Problem
The documentation for setting up InferencePool support does not mention that inferenceExtension.enabled=true must be set in the agentgateway Helm chart.
Without this flag, the controller uses an empty static collection for InferencePools (collection.go), so any HTTPRoute with an InferencePool backendRef will always return BackendNotFound and routing will fail with backends does not exist.
Expected
The docs should mention:
```bash
helm install agentgateway ...
--set inferenceExtension.enabled=true
```
Impact
Users following the current docs to set up InferencePool routing will hit a silent failure with no clear error pointing to this missing flag.
References
Problem
The documentation for setting up InferencePool support does not mention that
inferenceExtension.enabled=truemust be set in the agentgateway Helm chart.Without this flag, the controller uses an empty static collection for InferencePools (
collection.go), so any HTTPRoute with an InferencePool backendRef will always returnBackendNotFoundand routing will fail withbackends does not exist.Expected
The docs should mention:
```bash
helm install agentgateway ...
--set inferenceExtension.enabled=true
```
Impact
Users following the current docs to set up InferencePool routing will hit a silent failure with no clear error pointing to this missing flag.
References
collection.goline 189 — InferencePools only watched whenEnableInferExt=true