Problem

Test (evaluation) in the Secure Integration test workflow is failing on main (run 30494098538):

ERROR tests_integ/evaluation/test_third_party_adapters.py::TestDeepEvalAdapterIntegration::test_bias_metric_passes - ModuleNotFoundError: No module named 'deepeval'
ERROR tests_integ/evaluation/test_third_party_adapters.py::TestAutoEvalsAdapterIntegration::test_factuality_scorer - ModuleNotFoundError: No module named 'autoevals'
======= 36 passed, 1 skipped, 41 warnings, 6 errors in 78.52s (0:01:18) ========

test_third_party_adapters.py (added in #568) imports deepeval and autoevals, but the evaluation matrix group only installed strands-agents-evals. The bare import calls in the autouse fixtures raise ModuleNotFoundError, which pytest reports as errors rather than skips, so the job exits 1.

Both packages are already declared as optional extras in pyproject.toml ([deepeval], [autoevals]) — they were simply never installed in that job.

Change

One line — add both packages to the evaluation group's extra-deps.

Testing

• python3 -c "yaml.safe_load(...)" — workflow YAML parses, matrix entry correct
• Installed the full evaluation dep set on Python 3.10 (CI's version); deepeval==4.1.4 and autoevals both resolve and import cleanly, no conflicts with strands-agents-evals==1.0.3
• Confirmed the two ModuleNotFoundErrors are gone, and confirmed the residual OpenAI-credential failures documented above