apiVersion: v1
kind: Namespace
metadata:
  name: inference
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: mock-vllm
  namespace: inference
  labels:
    app: mock-vllm
spec:
  replicas: 1            # KEDA owns this from here — see scaledobject.yaml
  selector:
    matchLabels:
      app: mock-vllm
  template:
    metadata:
      labels:
        app: mock-vllm
    spec:
      containers:
        - name: mock-vllm
          image: mock-vllm:local
          imagePullPolicy: IfNotPresent     # image is side-loaded via `kind load`
          ports:
            - name: http
              containerPort: 8000
          env:
            - name: CAPACITY            # concurrent "decode slots" per replica
              value: "4"
            - name: GEN_TOKENS
              value: "200"
            - name: TPOT_SECONDS        # time per output token
              value: "0.05"
            - name: PREFILL_SECONDS
              value: "0.30"
          readinessProbe:
            httpGet:
              path: /health
              port: http
          resources:
            requests:
              cpu: 100m
              memory: 128Mi
            limits:
              cpu: "500m"
              memory: 256Mi