# =============================================================================
# ML Pipeline — Training Job + Inference Service
# =============================================================================

apiVersion: apps/v1
kind: Deployment
metadata:
  name: ml-inference
  namespace: ml-pipeline
  labels:
    app: ml-inference
    version: v1
spec:
  replicas: 1
  selector:
    matchLabels:
      app: ml-inference
  template:
    metadata:
      labels:
        app: ml-inference
        version: v1
      annotations:
        sidecar.istio.io/inject: "true"
    spec:
      serviceAccountName: ml-inference
      securityContext:
        runAsNonRoot: true
        runAsUser: 1000
        fsGroup: 1000
      containers:
        - name: inference
          image: "ecr.aws/devsecops/ml-inference:v1.0.0"
          ports:
            - containerPort: 8000
              protocol: TCP
          env:
            - name: MODEL_PATH
              value: "/models/latest"
            - name: HF_HOME
              value: "/cache/huggingface"
          resources:
            requests:
              cpu: "2"
              memory: 4Gi
              nvidia.com/gpu: "1"
            limits:
              cpu: "4"
              memory: 8Gi
              nvidia.com/gpu: "1"
          livenessProbe:
            httpGet:
              path: /health
              port: 8000
            initialDelaySeconds: 30
            periodSeconds: 30
          readinessProbe:
            httpGet:
              path: /ready
              port: 8000
            initialDelaySeconds: 10
            periodSeconds: 10
          volumeMounts:
            - name: model-storage
              mountPath: /models
            - name: huggingface-cache
              mountPath: /cache/huggingface
      volumes:
        - name: model-storage
          persistentVolumeClaim:
            claimName: model-pvc
        - name: huggingface-cache
          emptyDir:
            medium: Memory
            sizeLimit: 1Gi
      tolerations:
        - key: nvidia.com/gpu
          operator: Exists
          effect: NoSchedule
      nodeSelector:
        workload: ml
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: model-pvc
  namespace: ml-pipeline
spec:
  accessModes:
    - ReadWriteOnce
  storageClassName: gp3-encrypted
  resources:
    requests:
      storage: 50Gi
---
apiVersion: v1
kind: Service
metadata:
  name: ml-inference
  namespace: ml-pipeline
spec:
  selector:
    app: ml-inference
  ports:
    - port: 8000
      targetPort: 8000
---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: ml-inference
  namespace: ml-pipeline
---
# ML Training Job Template
apiVersion: batch/v1
kind: Job
metadata:
  name: ml-train-{{ .JobID }}
  namespace: ml-pipeline
spec:
  backoffLimit: 2
  ttlSecondsAfterFinished: 86400  # Clean up after 24h
  template:
    spec:
      serviceAccountName: ml-train
      securityContext:
        runAsNonRoot: true
        runAsUser: 1000
      containers:
        - name: trainer
          image: "ecr.aws/devsecops/ml-train:v1.0.0"
          command: ["python", "train.py"]
          env:
            - name: HF_TOKEN
              valueFrom:
                secretKeyRef:
                  name: hf-credentials
                  key: token
            - name: TRACKIO_URL
              value: "https://trackio.platform.internal"
          resources:
            requests:
              cpu: "4"
              memory: 16Gi
              nvidia.com/gpu: "1"
            limits:
              cpu: "8"
              memory: 32Gi
              nvidia.com/gpu: "1"
          volumeMounts:
            - name: training-data
              mountPath: /data
            - name: model-output
              mountPath: /output
      volumes:
        - name: training-data
          persistentVolumeClaim:
            claimName: training-data-pvc
        - name: model-output
          persistentVolumeClaim:
            claimName: model-output-pvc
      restartPolicy: Never
      tolerations:
        - key: nvidia.com/gpu
          operator: Exists
          effect: NoSchedule