Skip to content

fix: GenAI Client(evals) - Parse JSON-string and OpenAI-style conversation_history #7214

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
Oct 16, 2026
Merged

fix: GenAI Client(evals) - Parse JSON-string and OpenAI-style conversation_history #7214

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
17 changes: 17 additions & 0 deletions agentplatform/_genai/_evals_data_converters.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -230,6 +230,18 @@ def convert(self, raw_data: list[dict[str, Any]]) -> types.EvaluationDataset:
"Invalid prompt type for case %s: %s" % (i, type(prompt_data))
)

if (
isinstance(conversation_history_data, str)
and conversation_history_data.strip()
):
try:
conversation_history_data = json.loads(conversation_history_data)
except json.JSONDecodeError:
logger.warning(
"Could not decode conversation_history JSON string for case %s.",
eval_case_id,
)

conversation_history: Optional[list[types.evals.Message]] = None
if isinstance(conversation_history_data, list):
conversation_history = []
Expand All @@ -242,6 +254,11 @@ def convert(self, raw_data: list[dict[str, Any]]) -> types.EvaluationDataset:
)
)
elif isinstance(content, dict):
if isinstance(content.get("content"), str):
content = {
"role": content.get("role"),
"parts": [{"text": content["content"]}],
}
try:
validated_content = genai_types.Content.model_validate(
content
Expand Down
65 changes: 65 additions & 0 deletions tests/unit/agentplatform/genai/test_evals.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,6 +45,9 @@
types as agentplatform_genai_types,
)
from agentplatform._genai.types import common as common_types
from vertexai._genai import (
_evals_data_converters as vertexai_evals_data_converters,
)
from vertexai._genai import types as vertexai_genai_types
from google.genai import client
from google.genai import errors as genai_errors
Expand Down Expand Up @@ -5867,6 +5870,68 @@ def test_convert_with_conversation_history(self):
eval_case.conversation_history[1].content.parts[0].text == "Old model msg"
)

@pytest.mark.parametrize(
"converters",
[_evals_data_converters, vertexai_evals_data_converters],
ids=["agent_platform", "vertexai"],
)
@pytest.mark.parametrize(
"history, model_role",
[
(
'[{"role": "user", "parts": [{"text": "Old user msg"}]},'
' {"role": "model", "parts": [{"text": "Old model msg"}]}]',
"model",
),
(
'[{"role": "user", "content": "Old user msg"},'
' {"role": "assistant", "content": "Old model msg"}]',
"assistant",
),
(
[
{"role": "user", "content": "Old user msg"},
{"role": "assistant", "content": "Old model msg"},
],
"assistant",
),
],
ids=["json_string", "openai_json_string", "openai_dicts"],
)
def test_convert_parses_string_and_openai_conversation_history(
self, converters, history, model_role
):
result_dataset = converters._FlattenEvalDataConverter().convert(
[
{
"prompt": "Current prompt",
"response": "A response",
"conversation_history": history,
}
]
)

assert [
(message.content.role, message.content.parts[0].text)
for message in result_dataset.eval_cases[0].conversation_history
] == [("user", "Old user msg"), (model_role, "Old model msg")]

@pytest.mark.parametrize(
"history, warns",
[("not json", True), (" ", False)],
ids=["invalid_json", "blank"],
)
def test_convert_skips_unparsable_conversation_history(
self, history, warns, caplog
):
with caplog.at_level("WARNING", logger=_evals_data_converters.logger.name):
result_dataset = self.converter.convert(
[{"prompt": "p", "response": "r", "conversation_history": history}]
)

assert result_dataset.eval_cases[0].conversation_history is None
assert ("Could not decode conversation_history" in caplog.text) == warns

def test_convert_with_conversation_history_column_name(self):
"""Tests that 'conversation_history' is accepted as a column name alias for 'history'."""
raw_data_df = pd.DataFrame(
Expand Down
17 changes: 17 additions & 0 deletions vertexai/_genai/_evals_data_converters.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -230,6 +230,18 @@ def convert(self, raw_data: list[dict[str, Any]]) -> types.EvaluationDataset:
"Invalid prompt type for case %s: %s" % (i, type(prompt_data))
)

if (
isinstance(conversation_history_data, str)
and conversation_history_data.strip()
):
try:
conversation_history_data = json.loads(conversation_history_data)
except json.JSONDecodeError:
logger.warning(
"Could not decode conversation_history JSON string for case %s.",
eval_case_id,
)

conversation_history: Optional[list[types.evals.Message]] = None
if isinstance(conversation_history_data, list):
conversation_history = []
Expand All @@ -242,6 +254,11 @@ def convert(self, raw_data: list[dict[str, Any]]) -> types.EvaluationDataset:
)
)
elif isinstance(content, dict):
if isinstance(content.get("content"), str):
content = {
"role": content.get("role"),
"parts": [{"text": content["content"]}],
}
try:
validated_content = genai_types.Content.model_validate(
content
Expand Down
Loading