Billing and usage

Prepaid, per token

AI inference is billed from the same prepaid balance as your apps: you redeem a voucher, the value lands as credit, and token usage is metered against it. There is no subscription and no minimum. A request costs exactly what its tokens cost:

Per-model prices are listed on models and pricing; estimate a workload with the pricing calculator.

Tokens are counted on the server

Billing uses the token counts our gateway captures from the inference engine, never client-reported numbers. Every response includes them in the usage object:

"usage": {
  "prompt_tokens": 21,
  "completion_tokens": 38,
  "total_tokens": 59
}

In streaming mode the final chunk carries the same usage object (the gateway forces stream_options.include_usage on), so what you observe in the response is exactly what you are billed for: prompt_tokens at the input rate, completion_tokens at the output rate.

Hourly debits

Usage is recorded per request and debited from your balance once per hour, as a single ledger entry per hour of usage. The debits appear in the same transaction history as your voucher credits, on the Billing page in the portal, where credits and charges stay auditable side by side.

Track your token consumption per day and per model on the AI page in the portal, or programmatically (with your portal access token):

curl -sS "https://api.alawadi.cloud/v1/ai/usage?from=2026-06-01&to=2026-06-30" \
  -H "Authorization: Bearer $PORTAL_ACCESS_TOKEN"

The response is token and request counts only, broken down per day and per model. It does not return money: cost is computed from the rates above, and the actual charges land as the hourly debits in your transaction history.

{
  "from": "2026-06-01",
  "to": "2026-06-30",
  "rows": [
    {
      "day": "2026-06-12",
      "model": "qwen2.5-7b-instruct",
      "prompt_tokens": 18240,
      "completion_tokens": 9610,
      "requests": 142
    }
  ],
  "totals": {
    "prompt_tokens": 512300,
    "completion_tokens": 244180,
    "requests": 3120
  }
}

Export an audit report

For compliance reviews or chargeback, export everything your account did on the AI surface — inference usage events, the AI key lifecycle (prefixes only, never secrets), and your agents with their revision history — from GET /v1/ai/audit (owner-scoped, portal access token):

curl -sS "https://api.alawadi.cloud/v1/ai/audit?from=2026-06-01&to=2026-06-30&format=json" \
  -H "Authorization: Bearer $PORTAL_ACCESS_TOKEN"

The window defaults to the last 30 days. kind=usage|keys|agents selects a subset (default: all three); format=csv downloads a spreadsheet-ready file and requires a single kind, one table per file. Full parameters and schemas are in the AI Agents API reference.

Zero balance and suspension

A balance pre-check runs on every request, before any tokens are generated:

• Balance at zero (or no balance yet) → the request is rejected with 402 and code INSUFFICIENT_BALANCE. No tokens are consumed and nothing is charged.
• Account suspended for billing → requests are rejected with 403 and code ACCOUNT_SUSPENDED. This is the same suspension that stops your apps when a balance stays at zero past the grace period.

Both states clear themselves: redeem a voucher and requests start succeeding again. No support ticket needed.