Watchers and Notifications
A Watcher is automation without a screen. Write the condition you want monitored as a plain Korean sentence — something like "Tell me when line 3 pressure goes above 90" — and Studio checks the field data on a set interval by itself, notifying you only when the condition is met.
Open it from the radar icon in the left sidebar (fa-radar, Watchers). If there are unread notifications, a red count badge appears on the icon (9+ for 10 or more).
The screen is split into two tabs.
| Tab | What it does |
|---|---|
| Notifications | List of alerts triggered so far (default tab) |
| Settings | Create watchers, enable/disable, run immediately, delete |
Creating a Watcher
Create one under [Settings] tab → New Watcher.
1. Get a draft from a single natural-language line
Type your monitoring condition as one sentence in the top input field (예: 매시간 3번 라인 압력이 90 넘으면 알려줘) and click AI Fill (fa-wand-magic-sparkles). The Name, Instructions, and Run Interval (minutes) fields below are filled in automatically. Pressing Enter works too.
Here are sentence examples commonly used in the field.
용해로 1호기 온도가 850도를 넘으면 알려줘
3번 라인 설비가 10분 넘게 정지 상태면 경고해줘
지난 1시간 CRITICAL 알람이 5건 이상 발생하면 알려줘
2공장 압축기 가동률(OEE)이 80% 밑으로 떨어지면 매시간 알려줘
보일러 급수 유량 태그가 30분 동안 값이 안 들어오면 알려줘
2. Review and create
You can freely edit the auto-filled values.
- Name — The title shown on notification cards and in webhook messages. Choose a name your field staff will recognize.
- Run Interval (minutes) — 1 to 1440 minutes. Default is 60.
- Instructions — A description of the task the agent performs each run (
예: 3번 라인 OEE를 확인해서 80% 미만이면 경고. /oee?site_id=S1 조회.). The more specifically you state which equipment and tags to compare against which thresholds, the fewer false positives. - Webhook URL (optional · Slack/Teams) — See Sending to a webhook below. Leave it empty to show notifications in Studio only.
Click Create and the watcher is registered as active immediately, running automatically from the next interval.
For equipment stoppages and safety-related conditions, use 5–10 minutes; for process values such as temperature and pressure, 15–30 minutes; for slowly changing indicators such as OEE and production output, 60 minutes or more. Shorter intervals increase AI usage.
3. Create directly from chat
You don't have to open the Watchers screen. Saying this in Chat also creates a watcher.
3번 라인 압력이 8bar 넘으면 알려줘
용해로 온도 15분마다 감시해줘
The agent first locates the target equipment and tags, confirms their IDs, then creates the watcher and tells you in its reply what it monitors and how often. The new watcher appears in the [Settings] tab list right away.
Receiving notifications
When a condition is met, you are notified in three places at once.
- Live toast — If Studio is open, a
🛰 와처이름 — 내용notification appears at the bottom right of the screen for about 6 seconds. - Nav badge — Unread count on the sidebar Watchers icon.
- [Notifications] tab — The full trigger history. Warning icon, watcher name, time of occurrence, and content stack up as cards.
Opening the [Notifications] tab marks all notifications up to that point as read, and the badge disappears.
Use the search box above the list (알림 검색…) to filter by content or watcher name. 30 items are shown at a time, with Load more to continue. If there is nothing yet, "No new notifications / Alerts detected by watchers will appear here." is displayed.
Sending to a webhook
If you enter a Slack or Teams incoming webhook address in the Webhook URL field when creating a watcher, alerts are also sent to that channel. Messages use the 🚨 [와처이름] 내용 format.
Webhooks and live toasts are sent once, only at the moment the state changes from normal to alert. While the condition persists, nothing is resent on each interval; a new message is sent only after the state returns to normal and the condition triggers again. This is by design, to keep channels from being flooded overnight. (The run history itself is recorded every interval, so you can review the progression in the [Settings] tab.)
Managing watchers
From each watcher card in the [Settings] tab you can do the following.
- Enable/Disable — Toggle on the left of the card. Disabling stops scheduled runs; re-enabling resumes them. During a brief inspection, disable it rather than deleting it.
- Run now — Runs once immediately without waiting for the interval. Use this to verify that a newly created watcher's instructions work as intended.
- Delete — Trash icon (
fa-trash-can). Deleted after confirmation; this cannot be undone. - Run history (n) — Expand to list recent runs with timestamps and result color dots.
The card shows the run interval ({n}분마다), the Last run time (실행 전 if it has never run), and the last result badge.
| Badge | Meaning |
|---|---|
| Normal | Ran, and the condition was not met |
| Warning | Condition met — notification sent |
| ERR | Run failed (data query error, AI not configured, etc.) — the reason is shown in the body below |
The name, instructions, and interval of an existing watcher cannot be edited directly in the UI. To change the condition, create a new watcher and delete the old one. If you only need to pause it temporarily, just turn the toggle off.
If "Platform disconnected — live data and watchers paused" appears in the bottom status bar, watcher runs are skipped. The run history records "Run skipped due to Platform disconnection (auto-resumes when the connection is restored)," and once the connection is restored, runs resume automatically from the next interval. This behavior prevents false alerts. If the connection stays down for a long time, see Troubleshooting.
Permissions
- Admin and Builder — Can create, enable, disable, and delete watchers.
- Viewer — Can see the [Notifications] tab and the watcher list, but the creation form and the run/delete buttons are hidden.
Watchers and notifications can also be checked and operated from a phone in the field — see Field Mobile.