feat(inference)serve Qwen3-8B directly through Envoy, drop Semantic Router

clusters/tinycloud/platform/ai/inference/helm-release.yaml → clusters/tinycloud/platform/inference/helm-release.yaml

@@ -27,32 +27,34 @@ spec:
     routerSpec:
       enableRouter: false
     servingEngineSpec:
+      strategy:
+        type: Recreate
       serviceMonitor:
         enabled: true
         additionalLabels:
           platform.tinycloud.dev/monitoring: prometheus
       modelSpec:
-        - name: qwen-instruct
+        - name: qwen3-8b
           repository: vllm/vllm-openai
           tag: v0.29.0
-          modelURL: Qwen/Qwen3-4B-AWQ
+          modelURL: Qwen/Qwen3-8B-AWQ
           requestCPU: 2
-          requestMemory: 4Gi
+          requestMemory: 6Gi
           requestGPU: 1
           requestGPUType: nvidia.com/gpu
           limitCPU: "4"
-          limitMemory: 8Gi
+          limitMemory: 10Gi
           storageClass: local-path
           pvcStorage: 10Gi
           vllmConfig:
             tensorParallelSize: 1
             maxModelLen: 8192
-            dtype: float16
+            dtype: bfloat16
             quantization: awq
-            gpuMemoryUtilization: 0.8
+            gpuMemoryUtilization: 0.9
             extraArgs:
               - --served-model-name
-              - qwen3-4b
+              - qwen3-8b
               - --enable-auto-tool-choice
               - --tool-call-parser
               - hermes

clusters/tinycloud/platform/ai/router/http-route.yaml → clusters/tinycloud/platform/inference/http-route.yaml

@@ -1,8 +1,8 @@
 apiVersion: gateway.networking.k8s.io/v1
 kind: HTTPRoute
 metadata:
-  name: semantic-router-api
-  namespace: semantic-router
+  name: vllm-openai
+  namespace: vllm
 spec:
   parentRefs:
     - group: gateway.networking.k8s.io
@@ -22,6 +22,5 @@ spec:
       backendRefs:
         - group: ""
           kind: Service
-          name: vllm-qwen-instruct-engine-service
-          namespace: vllm
+          name: vllm-qwen3-8b-engine-service
           port: 80

clusters/tinycloud/platform/ai/inference/kustomization.yaml → clusters/tinycloud/platform/inference/kustomization.yaml

@@ -4,4 +4,8 @@ resources:
   - namespace.yaml
   - helm-repository.yaml
   - helm-release.yaml
-  - reference-grant.yaml
+  - http-route.yaml
+labels:
+  - pairs:
+      app.kubernetes.io/part-of: inference
+    includeSelectors: false

clusters/tinycloud/platform/ai/router/namespace.yaml → clusters/tinycloud/platform/inference/namespace.yaml

@@ -1,6 +1,6 @@
 apiVersion: v1
 kind: Namespace
 metadata:
-  name: semantic-router
+  name: vllm
   labels:
     platform.tinycloud.dev/gateway-access: envoy

clusters/tinycloud/platform/kustomization.yaml

@@ -1,7 +1,7 @@
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization
 resources:
-  - ai
+  - inference
   - gateway
   - observability
 labels: