Fleet dashboard
Exec summary across every registered service + schema. Top row is absolute fleet size; remaining rows are 7-day rolling activity. Cards link to the full detail surface.
Review activity
Edit rate = fraction of LLM-annotated elements a reviewer subsequently touched. Low is good — the annotator matched intent. High means the annotator needs training feedback. (PRD §16 north-star metric.)
Edits per LLM annotation (7d)—no annotations yet — infer a model first
Edits per LLM annotation (30d)—no annotations yet — infer a model first
Related: tokens — AI budget · audit — full trail · infer a new schema