Routing

Use premium models by policy, not by accident.

Create aliases like coding-prod or demo-safe, then resolve every request through quota, provider health, cost caps and fallback strategy.

coding-prod policy
active
coding-prod
alias requested by clients
cost-aware policy
premium cap $8/hour
healthy provider
gemini then fallback
FAILOVER

Switch on 429/5xx

Fallback without losing the reason why the provider was skipped or retried.

WEIGHTED

Split traffic

Distribute requests across pools for reliability, capacity or cost control.

SCOPED

Key-specific access

Let teams use only the aliases and providers they are allowed to spend on.