# Qwen3.5-4B (MTP, UD-Q4_K_XL) served by llama.cpp for low-latency assistants.
# The model is fully offloaded to the NUCBox Radeon 8060S. Model weights are
# stored on the shared llamacpp-models hostPath PVC.
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: llamacpp-qwen35-4b
  namespace: llamacpp
  labels:
    app: llamacpp
    model: qwen3.5-4b
spec:
  replicas: 1
  strategy:
    type: Recreate
  selector:
    matchLabels:
      app: llamacpp
      model: qwen3.5-4b
  template:
    metadata:
      labels:
        app: llamacpp
        model: qwen3.5-4b
    spec:
      nodeSelector:
        kubernetes.io/arch: amd64
        hardware: high-memory
      initContainers:
      - name: fetch-model
        image: alpine:3.20
        command: ["/bin/sh", "-c"]
        args:
        - |
          set -e
          if [ -s "/models/$MODEL_FILE" ]; then
            echo "Model $MODEL_FILE already present — skipping download."
            exit 0
          fi
          apk add --no-cache curl
          echo "Downloading $MODEL_FILE from $MODEL_URL ..."
          curl -fL --retry 5 --retry-delay 5 -o "/models/$MODEL_FILE.partial" "$MODEL_URL"
          mv "/models/$MODEL_FILE.partial" "/models/$MODEL_FILE"
          echo "Download complete: $(ls -lh /models/$MODEL_FILE)"
        env:
        - name: MODEL_URL
          value: "https://huggingface.co/unsloth/Qwen3.5-4B-MTP-GGUF/resolve/main/Qwen3.5-4B-UD-Q4_K_XL.gguf"
        - name: MODEL_FILE
          value: "Qwen3.5-4B-UD-Q4_K_XL.gguf"
        volumeMounts:
        - name: models
          mountPath: /models
      containers:
      - name: llama-server
        image: ghcr.io/ggml-org/llama.cpp:server-vulkan
        imagePullPolicy: IfNotPresent
        args:
        - -m
        - /models/Qwen3.5-4B-UD-Q4_K_XL.gguf
        - --alias
        - qwen3.5-4b
        - --host
        - 0.0.0.0
        - --port
        - "8080"
        - --jinja
        - -ngl
        - "999"
        - -c
        - "65536"
        - -np
        - "1"
        - --cont-batching
        - --cache-type-k
        - q8_0
        - --cache-type-v
        - q8_0
        - --threads
        - "8"
        ports:
        - name: http
          containerPort: 8080
        resources:
          requests:
            cpu: "1000m"
            memory: 2Gi
          limits:
            cpu: "4000m"
            memory: 8Gi
        readinessProbe:
          httpGet:
            path: /health
            port: 8080
          initialDelaySeconds: 30
          periodSeconds: 10
          failureThreshold: 6
        livenessProbe:
          httpGet:
            path: /health
            port: 8080
          initialDelaySeconds: 120
          periodSeconds: 30
          failureThreshold: 5
        securityContext:
          privileged: true
        volumeMounts:
        - name: models
          mountPath: /models
          readOnly: true
        - name: dri
          mountPath: /dev/dri
      volumes:
      - name: models
        persistentVolumeClaim:
          claimName: llamacpp-models
      - name: dri
        hostPath:
          path: /dev/dri
          type: Directory
---
apiVersion: v1
kind: Service
metadata:
  name: llamacpp-qwen35-4b
  namespace: llamacpp
  labels:
    app: llamacpp
    model: qwen3.5-4b
spec:
  type: ClusterIP
  selector:
    app: llamacpp
    model: qwen3.5-4b
  ports:
  - name: http
    port: 80
    targetPort: 8080