clean up and add routes to paas site

apps/base/customer1/openclaw/kustomization.yaml

@@ -1,14 +1,10 @@
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization
 resources:
- # - vllm-l4.yaml
-  - keda-a100.yaml
   - pvc.yaml
   - new-configmap.yaml
  # - deployment.yaml
   - service.yaml
   - openclaw-secrets.yaml
-  - pg-cluster-openclaw.yaml
-  #- a100-vllm.yaml 
+  #- pg-cluster-openclaw.yaml
   - xai-apikey.yaml
-  - gateway-routes/route.yaml

clusters/devops-lab/infra-gatewayapi.yaml

@@ -0,0 +1,19 @@
+
+apiVersion: kustomize.toolkit.fluxcd.io/v1
+kind: Kustomization
+metadata:
+  name: infrastructure-gatewayapi   
+  namespace: flux-system
+spec:
+  interval: 1m0s 
+  retryInterval: 1m
+  timeout: 5m
+  sourceRef:
+    kind: GitRepository
+    name: flux-system
+  path: ./infrastructure/gatewayapi 
+  prune: true
+  decryption:
+    provider: sops
+    secretRef:
+      name: sops-age 

clusters/devops-lab/infra-gpus.yaml

@@ -0,0 +1,19 @@
+
+apiVersion: kustomize.toolkit.fluxcd.io/v1
+kind: Kustomization
+metadata:
+  name: infrastructure-gpus  
+  namespace: flux-system
+spec:
+  interval: 1m0s 
+  retryInterval: 1m
+  timeout: 5m
+  sourceRef:
+    kind: GitRepository
+    name: flux-system
+  path: ./infrastructure/gpus/staging  
+  prune: true
+  decryption:
+    provider: sops
+    secretRef:
+      name: sops-age 

infrastructure/gatewayapi/apigateway.yaml

@@ -6,13 +6,21 @@ metadata:
 spec:
   gatewayClassName: gke-l7-global-external-managed  # Use gke-l7-global-external-http for external, or gke-l7-rilb for internal
   listeners:
-  - name: https 
+  - name: n8n-https 
     protocol: HTTPS 
     port: 443
     tls:
       mode: Terminate
       options:
         networking.gke.io/pre-shared-certs: n8n-sirius-sec-com
+  - name: siriusdevops-https
+    hostname: siriusdevops.com
+    protocol: HTTPS 
+    port: 443
+    tls:
+      mode: Terminate
+      options:
+        networking.gke.io/pre-shared-certs: siriusdevops-cert
     allowedRoutes:
       namespaces:
         from: "All" 

infrastructure/gatewayapi/gateway-routes/route.yaml

@@ -0,0 +1,18 @@
+apiVersion: gateway.networking.k8s.io/v1
+kind: HTTPRoute
+metadata:
+  name: paas-landing 
+  namespace: customer1
+spec:
+  parentRefs:
+  - name: external-http-gateway 
+    kind: Gateway
+  rules:
+  - matches:
+    - path:
+        type: PathPrefix
+        value: /paas 
+    backendRefs:
+    - name: paas-landing-svc  
+      port: 80  
+  

infrastructure/gatewayapi/kustomization.yaml

@@ -0,0 +1,6 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - apigateway.yaml
+  - gateway-routes/
+

infrastructure/gpus/base/keda-gpu-scaling/kustomization.yaml

@@ -0,0 +1,6 @@
+
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - keda-vllm.yaml
+

infrastructure/gpus/base/vllm-servers/kustomization.yaml

@@ -0,0 +1,6 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  #- a100-vllm.yaml
+  - rtx600-vllm.yaml
+  #- vllm-l4.yaml

infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml

@@ -0,0 +1,129 @@
+
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: openclaw-brain-vllm
+  namespace: customer1
+  labels:
+    app: openclaw-brain
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: openclaw-brain
+  template:
+    metadata:
+      labels:
+        app: openclaw-brain
+    spec:
+      nodeSelector:
+        cloud.google.com/gke-accelerator: "nvidia-a100-80gb"
+      tolerations:
+        - key: "nvidia.com/gpu-nvidia-rtx-pro-6000"
+          operator: "Equal"
+          value: "present"
+          effect: "NoSchedule"
+      containers:
+        - name: vllm-brain
+          image: vllm/vllm-openai:latest
+          command: ["python3", "-m", "vllm.entrypoints.openai.api_server"]
+          env:
+            - name: HF_TOKEN
+              value: ""
+            - name: TOKENIZER_MODE
+              value: "hf"
+          args:
+            - --model=Youssofal/Qwen3.6-27B-Abliterated-Heretic-Uncensored-BF16
+            - --host=0.0.0.0
+            - --port=8000
+            - --tensor-parallel-size=1
+            - --tokenizer-mode=hf 
+            - --gpu-memory-utilization=0.95
+            - --kv-cache-dtype=fp8
+            - --max-model-len=131072 
+            - --enable-auto-tool-choice 
+            - --enable-chunked-prefill
+            - --max-num-batched-tokens=8192
+            - --max-num-seqs=4
+            - --trust-remote-code
+            - --dtype=auto 
+            - --enable-prefix-caching
+            - --tool-call-parser=qwen3_xml
+            - --reasoning-parser=qwen3
+            - --disable-custom-all-reduce
+          ports:
+            - containerPort: 8000
+          resources:
+            limits:
+              nvidia.com/gpu: 1
+              memory: "100Gi"
+              cpu: "16"
+            requests:
+              nvidia.com/gpu: 1
+              memory: "80Gi"
+              cpu: "8"
+          
+          startupProbe:
+            httpGet:
+              path: /health
+              port: 8000
+            initialDelaySeconds: 30
+            periodSeconds: 10
+            timeoutSeconds: 10
+            failureThreshold: 60
+            successThreshold: 1 
+          
+          livenessProbe:
+            httpGet:
+              path: /health
+              port: 8000 
+            initialDelaySeconds: 180
+            periodSeconds: 15
+            failureThreshold: 5
+            successThreshold: 1 
+            timeoutSeconds: 10 
+            
+          
+          readinessProbe:
+            httpGet:
+              path: /v1/models 
+              port: 8000
+            initialDelaySeconds: 120
+            periodSeconds: 10 
+            timeoutSeconds: 15 
+            failureThreshold: 8
+            successThreshold: 1 
+          
+          volumeMounts:
+            - name: model-cache
+              mountPath: /root/.cache/huggingface
+      volumes:
+        - name: model-cache
+          persistentVolumeClaim:
+            claimName: vllm-model-qwen3.6-27b-uncensored
+---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: vllm-model-qwen3.6-27b-uncensored
+  namespace: customer1
+spec:
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 100Gi
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: openclaw-brain-service
+  namespace: customer1
+spec:
+  selector:
+    app: openclaw-brain
+  ports:
+    - protocol: TCP
+      port: 8000
+      targetPort: 8000
+  type: ClusterIP

infrastructure/gpus/staging/kustomization.yaml

@@ -0,0 +1,5 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - ../base/vllm-servers/
+  - ../base/keda-gpu-scaling/ 

modules/pro6000-nodepool.tf

@@ -28,7 +28,7 @@ resource "google_container_node_pool" "pro6000_pool" {
     }
 
     # SPOT is the modern equivalent of Preemptible
-    spot = false 
+    spot = true 
 
     # Specific labels to help the Autoscaler find this pool
     labels = {

plans/AI_ARCHITECTURE.md

@@ -13,7 +13,7 @@ Instead of running an expensive A100 GPU 24/7 for all requests, we split the cog
 
 ## Tier 2: The Deep Thinker (A100 GPU)
 - **Hardware:** 1x NVIDIA A100 (80GB VRAM)
-- **Model:** `coder3101/Qwen3.5-27B-heretic`
+- **Model:** `Qwen3.6-27B-heretic`
 - **Status:** Scaled to zero by default.
 - **Role:** Activated only for massive context tasks, deep research, and complex multi-file architectural reasoning.