Zorveus LogoZorveusDOCS
Resources

Frequently asked questions

View Markdown

Answers about credentials, product users, billing, Bring Your Own Key routing, latency, and stored usage data.

Does Zorveus add latency to AI inference?

Zorveus adds negligible gateway processing and network transit to the provider request. Streaming endpoints return Server-Sent Events in real time. Latency depends primarily on the upstream provider, selected model, and geographical region.


Can I mix BYOK keys and wallet-funded routing?

Yes. An eligible provider credential routes matching models through your upstream provider account. If auto_fallback_to_wallet is enabled and your organization wallet is funded, Zorveus automatically falls back to wallet-funded routing if your BYOK key encounters rate limits or quota exhaustion.


Does Bring Your Own Key (BYOK) mean inference is free?

No. Upstream model providers (OpenAI, Anthropic, Gemini) bill your credit card or invoice directly for model tokens.

Zorveus provides free BYOK routing for the first $25,000/month in traffic (sell_cost = $0.00). If your monthly BYOK volume exceeds $25k, Zorveus applies a volume platform fee to the incremental traffic above $25k, settled against your Zorveus wallet.


Are prompt inputs and completions stored?

Zorveus usage records contain accounting and observability fields (model, provider, token counts, cost, status, latency, and user attribution). Zorveus does not persist prompt inputs or completion text.


Which key calls inference?

Use an API key (zrv_...) against https://api.zorveus.com/v1 or provider-native wire endpoints. Scoped OAuth user tokens are also valid inference credentials.

Do not use an organization Service Key for inference. Service keys (zrv_svc_...) are for supported product-user routes under /product-users/by-external-id at https://api.zorveus.com.


Does an API-key cap give every customer a separate allowance?

No. An API key belongs to an app connection, and an app-connection cap limits aggregate usage across that key.

To enforce individual limits for each customer:

  • Use Product-User Spend Caps for monthly/daily spending caps per customer.
  • Use Product-User Credit Grants to issue one-time or promotional allowances to specific users.

Must I register product users in advance?

No. Simply pass metadata.external_user_id (e.g. usr_123) in your inference request payload or SDK client. Zorveus automatically provisions and tracks the product user on the fly.

Was this page helpful?
Edit this page on GitHub

On this page