fix(eval): handle failed inference results without invocations

Merge #5878

## What changed

- Return a failed `EvalCaseResult` when inference failed before producing any invocations.
- Preserve the existing session lookup path when a failed inference still has a session id.
- Add a regression test for `InferenceResult(status=FAILURE, inferences=None)`.
- Clean up two existing lint issues in the touched eval test file so the local changed-file ruff check passes.

This prevents the eval runner from replacing the original inference error with `TypeError: object of type 'NoneType' has no len()`.

Fixes #5876

## To verify

- `.\.venv\Scripts\python.exe -m py_compile src\google\adk\evaluation\local_eval_service.py tests\unittests\evaluation\test_local_eval_service.py`
- `.\.venv\Scripts\python.exe -m pytest tests\unittests\evaluation\test_local_eval_service.py -k "failed_without_inferences or evaluate_single_inference_result" -q --basetemp .tmp\pytest`
- `.\.venv\Scripts\python.exe -m ruff check src\google\adk\evaluation\local_eval_service.py tests\unittests\evaluation\test_local_eval_service.py`
- `.\.venv\Scripts\python.exe -m pyink --check src\google\adk\evaluation\local_eval_service.py tests\unittests\evaluation\test_local_eval_service.py`
- `git diff --check`

Co-authored-by: George Weale <gweale@google.com>
COPYBARA_INTEGRATE_REVIEW=#5878 from he-yufeng:fix/eval-failed-inference-none 5ed777d
PiperOrigin-RevId: 933176248

Author: 2 people

src/google/adk/evaluation/local_eval_service.py

@@ -277,6 +277,29 @@ async def _evaluate_single_inference_result(
         else 'test_user_id'
     )
 
+    if inference_result.inferences is None:
+      session_details = None
+      if inference_result.session_id is not None:
+        session_details = await self._session_service.get_session(
+            app_name=inference_result.app_name,
+            user_id=user_id,
+            session_id=inference_result.session_id,
+        )
+      return (
+          inference_result,
+          EvalCaseResult(
+              eval_set_file=inference_result.eval_set_id,
+              eval_set_id=inference_result.eval_set_id,
+              eval_id=inference_result.eval_case_id,
+              final_eval_status=EvalStatus.FAILED,
+              overall_eval_metric_results=[],
+              eval_metric_result_per_invocation=[],
+              session_id=inference_result.session_id or '',
+              session_details=session_details,
+              user_id=user_id,
+          ),
+      )
+
     if eval_case.conversation_scenario is None and len(
         inference_result.inferences
     ) != len(eval_case.conversation):

tests/unittests/evaluation/test_local_eval_service.py

@@ -15,7 +15,6 @@
 from __future__ import annotations
 
 import asyncio
-import sys
 from typing import Optional
 
 from google.adk.agents.llm_agent import LlmAgent
@@ -465,6 +464,39 @@ async def test_evaluate_single_inference_result(
     assert metric_result.eval_status == EvalStatus.PASSED
 
 
+@pytest.mark.asyncio
+async def test_evaluate_single_inference_result_failed_without_inferences(
+    eval_service, mock_eval_sets_manager, mocker
+):
+  inference_result = InferenceResult(
+      app_name="test_app",
+      eval_set_id="test_eval_set",
+      eval_case_id="case1",
+      inferences=None,
+      session_id="session1",
+      status=InferenceStatus.FAILURE,
+      error_message="auth failed",
+  )
+  eval_metric = EvalMetric(metric_name="fake_metric", threshold=0.5)
+  evaluate_config = EvaluateConfig(eval_metrics=[eval_metric], parallelism=1)
+
+  mock_eval_case = mocker.MagicMock(spec=EvalCase)
+  mock_eval_case.conversation = []
+  mock_eval_case.conversation_scenario = None
+  mock_eval_case.session_input = None
+  mock_eval_sets_manager.get_eval_case.return_value = mock_eval_case
+
+  _, result = await eval_service._evaluate_single_inference_result(
+      inference_result=inference_result, evaluate_config=evaluate_config
+  )
+
+  assert result.eval_id == "case1"
+  assert result.session_id == "session1"
+  assert result.final_eval_status == EvalStatus.FAILED
+  assert result.overall_eval_metric_results == []
+  assert result.eval_metric_result_per_invocation == []
+
+
 @pytest.mark.asyncio
 async def test_evaluate_single_inference_result_for_conversation_scenario(
     eval_service, mock_eval_sets_manager, mocker
@@ -520,7 +552,7 @@ async def test_evaluate_single_inference_result_for_conversation_scenario(
   for i in range(3):
     invocation_result = result.eval_metric_result_per_invocation[i]
     assert invocation_result.actual_invocation == inference_result.inferences[i]
-    assert invocation_result.expected_invocation == None
+    assert invocation_result.expected_invocation is None
     assert len(invocation_result.eval_metric_results) == 1
     metric_result = invocation_result.eval_metric_results[0]
     assert metric_result.metric_name == "fake_single_sided_metric"