Alerting keeps you informed about what’s happening in your AI deployment without having to watch dashboards. You set the conditions you care about, and Nebuly notifies your team in Slack or Microsoft Teams when they happen, or on a regular schedule.
What it does
There are two kinds of alerts:- Threshold alerts: fire a notification when a metric crosses a value you define (e.g. error rate goes above
5%). - Recurring summaries: send a periodic digest of a project’s activity on a schedule you choose.
Where to manage alerts
All alerts live under Settings → Alerts. From there you can see every alert configured for the organization and create, edit, or delete them. Each delivered notification is also recorded in a notifications table, where the date column is labelledStarted on {date} so you can see exactly when each one fired.
Threshold alerts
A threshold alert watches one metric and notifies you when it crosses your limit. 1. Choose the metric. Options are:| Metric | What it tracks |
|---|---|
| Error Rate | Share of interactions with an error |
| Interactions (Count) | Number of interactions |
| Users (Count) | Number of users |
| Topic | Volume of selected topics, as a count or a percentage |
- Error Rate and Topic (percentage): entered as a percentage.
- Interactions, Users, and Topic (count): entered as an absolute number.
Threshold alerts are edge-triggered: you are notified once, at the moment the threshold is crossed, rather than repeatedly for as long as the condition stays true. You will be notified again only when it crosses back and then breaches again.
Recurring summaries
A recurring summary is a scheduled digest of a project’s activity. You pick:- The schedule: daily or weekly.
- The start date and time the digest should begin from.
Setting up delivery
Alerts are sent to a Slack or Teams channel via an incoming webhook, which you create once in the destination tool and paste into Nebuly:- Slack: Sending messages using incoming webhooks
- Microsoft Teams: Send messages in Teams using incoming webhooks