Billing, quotas and PAYG

Request lifecycle

1. Authenticate the API key

   The request is attached to a customer account and API key.

2. Resolve the public model

   The public model ID is checked against enabled catalog entries and plan access.

3. Decide billing

   Subscription quota is reserved first. If unavailable and PAYG is enabled, wallet eligibility is checked before upstream and authoritative usage is debited after completion.

4. Call the model route

   The request is routed only after billing and rate-limit checks pass.

5. Settle usage

   Final usage settles subscription quota or debits PAYG. Failed requests are not charged and their quota reservations are released.

Quota and PAYG rules

• Subscription plans use fixed 5h and 7d quota windows that start with the first successful request.
• Each model can consume a different request multiplier per API call.
• PAYG fallback requires wallet credit and account-level PAYG enabled.
• Streaming PAYG requires final usage data to settle correctly.

Rate limits

• Account-level and API-key-level limits are enforced per minute.
• A 429 response means the request was blocked before generation.
• Create separate keys per integration to make rate-limit diagnostics easier.