feat: vLLM in K8s via Harbor mirror, pipelines point to internal svc

- vLLM image mirrored to registry.celestium.life/stonks-oracle/vllm-openai
- Deployment uses Harbor image (Docker Hub IPv6 unreachable from cluster)
- All 3 pipelines use vllm-external.vllm-service.svc.cluster.local:2701
- K8s manifests at infra/kube-vllm/ and synced to ~/sources/kube/vllm

infra/kube-vllm/service.yaml

@@ -0,0 +1,29 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: vllm
+  namespace: vllm-service
+spec:
+  ports:
+    - name: http
+      port: 8000
+      targetPort: 8000
+  selector:
+    app: vllm
+  type: ClusterIP
+---
+# External access via NodePort (like ollama's port 2701 pattern)
+apiVersion: v1
+kind: Service
+metadata:
+  name: vllm-external
+  namespace: vllm-service
+spec:
+  ports:
+    - nodePort: 31508
+      name: vllm-api
+      port: 2701
+      targetPort: 8000
+  selector:
+    app: vllm-metrics
+  type: LoadBalancer