dell-ai CLI and SDK

dell-ai provides a programmable interface to Dell Enterprise Hub (DEH). The CLI and Python SDK use the same underlying client, so interactive terminal workflows can be moved into notebooks, automation, and CI without changing the deployment model.

What dell-ai provides

Catalog discovery: search validated models, applications, and Dell platforms.
Deployment snippets: request the Docker, Kubernetes, or Helm configuration for an exact model and platform combination.
Local deployment: execute the validated snippet and manage ports and GPU allocation.
Goodput sizing: ask DEH to select an optimized configuration for a workload profile.
Deployment tracking: inspect endpoints and remove deployments through a local registry.
Environment management: store project-local or user-wide settings used by the CLI and SDK.

Discover models and platforms

# Browse the catalog
dell-ai models list --format table
dell-ai platforms list --format table
dell-ai apps list --format table

# Inspect compatibility and available container tags
dell-ai models compatible-platforms google/gemma-3-27b-it --format table
dell-ai models list-tags \
    --model-id google/gemma-3-27b-it \
    --platform-id xe9680-nvidia-h200

For gated models, dell-ai validates that the active Hugging Face token can access the repository before generating a deployment.

Generate a deployment snippet

Use get-snippet when you want to review, save, or customize the native command yourself:

dell-ai models get-snippet \
    --model-id google/gemma-3-27b-it \
    --platform-id xe9680-nvidia-h200 \
    --engine kubernetes \
    --gpus 8 \
    --replicas 1

Add --image-tag <tag> to pin a container version. The model deployment pages in this portal expose the same choices and can now switch between the native Docker or Kubernetes snippet and its corresponding dell-ai command.

Deploy and manage a model

Use deploy to fetch the snippet and execute it on the local node:

dell-ai models deploy \
    --model-id google/gemma-3-27b-it \
    --platform-id xe9680-nvidia-h200 \
    --engine docker \
    --gpus 8 \
    --replicas 1

Docker deployments are detached by default. dell-ai can remap a busy host port, allocate free NVIDIA or AMD GPUs, and record the resulting endpoint and container ID. Kubernetes deployments are applied through kubectl, so the corresponding local engine must already be installed and configured.

# Inspect active deployments, endpoints, and local system state
dell-ai status

# Remove one deployment
dell-ai models undeploy -d google/gemma-3-27b-it

Size by goodput scenario

Goodput scenarios let DEH select the GPU count and runtime configuration for a workload profile instead of requiring manual sizing. --goodput and --gpus are mutually exclusive.

dell-ai models goodput-scenarios --format table

dell-ai models deploy \
    --model-id google/gemma-3-27b-it \
    --platform-id xe9680-nvidia-h200 \
    --engine docker \
    --goodput balanced

See Goodput Scenarios for the workload profiles and SLO terminology used by Dell Enterprise Hub.

Use the Python SDK

The SDK mirrors the CLI and uses the authentication created by dell-ai login:

from dell_ai.client import DellAIClient

client = DellAIClient()

models = client.list_models()
platform = client.get_platform(platform_id="xe9680-nvidia-h200")

snippet = client.get_deployment_snippet(
    model_id="google/gemma-3-27b-it",
    platform_id=platform.id,
    engine="docker",
    num_gpus=8,
    num_replicas=1,
)
print(snippet)

result = client.deploy_model(
    model_id="google/gemma-3-27b-it",
    platform_id=platform.id,
    engine="docker",
    num_gpus=8,
    num_replicas=1,
    detach=True,
)
print(result["success"], result.get("endpoint"))

Environment variables and local state

dell-ai env stores configuration in two scopes:

Local project values live in .dell-ai-env.json in the current directory.
Global values live in ~/.config/dell-ai/env.json.

The active shell takes precedence over local values, which take precedence over global values.

dell-ai env set DELL_AI_ENDPOINT http://localhost:80
dell-ai env set DELL_AI_ENDPOINT http://localhost:80 --global
dell-ai env list
dell-ai env delete DELL_AI_ENDPOINT

Successful deployments are tracked in a similar local or global registry. This lets dell-ai status rediscover endpoints and lets dell-ai models undeploy tear them down later.

Documentation