Invoices, reconciliation & disputes

This page explains how billing records are generated, how to reconcile your usage, and how to handle disputes.

Billing cycle

Routic bills based on actual token consumption, recorded per request. Here's how the pipeline works:

Request — You call the API with your key.
Gateway processing — The gateway routes to a model provider and logs the spend.
Usage sync — A sync job (every ~5 minutes) pulls gateway spend logs into usage_records.
Billing debit — A billing job (every ~1 hour) deducts the cost from your account balance.

This means:

Usage appears with a short delay — up to 5 minutes after the actual request.
Balance deduction is not instant — it happens within the next billing cycle (up to 1 hour).
Your dashboard shows near-real-time data — the usage and balance views refresh automatically.

The Routic console provides:

Use the usage endpoints to query your consumption programmatically. See Models, SKUs & usage for details.

Reconciliation means comparing what you expect to be charged against what was actually charged. This is important for:

Budget tracking — ensure spending stays within expected bounds.
Audit — verify that every request was billed correctly.
Dispute resolution — identify specific requests where the charge is incorrect.

Export your request logs — from your application, export logs containing request_id, model, and timestamp.
Compare with usage records — match each request_id against the usage data in your dashboard or via the API.
Check totals — sum up token counts and costs for a time period and compare with the billed amount.

Discrepancy	Likely cause	What to do
Missing usage record	Sync delay (up to 5 min) or the request failed before logging	Wait 5 minutes, then check again. If still missing, contact support.
Token count mismatch	Different tokenization between your count and the model provider	Token counts are determined by the model provider's tokenizer. Use the `usage` field in the response as the source of truth.
Cost seems wrong	Markup applied to the model provider's base price	The charge includes a markup over the model provider's cost. Check the cost comparison for pricing context.
Double charge	Retry was not idempotent — two requests were processed	The chat completions endpoint is not idempotent. Avoid retrying on non-rate-limit errors. See Errors & retries.

Electronic invoices are not yet available in the self-service portal. Invoice issuance follows these rules:

Invoice type	How to request
Local statutory invoice	Contact support with your billing period and company details.
USD receipt	Available in your dashboard under billing history.
Proforma / Commercial Invoice	Contact support for cross-border documentation.

The following invoice capabilities are planned:

If you believe a charge is incorrect:

Collect the evidence — note the request_id, timestamp, expected vs. actual charge.
Contact support — provide the details above. Include your account ID.
Resolution timeline — disputes are reviewed within 1–3 business days.

Common valid disputes:

You were charged for a request that returned a 500 error (tokens may not have been consumed).
The same request_id appears twice in your usage records.
The token count in the usage record doesn't match the usage field in the original response.