This session added Redis-backed per-user weekly generation caps and a global daily stopcock (usageService.js), plus a minimal UI quota chip in ArchitectureView. There is no operational view of the shared budget.
Work needed:
- Extend /internal/metrics (already admin-secret-gated) with usageService stats: per-user gens used, global daily gens, chat throttle hit-rate
- A tiny internal dashboard route (HTML) or JSON endpoint the operator can curl
- Optional threshold alerts (warn when the daily gens stopcock is near the cap)
Why: one shared free Kilo key serves all users; the operator needs to see when the pipe saturates to tune limits before users hit errors.
This session added Redis-backed per-user weekly generation caps and a global daily stopcock (usageService.js), plus a minimal UI quota chip in ArchitectureView. There is no operational view of the shared budget.
Work needed:
Why: one shared free Kilo key serves all users; the operator needs to see when the pipe saturates to tune limits before users hit errors.