Dashboard
The first screen after sign-in, summarizing the tenant's key metrics, recent alerts, model status, GPUs, devices and deployments.
Required permission: View
This is the first screen after sign-in. It summarizes the selected tenant in six cards and reloads every 30 seconds. Every member of the tenant sees it, and nothing changes here. To act, follow the link at the top right of a card (View all (전체 보기) and similar) to the relevant screen.
| # | Area | Description |
|---|---|---|
| ① | Key metrics (KPI) (핵심 지표(KPI)) | Four tenant KPI tiles — data latency, automation (automation rate), API latency (API response p95), availability. Under each value is the target; if the value misses it, the value turns red and Below target (미달) appears. Without a target, "No target set (목표 미설정)" is shown. The system administrator sets the targets in the tenant settings. |
| ② | Recent alerts (최근 경보) | Up to 5 alerts that are not handled yet (active). Click a title to open the alert detail. |
| ③ | Model status (모델 상태) | For each of Production, Staging and Archived, the number of registered models that have at least one version in that stage (not the number of versions). Registry (레지스트리) opens the model registry. |
| ④ | GPU slots (GPU 슬롯) | GPU slots wired to this tenant. With dedicated (전용) wiring it shows always-allocated, in use and idle; with shared (공유) wiring it shows the pool name, pool slots, pool usage and idle (for the whole pool) plus this tenant's share. If Kubernetes is not connected, only the wired values appear and usage and idle show "—". If nothing is wired, "No GPU is wired (GPU 가 배선되지 않았습니다)" appears and training and serving run on CPU only. |
| ⑤ | Edge Fleet | All registered field devices and their status badges. Click a device name to open the device detail. |
| ⑥ | Recent deployments (최근 배포) | Model, target, status and time of up to 10 recent deployments. Click the model to open the deployment detail. |
| ⑦ | Tenant indicator | The tenant you are looking at. If you belong to several tenants, switch here. |
Only cards for features enabled in the tenant are shown. A tenant with the alerts feature turned off has no card ②. If one card fails to load, only that card shows "Failed to load the widget (위젯을 불러오지 못했습니다)" and the others stay visible.
Main tasks
- If there are alerts to handle, click an alert title in ② to go to the alert detail.
- To see model stages in detail, press Registry (레지스트리) in ③.
- If a device is unhealthy, use View all (전체 보기) in ⑤ to go to Edge Fleet and open the device.
Statuses and badges
| Card | Badge | Meaning |
|---|---|---|
| Recent alerts | CRITICAL · WARNING · INFO | Alert severity. Hover to see it spelled out as critical, warning, info (심각·경고·정보). |
| Model status | Production · Staging · Archived | Stage of the model version. See the stage table in Model registry. |
| Edge Fleet | Pending · Healthy · Unhealthy · Uninstalled (연결 대기 · 정상 · 비정상 · 미설치) | Device status. See the status table in Edge Fleet. |
| Recent deployments | ACTIVE · IN_PROGRESS · FAILED · ROLLED_BACK | Deployment status. See the status table in Deployments. |