"""Evaluate the target operating model agent on benchmark cases with real models. Each case runs one pre-approved Execution through the full assembly (the ADK runtime and the agent's own Skills and role files), then scores it with eval/scoring.py. It costs model usage: run it deliberately. Usage: GOOGLE_GENAI_USE_ENTERPRISE=true GOOGLE_CLOUD_PROJECT= GOOGLE_CLOUD_LOCATION=global \\ .venv/bin/python -m eval.operating_model_eval [--repeat N] [case-id ...] The model settings are read from the environment or from `.env`. With no case named, every case in eval/datasets/cases.json runs, one after another; `--repeat N` runs each N times, so the spread of what is staffed and priced can be seen. Results are written under eval/results/