# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0

apiVersion: nvidia.com/v1alpha1
kind: DynamoGraphDeployment
metadata:
  name: vllm-agg-lmcache
spec:
  services:
    Frontend:
      envFromSecret: hf-token-secret
      componentType: frontend
      replicas: 1
      extraPodSpec:
        mainContainer:
          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:my-tag
    VllmDecodeWorker:
      envFromSecret: hf-token-secret
      componentType: worker
      replicas: 1
      sharedMemory:
        disabled: true
      resources:
        limits:
          gpu: "1"
        requests:
          custom:
            # Increase this value for larger models
            ephemeral-storage: "2Gi"
      extraPodSpec:
        mainContainer:
          image: nvcr.io/nvidia/ai-dynamo/vllm-runtime:my-tag
          workingDir: /workspace/examples/backends/vllm
          env:
            - name: PYTHONHASHSEED
              value: "0"
          volumeMounts:
            - name: kv-transfer-config
              mountPath: /etc/lmcache
              readOnly: true
          command:
            - /bin/sh
            - -c
            - exec python3 -m dynamo.vllm "$@" --kv-transfer-config "$(cat /etc/lmcache/kv-transfer-config.json)"
            - --
          args:
            - --model
            - Qwen/Qwen3-0.6B
            - --disable-hybrid-kv-cache-manager
        volumes:
          - name: kv-transfer-config
            configMap:
              name: lmcache-mp-connection   # = <LMCacheEngine.metadata.name>-connection

        # Required for CUDA IPC between vLLM and LMCache
        hostIPC: true