This section covers the layout, buttons and badges of each workspace screen (the everyday menu in the left sidebar). For procedures, see the tutorial. The Administration chapter covers the admin console (members, tokens, training runtimes) and the system console (tenants, cluster).

Common to all screens

  • Header: on the left, where you are (breadcrumbs); on the right, the tenant indicator, the bell icon (inbox) and your avatar (account settings, sign out). If you belong to several tenants, the tenant indicator becomes a select box that switches tenants.
  • Screen help: press the ? next to the title to see "What you can do on this screen" and "Where it sits in the model operations flow". The small ? in a table header explains the term.
  • Permissions: buttons for actions you lack permission for are not greyed out — they are not shown at all. If a button is missing, check your role first. A few places show a read-only view or a notice instead of the buttons (drift and device policies, point label reading, the inference console, approval detail). A greyed-out (disabled) button is unavailable because of the current state, not your permissions. A menu you cannot view (for example Approvals (승인)) does not appear in the sidebar, and its address shows an "Access denied (접근 권한이 없습니다)" screen. The server makes the final decision: even if you type the address directly, it refuses the action.
  • Disabled features: a module that is turned off for the tenant disappears from the sidebar, and if you go to its address you see a "This feature is disabled (꺼진 기능입니다)" screen. To turn a feature on, ask the system administrator.

Screen map

In menu order. "View permission" is what you need to see the screen. Each page lists the permissions for its actions.

MenuSummaryView permissionMain action permissions
DashboardSummary of the tenant's key metrics, alerts, models and devicesEveryone—
ExperimentsMLflow experiments and run records, run comparison and detailEveryoneRegister model model_register, stop or delete a run, delete an experiment develop
TrainingTraining run list and detail, pretrained weightsEveryoneNew training and weights develop, register model model_register, delete a run operate
New training wizardConfigure and start a training run in 11 stepsdevelop—
Model registryRegistered models, versions and stages, promotion requests, serving image buildsEveryoneRequest promotion production_promote, archive staging_promote, delete model_register, build image deploy
Container registryRepositories, tags and images of the built-in registry, image importEveryoneImport and delete develop
PipelinesRun the Train, Register, Deploy and CT pipelinesEveryoneTrigger, rerun, stop operate
DeploymentsField deployment history of model versions, and rollbackEveryoneNew deployment deploy, rollback rollback
API servicesCentral serving endpoints and the inference consoleEveryoneCreate and delete deploy, stop and call operate
MonitoringInference metrics, training queue and GPU trends, model performance and input distributionEveryone—
AlertsAlerts the team handles together, silences, rules, channelsEveryoneAcknowledge, resolve, silence operate; rules and channels settings
ApprovalsApprove or reject Production promotion requestsapproveApprove and reject approve
DatasetsDatasets and files, previews (images, PLY 3D, time series)EveryoneCreate, upload, delete dataset_write
Edge FleetList of field devices and six tabs per deviceEveryoneRestart, Offline, commands operate; registration, policy, tokens settings
Collected dataFiles, records and inference results from the devices, across the tenantEveryoneDeleting files dataset_write
DriftInput distribution change per model (PSI, KS) and policyEveryoneSave policy settings
Inbox (bell icon)Personal notifications sent to youEveryone—
Account settings (avatar)Name, contact, password, notification preferencesEveryone—

Roles and permissions

Permissions come from your role in the tenant. One person can have a different role in each tenant.

PermissionVIEWERDEVELOPEROPERATORAPPROVERADMIN
view View●●●●●
develop Develop●●●
model_register Register models●●●
staging_promote Promote to Staging●●●
dataset_write Write datasets●●●
production_promote Promote to Production●●
approve Approve●●
deploy Deploy●●
operate Operate●●
rollback Roll back●
settings Settings●

Written for the platform as of 2026-10-02.

© MLOps Central Platform