Spend controls

Three controls decide how fast this workspace can spend, and when it stops. They live on the Settings screen in the dashboard, in its Limits section.

​

Pause every request

Pause is the control for a runaway client, a key you are not sure about, or a weekend you would rather not be surprised by. It refuses requests without revoking anything, so lifting it starts everything again with no keys to re-create.

​

Before you press it

The screen states what it is about to do, in numbers, before anything happens:

Pausing refuses every request on all N live keys straight away, with nothing charged for a refused request. Over the last 7 days this account made N requests, which is the traffic that would have stopped. Nothing is deleted and no key is revoked, so lifting the pause starts everything again with no keys to re-create.

Both counts are your workspace’s own, read at the moment the screen loads. The request count is the trailing seven days of what actually happened, not a forecast of what will stop. Nobody can count future requests, so the screen quotes the honest stand-in and says which it is.

​

While it is on

Requests are refused with HTTP 503 and code paused_by_owner, and nothing is charged for them. The screen explains the state rather than only showing a toggle, because somebody who paused on Friday and is reading this on Monday has forgotten:

Every request on this account is refused while this is on, on all N live keys, and nothing is charged for a refused request. Credit already held for requests that were running when you paused is still theirs: each one finishes and is charged for what it really used. Lift the pause here and requests start working again immediately, with no keys to re-create.

Requests already in flight are not cut off. Each finishes and is charged for what it really used.

​

Lifting it

Press it again. The next request is admitted immediately: there is no cache to expire and no deploy involved. Your keys, your members and your credit are exactly as they were.

​

Set your spend rate cap

The spend rate cap limits how much money may be reserved per rolling 60 seconds. The default is 5000 minor units. Over it, requests are refused with HTTP 429 and code spend_rate_exceeded, and they are admitted again within 60 seconds with nothing for you to do. Read Rate limits and spend controls first if you have not: the cap counts estimated maximums rather than money spent, and sending max_tokens is the lever that makes it bind less often. Two rules worth knowing:

• A cap of 0 is refused. Stopping requests entirely is what the pause is for, and it has its own control, its own state and its own sentence. A cap of zero would be a second way to do it that no screen would explain.
• Lowering always works. If your cap is above the self service ceiling, you can still lower it to any figure at or above 100 minor units. You just cannot raise it further until we raise the ceiling.

​

The rows you cannot raise

The request rate and the token rates protect a provider account shared across every customer, and the screen says so rather than only enforcing it:

This protects a provider account we share across every customer. Raising it is a decision we make, not one you can make yourself.

There is no self service path for those figures. If you need more room, ask us. Two of the rows carry no figure yet:

Not yet enforced. We are not checking your traffic against this figure yet.

They are drawn because the units we meter you in are fixed now rather than later. No figure means nobody is checking, not that you have no room. The Limits section and the self service cap control are shipped. Every row reads the same live data the platform enforces against, and changing your spend rate cap here takes effect exactly the way the steps above describe.

​

Choose when we warn you

The low credit notice emails you once, at your verified address, when your available balance falls below a figure you choose.

• A whole number of minor units, from 0 to 100000. 500 means 5.00.
• 0 turns the email off, and the screen says so in words.
• One email per crossing. No second email is sent until the balance goes back above the figure and falls below it again, and the screen says which of those three states you are in.

The email names the balance you hold and the threshold you set, says that requests are refused once credit runs out and that nothing is charged for a request we refuse, and points at Billing. It carries no link with a token in it.

​

What each refusal costs you

Nothing. A paused request, a rate capped request and a request refused for want of credit are all charged nothing. That is why the pause is safe to press and the cap is safe to test.

​

Next