Billing support
Usage, Budgets, and Billing
Attribute usage by project, key, model, and business scenario, and plan budgets, alerts, top-ups, and billing exports.
Cost governance needs more than a balance display. Teams should be able to determine who used which model, in which project, at what cost, and whether the usage is unusual.
Recommended attribution dimensions
| Dimension | Purpose |
|---|---|
| Account / team / project | Separate cost centers and environment budgets. |
| API key | Detect key leaks, abnormal scripts, or growth in one service. |
| Model and version | Compare prices, average tokens, latency, and success rates. |
| Business labels | Attribute chat, summarization, RAG, code, or individual tenant costs. |
| Time and status | Distinguish successes, failures, retries, cache hits, and peak periods. |
Recommended cost controls
- Configure monthly soft budgets, hard caps, low-balance reminders, and unusual-growth alerts.
- Automatic top-ups need per-payment amounts, monthly limits, payment-failure notices, and audit records.
- Exports should include time, model, tokens, cost, status, request ID, and attribution labels.
- Reconcile discrepancies between billed amounts and response
usageagainst downloadable settlement details.
These are governance recommendations, not a claim that every control is available in every deployment.
Settlement boundaries to publish
APIQIK must define billing for failures, timeouts, user cancellations, internal routing retries, cache hits, reasoning tokens, tool calls, and multimodal input, as well as billing-record delays and price-change notice periods.