Skip to content

chore: GenAI Client - Add replay tests for additional RubricMetrics coverage in evals SDK #6602

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
copybara-service merged 1 commit into from
Apr 25, 2026
+183 −3
Merged

chore: GenAI Client - Add replay tests for additional RubricMetrics coverage in evals SDK #6602

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
122 changes: 121 additions & 1 deletion tests/unit/vertexai/genai/replays/test_evaluate.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -458,10 +458,130 @@ def parse_results(responses):
"my_custom_metric"
]
assert metric_result.score is not None
assert metric_result.score > 0.2
assert metric_result.score >= 0.0
assert metric_result.error_message is None


def test_evaluation_single_turn_agent_data(client):
"""Tests single-turn AgentData eval with agent quality metrics."""
client._api_client._http_options.api_version = "v1beta1"

weather_agent = {
"weather_bot": types.evals.AgentConfig(
agent_id="weather_bot",
agent_type="SpecialistAgent",
description="Handles weather queries.",
instruction=(
"You are a weather assistant. Use the get_weather tool to"
" answer weather questions."
),
tools=[
genai_types.Tool(
function_declarations=[
genai_types.FunctionDeclaration(
name="get_weather",
description=(
"Gets the current weather for a given location."
),
)
]
)
],
),
}

eval_case = types.EvalCase(
eval_case_id="successful-tool-use",
agent_data=types.evals.AgentData(
agents=weather_agent,
turns=[
types.evals.ConversationTurn(
turn_index=0,
events=[
types.evals.AgentEvent(
author="user",
content=genai_types.Content(
role="user",
parts=[
genai_types.Part(
text="What is the weather in Tokyo?"
)
],
),
),
types.evals.AgentEvent(
author="weather_bot",
content=genai_types.Content(
role="model",
parts=[
genai_types.Part(
function_call=genai_types.FunctionCall(
id="tool_call_0",
name="get_weather",
args={"location": "Tokyo"},
)
)
],
),
),
types.evals.AgentEvent(
author="weather_bot",
content=genai_types.Content(
role="tool",
parts=[
genai_types.Part(
function_response=genai_types.FunctionResponse(
id="tool_call_0",
name="get_weather",
response={"weather": "75F and sunny"},
)
)
],
),
),
types.evals.AgentEvent(
author="weather_bot",
content=genai_types.Content(
role="model",
parts=[
genai_types.Part(
text=(
"It is currently 75F and sunny in" " Tokyo."
)
)
],
),
),
],
)
],
),
)

eval_dataset = types.EvaluationDataset(eval_cases=[eval_case])

metrics = [
types.RubricMetric.FINAL_RESPONSE_QUALITY,
types.RubricMetric.TOOL_USE_QUALITY,
types.RubricMetric.HALLUCINATION,
types.RubricMetric.SAFETY,
types.RubricMetric.GENERAL_QUALITY,
types.RubricMetric.TEXT_QUALITY,
]

evaluation_result = client.evals.evaluate(dataset=eval_dataset, metrics=metrics)

assert isinstance(evaluation_result, types.EvaluationResult)
assert evaluation_result.summary_metrics is not None
assert len(evaluation_result.summary_metrics) > 0
for summary in evaluation_result.summary_metrics:
assert isinstance(summary, types.AggregatedMetricResult)
assert summary.metric_name is not None

assert evaluation_result.eval_case_results is not None
assert len(evaluation_result.eval_case_results) == 1


pytestmark = pytest_helper.setup(
file=__file__,
globals_for_file=globals(),
Expand Down
64 changes: 62 additions & 2 deletions tests/unit/vertexai/genai/replays/test_evaluate_predefined_metrics.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -224,6 +224,7 @@ def test_multi_turn_predefined_metric(client):

predefined_metrics = [
types.RubricMetric.MULTI_TURN_GENERAL_QUALITY,
types.RubricMetric.MULTI_TURN_TEXT_QUALITY,
]

evaluation_result = client.evals.evaluate(
Expand All @@ -233,11 +234,16 @@ def test_multi_turn_predefined_metric(client):

assert isinstance(evaluation_result, types.EvaluationResult)
assert evaluation_result.summary_metrics is not None
assert len(evaluation_result.summary_metrics) > 0
assert len(evaluation_result.summary_metrics) == 2
metric_names = set()
for summary in evaluation_result.summary_metrics:
assert isinstance(summary, types.AggregatedMetricResult)
assert summary.metric_name == "multi_turn_general_quality_v1"
metric_names.add(summary.metric_name)
assert isinstance(summary.mean_score, float)
assert metric_names == {
"multi_turn_general_quality_v1",
"multi_turn_text_quality_v1",
}

assert evaluation_result.eval_case_results is not None
assert len(evaluation_result.eval_case_results) > 0
Expand Down Expand Up @@ -415,6 +421,60 @@ def test_evaluation_gecko_text2video_metric(client):
assert case_result.response_candidate_results is not None


def test_single_turn_rubric_metrics(client):
"""Tests single-turn text quality RubricMetrics with reference."""
prompts_df = pd.DataFrame(
{
"prompt": ["Summarize the benefits of regular exercise."],
"response": [
"Exercise improves cardiovascular health, boosts mood through"
" endorphin release, strengthens muscles and bones, and enhances"
" sleep quality. Regular physical activity also helps maintain a"
" healthy weight and reduces the risk of chronic diseases."
],
"reference": [
"Exercise improves heart health, mood, muscle strength," " and sleep."
],
"context": [
"Exercise improves heart health, mood, muscle strength," " and sleep."
],
}
)

eval_dataset = types.EvaluationDataset(
eval_dataset_df=prompts_df,
candidate_name="gemini-2.5-flash",
)

predefined_metrics = [
types.RubricMetric.INSTRUCTION_FOLLOWING,
types.RubricMetric.GENERAL_QUALITY,
types.RubricMetric.TEXT_QUALITY,
types.RubricMetric.GROUNDING,
types.RubricMetric.SAFETY,
types.RubricMetric.FINAL_RESPONSE_MATCH,
]

evaluation_result = client.evals.evaluate(
dataset=eval_dataset,
metrics=predefined_metrics,
)

assert isinstance(evaluation_result, types.EvaluationResult)
assert evaluation_result.summary_metrics is not None
assert len(evaluation_result.summary_metrics) > 0
for summary in evaluation_result.summary_metrics:
assert isinstance(summary, types.AggregatedMetricResult)
assert summary.metric_name is not None

assert evaluation_result.eval_case_results is not None
assert len(evaluation_result.eval_case_results) > 0
for case_result in evaluation_result.eval_case_results:
assert isinstance(case_result, types.EvalCaseResult)
assert case_result.eval_case_index is not None
assert case_result.response_candidate_results is not None


pytestmark = pytest_helper.setup(
file=__file__,
globals_for_file=globals(),
Expand Down
Loading