You set the limits.
Every API key can have a monthly limit in euros, and your team can have a shared one. Once a limit is reached, the API rejects the call before a provider is called.
How the check adds up.
Before every call, FlowModels reserves the highest possible amount. The call only runs if the limit still holds with that reservation.
Otherwise the API rejects it
429 spend_limit_exceeded
Three checks, in this order.
A call must pass all three. The first one that fails decides the response.
| Check | Applies to | Response |
|---|---|---|
| Credit covers the reservation | your account | 429 insufficient_quota |
| Monthly limit of the team, if set | all keys of the team | 429 spend_limit_exceeded |
| Monthly limit of the key, if set | exactly this key | 429 spend_limit_exceeded |
No budget per end user.
FlowModels has no limit per end user. Limits apply to keys and teams. You can still see how much one person uses: GET /v1/usage?group_by=end_user. You build a ceiling per person in your application.
Where you set what.
| Limit | Values | Setting |
|---|---|---|
| Monthly limit per key | Whole euro amounts from €1 to €100,000. Empty means no limit. | Edit key |
| Monthly limit of the team | Euro amount with up to two decimals. 0 means no calls, empty means no limit. | Credit, monthly spending limit |
| Requests per minute per key | Your own ceiling below the tier of your team. | Edit key |
A month is a calendar month. The limit only says how much you want to spend; it does not add any credit.
Warning thresholds, before it gets tight.
In the dashboard you set up to ten amounts per list. We notify you by email, in the app and, if you subscribed to it, through the credit.threshold webhook.
| List | You are warned | Value in data.metric |
|---|---|---|
| Available amount falls below | when credit minus reservations falls below the threshold, again the next time it falls below | balance_floor |
| Spending in the month reaches | when billed calls reach the amount, once per calendar month | spend_month |
Your key knows its limit.
GET /v1/key states, for the calling key, the monthly limit in micro-euros, the per-minute limit and the concurrent requests. The response never returns the key itself.
{
"object": "api_key",
"uuid": "018f1234-5678-7abc-9def-0123456789ab",
"name": "support-assistant",
"status": "active",
"kind": "inference",
"limits": {
"spending": {
"currency": "EUR",
"monthly_limit_micro_eur": 50000000
},
"requests": {
"tier": 1,
"per_minute": 60
},
"concurrency": {
"max_concurrent_requests": 10
}
}
}