Skip to main content
Caronte
Docs

Admin guide · Platform

Limits — budgets & request caps

The Limits page caps usage per API key, in two flavors toggled top-right: spend ceilings (budgets) and request-count ceilings (request caps). Either view opens on a counter strip — how many limits exist, how many are active, how many are already exceeded, and the spend so far — with the left rail filtering by key, period, status and usage state.

Budgets

A spend ceiling (e.g. $10) over a period — Lifetime, Monthly, …. Each row shows the bound key, the limit, a live usage bar, status, and the email recipients notified when thresholds are crossed.

The Limits page (Budgets view): counters for budgets, active, exceeded and current spend, over rows with the bound key, period, limit, a usage bar, status and the alert recipients. Recipient emails are redacted.
Budgets — a per-key spend ceiling with live usage (recipient emails redacted).

Request caps

The same idea for volume: a request-count ceiling (e.g. 15,000 req) over a period. Create either with New budget / New request cap; the auto-refresh control polls live usage.

The Limits page (Requests view): counters for request caps, active, exceeded and current requests, over a per-key monthly cap of 15,000 requests with its usage bar. The recipient email is redacted.
Request caps — a per-key volume ceiling (recipient email redacted).

What happens at the limit

Limits are enforced by the gateway itself, not just reported.

  • When a key exceeds its budget or cap, further requests are rejected until the period resets (resets run automatically).
  • A key can have at most one budget and one request cap.
  • Email recipients are notified as thresholds are crossed, before the hard stop.

tip

Budgets pair naturally with per-key model scoping: scope the key to the models it needs, then cap what it can spend.