Reference

Usage and quota

Use when: Usage history and quota checks
Updated: September 4, 2026

Usage and quota views show which clients, providers, models, and upstream accounts are consuming prompt tokens and how routing choices affect account health.

Context usage chart showing input, output, cache, and reasoning token trends across a day. — Usage views show prompt-token pressure over time before you inspect provider and account details.

What to watch

Prompt tokens

Track client-side limits and upstream account spending.

Context size

Watch large prompts that may trigger provider failures or client-limit rejections.

Account state

Disabled, cooling-down, failed, and exhausted accounts are excluded from normal routing.

Route selection

Compare expected priority or custom-model behavior against observed account usage.

Useful endpoints

Endpoint	Use
`GET /health`	Basic process health.
`GET /ready`	Readiness check for dependencies and route serving.
`GET /admin/account-routing`	Inspect priority-routing configuration.
`GET /api-docs/openapi.json`	Runtime OpenAPI JSON from the gateway app.

Investigate unexpected usage

1
Check the managed API key.
Confirm the key allows the resolved provider and account. A custom alias is authorized through its concrete targets, not as a separate scope entry.
2
Check priority routing.
Priority accounts should receive traffic before normal accounts while eligible.
3
Check fallback behavior.
Provider failures can move traffic to another target in a custom model.