modelserving/k8s/llm-server.yaml

kind: Deployment
apiVersion: apps/v1
metadata:
  name: llm-server
spec:
  replicas: 1
  selector:
    matchLabels:
      app: llm-server
  template:
    metadata:
      labels:
        app: llm-server
    spec:
      serviceAccountName: llm-server
      containers:
      - name: llm-server
        image: llamacpp-gemma3-12b-it-cuda:latest # placeholder value, replaced by deployment scripts
        env:
        - name: LLAMA_ARG_N_GPU_LAYERS
          value: "99"
        - name: LLAMA_ARG_FLASH_ATTN
          value: "yes"
        args:
        - --jinja # Needed for tool use, no env var
        resources:
          limits:
            nvidia.com/gpu: "1"
          requests:
            nvidia.com/gpu: "1"
      nodeSelector:
        cloud.google.com/gke-accelerator: nvidia-l4

---

kind: ServiceAccount
apiVersion: v1
metadata:
  name: llm-server

---

kind: Service
apiVersion: v1
metadata:
  name: llm-server
  labels:
    app: llm-server
spec:
  selector:
    app: llm-server
  ports:
  - port: 80
    targetPort: 8080
    protocol: TCP