> For clean Markdown content of this page, append .md to this URL. For the complete documentation index, see https://docs.nvidia.com/dynamo/llms.txt. For full content including API reference and SDK examples, see https://docs.nvidia.com/dynamo/llms-full.txt.

# Install

The Dynamo Enterprise install path parallels the open-source install path. It uses the same Kubernetes shape, the same values, and the same operational surface. The substantive differences are the container image references and the Helm chart name.

For the full walkthrough of cluster prerequisites, accelerator support, optional multinode orchestration, RDMA, Prometheus, and shared model storage, follow the [Kubernetes Installation Guide](/dynamo/dev/kubernetes/installation/install-dynamo). The steps below cover only what changes for the enterprise images and chart.

## Container Images

The enterprise images are public. Anonymous `docker pull` works from any host with network access to `nvcr.io`. Environments that require authenticated pulls (a bastion registry, an air-gapped mirror, or a policy that blocks anonymous access to `nvcr.io`) can configure an NGC API key on the pulling host; see the [NGC User Guide](https://docs.nvidia.com/ngc/ngc-catalog-user-guide/index.html) for the API-key workflow.

```bash
docker pull nvcr.io/nvidia/ai-dynamo/sglang-runtime-enterprise:1.4.2
docker pull nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime-enterprise:1.4.2
docker pull nvcr.io/nvidia/ai-dynamo/vllm-runtime-enterprise:1.4.2
docker pull nvcr.io/nvidia/ai-dynamo/dynamo-frontend-enterprise:1.4.2
docker pull nvcr.io/nvidia/ai-dynamo/kubernetes-operator-enterprise:1.4.2
```

Kubernetes clusters that already pull from `nvcr.io` need no additional pull secret for the enterprise images. Air-gapped and internal-mirror installs should retag these into the internal registry using the same repository names and versions to keep the manifests portable.

## Platform Helm Chart

The enterprise Helm chart installs the Dynamo Operator, using the enterprise Operator image. etcd, NATS, KAI Scheduler, Grove, and Snapshot ship as bundled subcharts that are all disabled by default; enable the ones a deployment needs, or point the operator at instances you already run. The chart shape, its values, and its Kubernetes surface match the open-source `dynamo-platform` chart. The differences are the chart name and the default image references the chart resolves to.

```bash
export NAMESPACE=dynamo-system
export RELEASE_VERSION=1.4.2

helm fetch https://helm.ngc.nvidia.com/nvidia/ai-dynamo/charts/dynamo-platform-enterprise-$RELEASE_VERSION.tgz
helm install dynamo-platform dynamo-platform-enterprise-$RELEASE_VERSION.tgz \
  --namespace $NAMESPACE \
  --create-namespace
```

Verify the operator:

```bash
kubectl get crd | grep dynamo
kubectl get pods -n $NAMESPACE
```

The expected CRDs and pods are the same as in the open-source install; see the [Kubernetes Installation Guide](/dynamo/dev/kubernetes/installation/install-dynamo) for the full verification and troubleshooting steps, and for the values that enable the optional subcharts.

## Runtime Images in Deployments

The Dynamo Operator watches `DynamoGraphDeployment` (DGD) resources. Deployments that must run on the enterprise runtime images reference them by their full `-enterprise` image path in the DGD spec. A minimal example:

```yaml
apiVersion: nvidia.com/v1beta1
kind: DynamoGraphDeployment
metadata:
  name: my-agg-deployment
  namespace: dynamo-system
spec:
  components:
    - name: Frontend
      type: frontend
      replicas: 1
      podTemplate:
        spec:
          containers:
            - name: main
              image: nvcr.io/nvidia/ai-dynamo/dynamo-frontend-enterprise:1.4.2
    - name: VllmWorker
      type: worker
      replicas: 1
      podTemplate:
        spec:
          containers:
            - name: main
              image: nvcr.io/nvidia/ai-dynamo/vllm-runtime-enterprise:1.4.2
              envFrom:
                - secretRef:
                    name: hf-token-secret
              command:
                - python3
                - -m
                - dynamo.vllm
              args:
                - --model
                - your-org/your-model
```

For the full DGD walkthrough (choosing a model, sizing parallelism, applying and testing the deployment), see [Deploy with DGD](/dynamo/dev/kubernetes/model-deployment/deploy-with-dgd).

## Differences from the Open-Source Install

* **Image references.** Enterprise and open-source images both resolve under `nvcr.io/nvidia/ai-dynamo/`. The enterprise repository names carry an `-enterprise` suffix, which is the only difference.
* **Helm chart.** Enterprise chart: `dynamo-platform-enterprise`. Open-source chart: `dynamo-platform`. The values schema is the same.
* **Support scope.** Only the images and chart listed above are eligible for commercial support. The Python wheels, Rust crates, and non-listed component containers in the same release remain community-supported.