fix math routing

Signed-off-by: GitHub <[email protected]>

Author: samzong

e2e/profiles/llm-d/manifests/httproute-services.yaml

@@ -1,26 +1,5 @@
 apiVersion: gateway.networking.k8s.io/v1
 kind: HTTPRoute
-metadata:
-  name: vsr-default-svc
-  namespace: default
-spec:
-  parentRefs:
-    - group: gateway.networking.k8s.io
-      kind: Gateway
-      name: inference-gateway
-  rules:
-    - backendRefs:
-        - name: vllm-llama3-8b-instruct
-          port: 8000
-      matches:
-        - path:
-            type: PathPrefix
-            value: /
-      timeouts:
-        request: 300s
----
-apiVersion: gateway.networking.k8s.io/v1
-kind: HTTPRoute
 metadata:
   name: vsr-llama8b-svc
   namespace: default

e2e/profiles/llm-d/values.yaml

@@ -1,3 +1,4 @@
+# Profile revision 2025-11-21: math -> phi4-mini; cs/default -> llama3-8b; removed global HTTPRoute catch-all.
 config:
   # Allow Envoy to re-run route matching after Semantic Router sets x-selected-model.
   # Without this, Gateway API routes that depend on that header won't be chosen and return 404.
@@ -30,7 +31,7 @@ config:
           - type: domain
             name: math
       modelRefs:
-        - model: llama3-8b
+        - model: phi4-mini
           use_reasoning: false
     - name: cs_route
       priority: 10