get_metric_info(
         PrebuiltMetrics.RESPONSE_MATCH_SCORE.value
@@ -136,7 +138,7 @@ class TestResponseEvaluator:
     assert metric_info.metric_value_info.interval.min_value == 0.0
     assert metric_info.metric_value_info.interval.max_value == 1.0
 
-  def test_get_metric_info_invalid(self, mock_perform_eval):
+  def test_get_metric_info_invalid(self):
     """Test get_metric_info function for invalid metric."""
     with pytest.raises(ValueError):
       ResponseEvaluator.get_metric_info("invalid_metric")

tests/unittests/evaluation/test_safety_evaluator.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 """Tests for the Response Evaluator."""
-from unittest.mock import patch
 
 from google.adk.evaluation.eval_case import Invocation
 from google.adk.evaluation.eval_metrics import EvalMetric
@@ -24,16 +23,14 @@ from google.genai import types as genai_types
 from vertexai import types as vertexai_types
 
 
-@patch(
-    "google.adk.evaluation.vertex_ai_eval_facade._VertexAiEvalFacade._perform_eval"
-)
 class TestSafetyEvaluatorV1:
   """A class to help organize "patch" that are applicable to all tests."""
 
-  def test_evaluate_invocations_coherence_metric_passed(
-      self, mock_perform_eval
-  ):
+  def test_evaluate_invocations_coherence_metric_passed(self, mocker):
     """Test evaluate_invocations function for Coherence metric."""
+    mock_perform_eval = mocker.patch(
+        "google.adk.evaluation.vertex_ai_eval_facade._VertexAiEvalFacade._perform_eval"
+    )
     actual_invocations = [
         Invocation(
             user_content=genai_types.Content(
@@ -78,7 +75,7 @@ class TestSafetyEvaluatorV1:
         vertexai_types.PrebuiltMetric.SAFETY.name
     ]
 
-  def test_get_metric_info(self, mock_perform_eval):
+  def test_get_metric_info(self):
     """Test get_metric_info function for Safety metric."""
     metric_info = SafetyEvaluatorV1.get_metric_info()
     assert metric_info.metric_name == PrebuiltMetrics.SAFETY_V1.value

tests/unittests/evaluation/test_vertex_ai_eval_facade.py

@@ -12,10 +12,11 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from __future__ import annotations
+
 """Tests for the Response Evaluator."""
 import math
 import random
-from unittest.mock import patch
 
 from google.adk.evaluation.eval_case import Invocation
 from google.adk.evaluation.evaluator import EvalStatus
@@ -25,14 +26,14 @@ import pytest
 from vertexai import types as vertexai_types
 
 
-@patch(
-    "google.adk.evaluation.vertex_ai_eval_facade._VertexAiEvalFacade._perform_eval"
-)
 class TestVertexAiEvalFacade:
   """A class to help organize "patch" that are applicable to all tests."""
 
-  def test_evaluate_invocations_metric_passed(self, mock_perform_eval):
+  def test_evaluate_invocations_metric_passed(self, mocker):
     """Test evaluate_invocations function for a metric."""
+    mock_perform_eval = mocker.patch(
+        "google.adk.evaluation.vertex_ai_eval_facade._VertexAiEvalFacade._perform_eval"
+    )
     actual_invocations = [
         Invocation(
             user_content=genai_types.Content(
@@ -77,8 +78,11 @@ class TestVertexAiEvalFacade:
         vertexai_types.PrebuiltMetric.COHERENCE.name
     ]
 
-  def test_evaluate_invocations_metric_failed(self, mock_perform_eval):
+  def test_evaluate_invocations_metric_failed(self, mocker):
     """Test evaluate_invocations function for a metric."""
+    mock_perform_eval = mocker.patch(
+        "google.adk.evaluation.vertex_ai_eval_facade._VertexAiEvalFacade._perform_eval"
+    )
     actual_invocations = [
         Invocation(
             user_content=genai_types.Content(
@@ -133,9 +137,12 @@ class TestVertexAiEvalFacade:
       ],
   )
   def test_evaluate_invocations_metric_no_score(
-      self, mock_perform_eval, summary_metric_with_no_score
+      self, mocker, summary_metric_with_no_score
   ):
     """Test evaluate_invocations function for a metric."""
+    mock_perform_eval = mocker.patch(
+        "google.adk.evaluation.vertex_ai_eval_facade._VertexAiEvalFacade._perform_eval"
+    )
     actual_invocations = [
         Invocation(
             user_content=genai_types.Content(
@@ -180,10 +187,11 @@ class TestVertexAiEvalFacade:
         vertexai_types.PrebuiltMetric.COHERENCE.name
     ]
 
-  def test_evaluate_invocations_metric_multiple_invocations(
-      self, mock_perform_eval
-  ):
+  def test_evaluate_invocations_metric_multiple_invocations(self, mocker):
     """Test evaluate_invocations function for a metric with multiple invocations."""
+    mock_perform_eval = mocker.patch(
+        "google.adk.evaluation.vertex_ai_eval_facade._VertexAiEvalFacade._perform_eval"
+    )
     num_invocations = 6
     actual_invocations = []
     expected_invocations = []