apiVersion: apps/v1
kind: Deployment
metadata:
  name: decode
spec:
  template:
    spec:
      containers:
        - name: modelserver
          command: ["vllm", "serve"]
          args:
            - "Qwen/Qwen3-32B"
            - "--disable-access-log-for-endpoints=/health,/metrics,/v1/models"
          env:
            - name: HF_TOKEN
              valueFrom:
                secretKeyRef:
                  name: llm-d-hf-token
                  key: HF_TOKEN
            - name: VLLM_ALLOW_RUNTIME_LORA_UPDATING
              value: "True"
            - name: VLLM_PLUGINS
              value: "lora_filesystem_resolver"
            - name: VLLM_LORA_RESOLVER_CACHE_DIR
              value: "/adapters"
          volumeMounts:
            - mountPath: /adapters
              name: lora-adapters
            - mountPath: /dev/shm
              name: shm
          resources:
            limits:
              cpu: '8'
              memory: 64Gi
              nvidia.com/gpu: 1
            requests:
              cpu: '4'
              memory: 32Gi
              nvidia.com/gpu: 1
      volumes:
        - name: lora-adapters
          emptyDir: {}
        - name: shm
          emptyDir:
            medium: Memory
            sizeLimit: 10Gi