heyyy / testing environment
Assistant lab
Emulated tools · real model calls · synthetic test data. Browserbase is not used.
Try the assistant
Test a conversation with the selected fixture.
Or run a fresh, identical scenario across models.
Or run a fresh, identical scenario across models.
Run evidence
Outputs, tokens and tool provenance. “Real” means browser execution, not real-world availability. Search is always emulated; reactions are UI-only. Quality is reviewed against the rubric, not auto-scored.
No runs yet. Results remain in this tab until you reload; export to keep them.