Budgets and limits
How spend limits work and when they reset.
A budget is a limit in USD, tokens, or both, for one period: a day, a week (Monday start), a month, or total. Periods follow the instance time zone (Asia/Makassar).
Budgets can sit at four levels, and every one that applies is checked:
| Level | Set by |
|---|---|
| Instance | Simpang admin |
| Workspace (shared by everyone in it) | Simpang admin |
| You, inside the workspace | workspace owner or admin |
| One key | you, when you create the key |
When a call would go over a limit, it's refused with 402 budget_exceeded. The message says which limit and when it resets.
Extra budget
A workspace owner or admin can grant extra on your cap for the current period, with a reason. Extra on the shared workspace budget comes from the Simpang admin. Your limit for the period is the base limit plus grants.
How exact is it
Before a call starts, Simpang reserves an estimate (input size plus the output cap) and checks it against your limit. After the call it charges the real cost. Several long calls running at once can each pass the check, so spend can end up slightly over a limit by the gap between estimates and real usage.
The overview shows your pace: at the current rate, when you'd hit the limit.