Limits and the spend cap

Every request your company sends to DataMind Platform passes the limits your company holds before it reaches a model: how many requests can go out per minute, how many can run at once, and how much billable spending is allowed in the month. This page explains what each limit gates, where your company's values are shown, and what a request meets when a limit stops it. The full status and code reference is on Errors and retries.

The limits at a glance

Each value is set for your company, and the console shows the current ones.

Limit	What it caps	Where you see it
Requests per minute	How many requests your company can send each minute.	Settings, the Rate / concurrency row.
Requests at once	How many requests your company can have running at the same time.	Settings, the same row.
Monthly spend cap	Billable extras in the billing window: Premium overage, Standard overage and Pay-as-you-go spend together.	Settings, the Enforced cap row; an Administrator edits Monthly spend cap (billable extras).
Hard ceiling	The highest spend cap DataMind allows for your company.	Settings, the Hard ceiling row.

The number the platform applies to billable usage is the Enforced cap: your spend cap or the hard ceiling, whichever is lower. The Dashboard fills the Extra spend bar toward the enforced cap and shows the Ceiling beside it.

How a request passes the limits

The requests-per-minute rate and the requests-at-once limit gate every request, whatever model it names. Chat requests and web search and web read calls pass the same gates.
The spend cap gates only billable usage. Tokens included in your plan serve first and are never refused by the cap; pay-as-you-go requests are always billable, and Premium or Standard usage counts once its allowance is spent and the plan prices that overage.
A company without an active plan is refused until an Administrator selects one on Pricing.

If your plan prices no Standard overage, Standard requests are refused the moment the allowance is spent, and the console shows Standard requests are being refused. Premium keeps serving, and a higher spend cap does not help. Pricing lists the ways to keep working.

What a request meets when a limit stops it

Status	`code`	When it happens	What to do
429	`rate_limit_exceeded`	Your company sent more requests this minute than the rate allows.	Wait the seconds in the `retry-after` header.
429	`concurrency_limit_exceeded`	Your company is running more requests at once than the limit allows.	Wait for a running request to finish; `retry-after` is `1`.
402	`budget_exceeded`	Extra spend reached the enforced cap.	Wait for the next billing window, or raise the cap in Settings.
402	`no_active_subscription`	The company has no active plan.	An Administrator selects a plan on Pricing.
402	`standard_quota_exceeded`	The Standard allowance is spent and the plan prices no Standard overage.	Use Kartlos Premium or a pay-as-you-go model, or ask about a plan that prices Standard overage.
503	`service_unavailable`	The limits cannot be checked right now.	Retry shortly.

Money refusals answer 402 and not 429 on purpose: automatic retry logic retries 429, and a billing state cannot be fixed by retrying.
A limit gates a request before it starts. An answer that is already streaming is never cut off.
A refused request still counts toward your requests-per-minute rate for that minute, so wait the retry-after seconds before sending it again.
A request refused at a limit does not reach a model and is not billed.
If the platform cannot read your company's limits, the request is refused rather than served: a request that cannot be checked is not forwarded. The refusal reads Account limits are unavailable - retry shortly.

Where the values come from and how they change

The current values are on the Spend cap card in Settings: the Enforced cap, the Hard ceiling, the Rate / concurrency row and the allowance lines.
Pricing repeats the Enforced cap and the Ceiling in the current billing window strip.
An Administrator changes the monthly spend cap in Settings: enter the amount in Monthly spend cap (billable extras) and click Save. The cap cannot be higher than the hard ceiling; to raise the ceiling, contact DataMind support.
A saved change applies to new requests straight away.

Questions, answered

What limits does my company have?

Four: requests per minute, requests at once, a monthly spend cap and a hard ceiling. The first two gate every request; the two money values form the Enforced cap: whichever is lower. Your company's current values are on the Spend cap card in Settings.

Which requests does the spend cap gate?

Only billable usage: Premium overage, Standard overage and Pay-as-you-go spend. Tokens included in your plan serve first and the cap never refuses them; pay-as-you-go requests are always billable. What makes up Extra spend is on the Dashboard.

Why was my request refused when my plan still has tokens?

Another limit usually stopped it. The requests-per-minute rate and the requests-at-once limit gate every request, so they can refuse one while your allowance is untouched; and a company without an active plan is refused until an Administrator selects one on Pricing. Usage inside your included allowance is never refused by the spend cap itself.

What happens when the spend cap is reached?

Billable requests are refused with status 402 and code budget_exceeded, until the next billing window opens or an Administrator raises the cap in Settings. Requests covered by your plan's included allowance keep serving, and no answer is cut off mid-stream. Retrying cannot succeed until the cap or the window changes.

Why can extra spend pass the cap?

The check happens before a request starts, and usage is counted when a request finishes. Requests already running when the cap is reached can add a little more spend before the refusals begin. The platform lets them finish on purpose rather than stopping an answer mid-stream.

Can I change the spend cap myself?

Yes, if you are an Administrator: on the Spend cap card in Settings, enter the amount in Monthly spend cap (billable extras) and click Save. The cap cannot be higher than the hard ceiling; to raise the ceiling, contact DataMind support. A Viewer sees the values and cannot change them.

Do refused requests count against my rate limit?

Yes. A refused request keeps its place in the current minute's count, so wait the retry-after seconds and send it again rather than retrying in a loop.

What happens if the platform cannot check my limits?

The request is refused rather than served, with status 503 and the message Account limits are unavailable - retry shortly. A request the platform cannot measure against your limits is never forwarded. Retry shortly.