How provider routing and fallback keep your app online

Routing as a first-class concern

Every request flows through a routing layer that picks a healthy provider for the requested model. Static rules and live health checks both feed the decision.

Fallback that just works

When a provider times out or errors, the gateway retries the next candidate without your app noticing. Errors are classified as platform, upstream, user, or quota so retries stay safe.

Observability

Latency, success rate, and provider health are tracked per route so you can reason about reliability.