Skip to content

provider-usage.md

Latest commit

聽

History

History
43 lines (34 loc) 路 2.2 KB

provider-usage.md

File metadata and controls

43 lines (34 loc) 路 2.2 KB
Raw

Provider usage

Use kit usage [provider] to inspect provider usage and quota without starting a model prompt:

kit usage
kit usage openai
kit usage openrouter --credential-store keychain

Supported providers are openai (also openai-subscription) and openrouter. Omit the provider to check authenticated supported providers. The command uses the effective credential storage from configuration and CLI overrides (--credential-store and --credential-dir). OpenRouter also uses the effective --openrouter-api-key or OPENROUTER_API_KEY override. Missing credentials for an explicitly selected provider or unavailable provider data are reported rather than treated as zero usage. Provider usage is account/key-level information, not the current session's accumulated token count.

In the TUI, /usage [provider] performs the same check in the background and shows local informational output. It does not send a model prompt.

OpenAI reports ChatGPT subscription/Codex quotas, not every ChatGPT product quota or API billing balance. Every reported main and additional rate-limit window shows its interval, percentage remaining, and provider-reported reset countdown and absolute UTC time. Fixed-duration windows are not calendar months. Missing quota information does not mean unlimited.

OpenRouter reports API-key spend in USD, including BYOK spend. The counters cover all time and the current UTC day, Monday鈥揝unday week, and calendar month. The key's remaining budget is reported directly; lifetime spend is not subtracted from a recurring cap. No key cap does not imply unlimited account credits. Budget recurrence is shown separately: a nonrecurring cap has no recurring reset, and an exact next budget reset timestamp is not reported by this endpoint.

Checks run concurrently with bounded HTTP timeouts and response sizes. Successful providers remain visible if another provider fails. Usage calls use the providers' canonical usage endpoints, not custom model API base URLs. OpenAI authentication may refresh through the existing login mechanism; no model requests are made.

Related documentation