Seehttps://github.com/tensorzero/tensorzero/runs/56187031478

The actual failing test is e2e_tests/evaluations.spec.ts:58:1 › push the new run button, launch an image evaluation, but it causes Error: 2) [chromium] › e2e_tests/optimization.supervised-fine-tuning.spec.ts:168:3 › Custom user agent › @slow should fine-tune on image data with a mocked OpenAI server, because one of the selectors expects "select a metric" which requires an image eval.