apiVersion: apps/v1
kind: Deployment
metadata:
  name: vllm
spec:
  replicas: 1
  strategy: { type: Recreate }
  selector:
    matchLabels: { app: vllm }
  template:
    metadata:
      labels: { app: vllm }
    spec:
      enableServiceLinks: false
      nodeSelector:
        kubernetes.io/arch: arm64
      containers:
        - name: vllm
          image: vllm/vllm-openai-cpu:v0.27.0-arm64@sha256:d8ad98d4aa9ae41674b825d505184d55c2ef1f4a9c77e732e53c5caf9a9e5d51
          args:
            - Qwen/Qwen2.5-0.5B-Instruct
            - --served-model-name=local-chat
            - --dtype=float32
            - --max-model-len=2048
            - --max-num-seqs=2
          env:
            - name: VLLM_API_KEY
              valueFrom:
                secretKeyRef: { name: proxy-auth, key: api-key }
            - { name: VLLM_CPU_KVCACHE_SPACE, value: "1" }
            - { name: VLLM_CPU_OMP_THREADS_BIND, value: "0-3" }
            - { name: HF_HOME, value: /models }
          ports:
            - { name: http, containerPort: 8000 }
          resources:
            requests: { cpu: "2", memory: 3Gi }
            limits: { cpu: "4", memory: 6Gi }
          startupProbe:
            httpGet: { path: /health, port: http }
            periodSeconds: 10
            failureThreshold: 90
          readinessProbe:
            httpGet: { path: /health, port: http }
            periodSeconds: 10
          volumeMounts:
            - { name: models, mountPath: /models }
            - { name: shm, mountPath: /dev/shm }
      volumes:
        - name: models
          emptyDir: { sizeLimit: 5Gi }
        - name: shm
          emptyDir: { medium: Memory, sizeLimit: 1Gi }
---
apiVersion: v1
kind: Service
metadata:
  name: vllm
spec:
  selector: { app: vllm }
  ports:
    - { port: 8000, targetPort: http }