What direct, apim, and indirect mean, and when to use each mode.
A direct backend uses mode=direct. The proxy does not send active health probes to it and always includes it in the active host set. Real request failures are still recorded by the circuit breaker.
Host_<name>="host=https://model.example.com;mode=direct;path=/model"Use it when probing would be unsafe or undesirable—for example, when a serverless target scales to zero or has no suitable probe endpoint. Because direct mode has no probe-derived latency, it sorts first when LoadBalanceMode=latency.
Use APIM in the backend path when requests need gateway policies, transformations, subscriptions, caller authentication, or priority-aware selection across the services behind APIM. This adds APIM as an operational dependency, so use it for capabilities the direct path does not provide.
An APIM backend points a Host_<name> entry at Azure API Management. mode=apim is standard non-direct behavior: the proxy sends the configured probe on every PollInterval, recording both a rolling success rate and latency on each successful probe. It can remove APIM from the active set when health falls below the required success rate, and uses the recorded latency to order hosts when LoadBalanceMode=latency.
Host_<name>="host=https://gateway.azure-api.net;mode=apim;path=/shared;probe=/health"An indirect backend is endpoint metadata that APIM may select. The proxy never calls or probes it; instead, via names the APIM gateway that receives the request. Indirect backends must be referenced by a named Path_* route.
Host_apim="host=https://gateway.azure-api.net;mode=apim;probe=/health"
Host_ptu="host=https://ptu.openai.azure.com;mode=indirect;via=Host_apim"
Path_models="prefix=/models;hosts=Host_ptu;stripprefix=false"mode=indirect requires via, and the via target must use mode=apim. Existing APIM hosts without via remain valid and continue to work as opaque gateways.
APIM can supply API gateway capabilities such as caller authentication, subscriptions, transformations, and priority-aware backend policies. The proxy adds its own queue, worker controls, health tracking, circuit breaking, and telemetry around that gateway path.
See Backend Host Configuration for all host options.