# Acceptance - [ ] Model choice, reasoning flag, token counts, cost, and reason codes are emitted - [ ] Warning/alerts are configurable.