Skip to content

feat: Add PromptTemplateData to support context and history columns when creating Evaluation run from dataframe #6344

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

Open
copybara-service wants to merge 1 commit into
base: main
Choose a base branch
from
+489 −141
Open

feat: Add PromptTemplateData to support context and history columns when creating Evaluation run from dataframe #6344

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
282 changes: 244 additions & 38 deletions tests/unit/vertexai/genai/replays/test_create_evaluation_run.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,7 @@
from vertexai import types
from google.genai import types as genai_types
import pytest
import pandas as pd

GCS_DEST = "gs://lakeyk-limited-bucket/eval_run_output"
GENERAL_QUALITY_METRIC = types.EvaluationRunMetric(
Expand Down Expand Up @@ -63,28 +64,46 @@
)
),
)

TOOL = genai_types.Tool(
function_declarations=[
genai_types.FunctionDeclaration(
name="get_weather",
description="Get weather in a location",
parameters={
"type": "object",
"properties": {"location": {"type": "string"}},
},
)
]
)
AGENT_INFO = types.evals.AgentInfo(
agent_resource_name="projects/123/locations/us-central1/reasoningEngines/456",
name="agent-1",
instruction="agent-1 instruction",
tool_declarations=[TOOL],
)
DEFAULT_PROMPT_TEMPLATE = "{prompt}"
INPUT_DF_WITH_CONTEXT_AND_HISTORY = pd.DataFrame(
{
"prompt": ["prompt1", "prompt2"],
"reference": ["reference1", "reference2"],
"response": ["response1", "response2"],
"context": ["context1", "context2"],
"conversation_history": ["history1", "history2"],
}
)
CANDIDATE_NAME = "candidate_1"
MODEL_NAME = "projects/503583131166/locations/us-central1/publishers/google/models/gemini-2.5-flash"
EVAL_SET_NAME = "projects/503583131166/locations/us-central1/evaluationSets/6619939608513740800"

def test_create_eval_run_data_source_evaluation_set(client):
"""Tests that create_evaluation_run() creates a correctly structured EvaluationRun."""
client._api_client._http_options.api_version = "v1beta1"
tool = genai_types.Tool(
function_declarations=[
genai_types.FunctionDeclaration(
name="get_weather",
description="Get weather in a location",
parameters={
"type": "object",
"properties": {"location": {"type": "string"}},
},
)
]
)
evaluation_run = client.evals.create_evaluation_run(
name="test4",
display_name="test4",
dataset=types.EvaluationRunDataSource(
evaluation_set="projects/503583131166/locations/us-central1/evaluationSets/6619939608513740800"
evaluation_set=EVAL_SET_NAME
),
dest=GCS_DEST,
metrics=[
Expand All @@ -94,21 +113,14 @@ def test_create_eval_run_data_source_evaluation_set(client):
EXACT_MATCH_COMPUTATION_BASED_METRIC,
BLEU_COMPUTATION_BASED_METRIC,
],
agent_info=types.evals.AgentInfo(
agent_resource_name="project/123/locations/us-central1/reasoningEngines/456",
name="agent-1",
instruction="agent-1 instruction",
tool_declarations=[tool],
),
agent_info=AGENT_INFO,
labels={"label1": "value1"},
)
assert isinstance(evaluation_run, types.EvaluationRun)
assert evaluation_run.display_name == "test4"
assert evaluation_run.state == types.EvaluationRunState.PENDING
assert isinstance(evaluation_run.data_source, types.EvaluationRunDataSource)
assert evaluation_run.data_source.evaluation_set == (
"projects/503583131166/locations/us-central1/evaluationSets/6619939608513740800"
)
assert evaluation_run.data_source.evaluation_set == EVAL_SET_NAME
assert evaluation_run.evaluation_config == types.EvaluationRunConfig(
output_config=genai_types.OutputConfig(
gcs_destination=genai_types.GcsDestination(output_uri_prefix=GCS_DEST)
Expand All @@ -122,13 +134,13 @@ def test_create_eval_run_data_source_evaluation_set(client):
],
)
assert evaluation_run.inference_configs[
"agent-1"
AGENT_INFO.name
] == types.EvaluationRunInferenceConfig(
agent_config=types.EvaluationRunAgentConfig(
developer_instruction=genai_types.Content(
parts=[genai_types.Part(text="agent-1 instruction")]
),
tools=[tool],
tools=[TOOL],
)
)
assert evaluation_run.labels == {
Expand Down Expand Up @@ -190,13 +202,16 @@ def test_create_eval_run_with_inference_configs(client):
"""Tests that create_evaluation_run() creates a correctly structured EvaluationRun with inference_configs."""
client._api_client._http_options.api_version = "v1beta1"
inference_config = types.EvaluationRunInferenceConfig(
model="projects/503583131166/locations/us-central1/publishers/google/models/gemini-2.5-flash"
model=MODEL_NAME,
prompt_template=types.EvaluationRunPromptTemplate(
prompt_template="test prompt template"
),
)
evaluation_run = client.evals.create_evaluation_run(
name="test_inference_config",
display_name="test_inference_config",
dataset=types.EvaluationRunDataSource(
evaluation_set="projects/503583131166/locations/us-central1/evaluationSets/6619939608513740800"
evaluation_set=EVAL_SET_NAME
),
dest=GCS_DEST,
metrics=[GENERAL_QUALITY_METRIC],
Expand All @@ -207,9 +222,7 @@ def test_create_eval_run_with_inference_configs(client):
assert evaluation_run.display_name == "test_inference_config"
assert evaluation_run.state == types.EvaluationRunState.PENDING
assert isinstance(evaluation_run.data_source, types.EvaluationRunDataSource)
assert evaluation_run.data_source.evaluation_set == (
"projects/503583131166/locations/us-central1/evaluationSets/6619939608513740800"
)
assert evaluation_run.data_source.evaluation_set == EVAL_SET_NAME
assert evaluation_run.evaluation_config == types.EvaluationRunConfig(
output_config=genai_types.OutputConfig(
gcs_destination=genai_types.GcsDestination(output_uri_prefix=GCS_DEST)
Expand All @@ -223,9 +236,11 @@ def test_create_eval_run_with_inference_configs(client):
assert evaluation_run.error is None


# Test fails in replay mode because of UUID generation mismatch.
# Dataframe tests fail in replay mode because of UUID generation mismatch.
# def test_create_eval_run_data_source_evaluation_dataset(client):
# """Tests that create_evaluation_run() creates a correctly structured EvaluationRun with EvaluationDataset."""
# """Tests that create_evaluation_run() creates a correctly structured
# EvaluationRun with EvaluationDataset.
# """
# input_df = pd.DataFrame(
# {
# "prompt": ["prompt1", "prompt2"],
Expand Down Expand Up @@ -275,7 +290,7 @@ def test_create_eval_run_with_inference_configs(client):
# name="test6",
# display_name="test6",
# dataset=types.EvaluationDataset(
# candidate_name="candidate_1",
# candidate_name=CANDIDATE_NAME,
# eval_dataset_df=input_df,
# ),
# dest=GCS_DEST,
Expand Down Expand Up @@ -319,6 +334,196 @@ def test_create_eval_run_with_inference_configs(client):
# assert evaluation_run.error is None


# def test_create_eval_run_data_source_evaluation_dataset_with_inference_configs_and_prompt_template_data(
# client,
# ):
# """Tests that create_evaluation_run() creates a correctly structured
# EvaluationRun with EvaluationDataset and inference_configs.
# Prompt template data is inferred from the dataset and a default prompt
# template should be used.
# """
# evaluation_run = client.evals.create_evaluation_run(
# name="test9",
# display_name="test9",
# dataset=types.EvaluationDataset(
# candidate_name=CANDIDATE_NAME,
# eval_dataset_df=INPUT_DF_WITH_CONTEXT_AND_HISTORY,
# ),
# dest=GCS_DEST,
# metrics=[GENERAL_QUALITY_METRIC],
# inference_configs={
# CANDIDATE_NAME: types.EvaluationRunInferenceConfig(
# model=MODEL_NAME,
# )
# },
# )
# assert isinstance(evaluation_run, types.EvaluationRun)
# assert evaluation_run.display_name == "test9"
# assert evaluation_run.state == types.EvaluationRunState.PENDING
# assert isinstance(evaluation_run.data_source, types.EvaluationRunDataSource)
# assert evaluation_run.inference_configs[
# CANDIDATE_NAME
# ] == types.EvaluationRunInferenceConfig(
# model=MODEL_NAME,
# prompt_template=types.EvaluationRunPromptTemplate(
# prompt_template=DEFAULT_PROMPT_TEMPLATE
# ),
# )
# # Check evaluation set
# assert evaluation_run.data_source.evaluation_set
# eval_set = client.evals.get_evaluation_set(
# name=evaluation_run.data_source.evaluation_set
# )
# assert len(eval_set.evaluation_items) == 2
# # Check evaluation items
# for i, eval_item_name in enumerate(eval_set.evaluation_items):
# eval_item = client.evals.get_evaluation_item(name=eval_item_name)
# assert eval_item.evaluation_item_type == types.EvaluationItemType.REQUEST
# assert (
# eval_item.evaluation_request.prompt.prompt_template_data.values[
# "prompt"
# ]
# == genai_types.Content(
# parts=[
# genai_types.Part(
# text=INPUT_DF_WITH_CONTEXT_AND_HISTORY.iloc[i]["prompt"]
# )
# ],
# role="user",
# )
# )
# assert (
# eval_item.evaluation_request.prompt.prompt_template_data.values[
# "context"
# ]
# == genai_types.Content(
# parts=[
# genai_types.Part(
# text=INPUT_DF_WITH_CONTEXT_AND_HISTORY.iloc[i]["context"]
# )
# ],
# role="user",
# )
# )
# assert (
# eval_item.evaluation_request.prompt.prompt_template_data.values[
# "conversation_history"
# ]
# == genai_types.Content(
# parts=[
# genai_types.Part(
# text=(
# INPUT_DF_WITH_CONTEXT_AND_HISTORY.iloc[i][
# "conversation_history"
# ]
# )
# )
# ],
# role="user",
# )
# )
# assert (
# eval_item.evaluation_request.candidate_responses[0].text
# == INPUT_DF_WITH_CONTEXT_AND_HISTORY.iloc[i]["response"]
# )
# assert evaluation_run.error is None


# def test_create_eval_run_data_source_evaluation_dataset_with_agent_info_and_prompt_template_data(
# client,
# ):
# """Tests that create_evaluation_run() creates a correctly structured
# EvaluationRun with EvaluationDataset and agent_info.
# Prompt template data is inferred from the dataset and a default prompt
# template should be used.
# """
# evaluation_run = client.evals.create_evaluation_run(
# name="test9",
# display_name="test9",
# dataset=types.EvaluationDataset(
# candidate_name=CANDIDATE_NAME,
# eval_dataset_df=INPUT_DF_WITH_CONTEXT_AND_HISTORY,
# ),
# dest=GCS_DEST,
# metrics=[GENERAL_QUALITY_METRIC],
# agent_info=AGENT_INFO,
# )
# assert isinstance(evaluation_run, types.EvaluationRun)
# assert evaluation_run.display_name == "test9"
# assert evaluation_run.state == types.EvaluationRunState.PENDING
# assert isinstance(evaluation_run.data_source, types.EvaluationRunDataSource)
# assert evaluation_run.inference_configs[
# AGENT_INFO.name
# ] == types.EvaluationRunInferenceConfig(
# agent_config=types.EvaluationRunAgentConfig(
# developer_instruction=genai_types.Content(
# parts=[genai_types.Part(text=AGENT_INFO.instruction)]
# ),
# tools=[TOOL],
# ),
# prompt_template=types.EvaluationRunPromptTemplate(
# prompt_template=DEFAULT_PROMPT_TEMPLATE
# ),
# )
# # Check evaluation set
# assert evaluation_run.data_source.evaluation_set
# eval_set = client.evals.get_evaluation_set(
# name=evaluation_run.data_source.evaluation_set
# )
# assert len(eval_set.evaluation_items) == 2
# # Check evaluation items
# for i, eval_item_name in enumerate(eval_set.evaluation_items):
# eval_item = client.evals.get_evaluation_item(name=eval_item_name)
# assert eval_item.evaluation_item_type == types.EvaluationItemType.REQUEST
# assert (
# eval_item.evaluation_request.prompt.prompt_template_data.values[
# "prompt"
# ]
# == genai_types.Content(
# parts=[
# genai_types.Part(
# text=INPUT_DF_WITH_CONTEXT_AND_HISTORY.iloc[i]["prompt"]
# )
# ],
# role="user",
# )
# )
# assert (
# eval_item.evaluation_request.prompt.prompt_template_data.values[
# "context"
# ]
# == genai_types.Content(
# parts=[
# genai_types.Part(
# text=INPUT_DF_WITH_CONTEXT_AND_HISTORY.iloc[i]["context"]
# )
# ],
# role="user",
# )
# )
# assert (
# eval_item.evaluation_request.prompt.prompt_template_data.values[
# "conversation_history"
# ]
# == genai_types.Content(
# parts=[
# genai_types.Part(
# text=(
# INPUT_DF_WITH_CONTEXT_AND_HISTORY.iloc[i][
# "conversation_history"
# ]
# )
# )
# ],
# role="user",
# )
# )
# assert (
# eval_item.evaluation_request.candidate_responses[0].text
# == INPUT_DF_WITH_CONTEXT_AND_HISTORY.iloc[i]["response"]
# )
# assert evaluation_run.error is None

pytest_plugins = ("pytest_asyncio",)


Expand Down Expand Up @@ -371,13 +576,16 @@ async def test_create_eval_run_async_with_inference_configs(client):
"""Tests that create_evaluation_run() creates a correctly structured EvaluationRun with inference_configs asynchronously."""
client._api_client._http_options.api_version = "v1beta1"
inference_config = types.EvaluationRunInferenceConfig(
model="projects/503583131166/locations/us-central1/publishers/google/models/gemini-2.5-flash"
model=MODEL_NAME,
prompt_template=types.EvaluationRunPromptTemplate(
prompt_template="Test the {prompt}"
),
)
evaluation_run = await client.aio.evals.create_evaluation_run(
name="test_inference_config_async",
display_name="test_inference_config_async",
dataset=types.EvaluationRunDataSource(
evaluation_set="projects/503583131166/locations/us-central1/evaluationSets/6619939608513740800"
evaluation_set=EVAL_SET_NAME
),
dest=GCS_DEST,
metrics=[GENERAL_QUALITY_METRIC],
Expand All @@ -388,9 +596,7 @@ async def test_create_eval_run_async_with_inference_configs(client):
assert evaluation_run.display_name == "test_inference_config_async"
assert evaluation_run.state == types.EvaluationRunState.PENDING
assert isinstance(evaluation_run.data_source, types.EvaluationRunDataSource)
assert evaluation_run.data_source.evaluation_set == (
"projects/503583131166/locations/us-central1/evaluationSets/6619939608513740800"
)
assert evaluation_run.data_source.evaluation_set == EVAL_SET_NAME
assert evaluation_run.evaluation_config == types.EvaluationRunConfig(
output_config=genai_types.OutputConfig(
gcs_destination=genai_types.GcsDestination(output_uri_prefix=GCS_DEST)
Expand Down
Loading
Loading