Routing Logic

Routing logic combines historical merit signals with live operational data so each request can be assigned to the provider most likely to satisfy the current objective.

Request classification

Incoming requests are first mapped to a known workload category and policy context. That category determines which benchmark evidence and routing constraints are relevant to the decision.

Selection inputs

The router considers merit ranking, provider health, regional availability, policy constraints, and the request's latency or quality target. No single signal decides the outcome in isolation.

Policy evaluation

Policies define which providers are eligible, how scores are weighted, and when the router should prefer stability over raw benchmark rank. This gives teams a way to encode operational preferences without hardcoding single-provider routes.

Decision output

Each routing decision should produce enough metadata to explain why a provider was selected. That explanation becomes important when teams need to audit a request or investigate a divergence from expected behaviour.

Next step

Use Provider Health to understand one of the key live inputs to routing.