Dashboard layout

The dashboard provides an integrated web UI for accessing assigned Profile and Namespace resources and Modu AI serving features. From the dashboard, you can view and manage LLM serving with InferenceService, Notebook execution, training with TrainJob, volumes, and monitoring in one place.

Dashboard menu layout

Modu AI dashboard menu layout

Menu	Description
Home	Main dashboard page
Notebooks	Create and manage Jupyter Notebook and Code Server instances for development and debugging
Volumes	Manage volumes (PVCs) used by Notebook, training, and serving workloads
Endpoints	Manage KServe, LLM-D, and Dynamo serving endpoints with dedicated FQDNs
TrainJobs	Manage distributed training jobs based on Kubeflow Training Operator
Pipelines	Manage pipelines, experiments, runs, recurring runs, artifacts, and executions
Model Registry	Manage trained model metadata and versions
Monitoring	Monitor Namespace GPU, CPU, disk, and Pod usage in Grafana
Manage Account	Manage your user account, including password changes and SSH password verification
Manage Group Users	Manage users in the group Namespace. This menu is visible to users with Admin permissions.

Check Namespace

You can check the Namespace at the top of the dashboard.

Namespace selector

Field	Description
Namespace	kbm-g-np-<organization-name> — Workspace assigned through a Profile
GPU Quota	Maximum number of GPUs allocated to each organization. For example, 128 GPUs are equivalent to 16 nodes with 8 GPUs each.
managed label	Namespace integrated with Modu AI serving and Gateway features

Select your assigned Namespace before creating InferenceService, Notebook, or TrainJob resources.

Guides and notices

Guide links and Modu AI notices are available on the right side of the dashboard.

Guide cards on the right side of the Home page

Serving architecture (reference)

Client
  → Dedicated LB → Envoy Gateway
       → HTTPRoute → InferenceService (such as vLLM)
       → Pod (dedicated B200 node pool)

From Endpoints in the dashboard, check the status and URL for each endpoint type. Use the dedicated FQDN (*.moduai.kakaocloud.com) for API calls.