Alerts tell you that deliveries are failing before your customers do. Channels are created on a source’s Alerts tab and can be scoped to that one source or to every source in the workspace.
Channels
Section titled “Channels”| Channel | Fields |
|---|---|
| Recipient email | |
| Telegram | Bot token, Chat ID |
Scope is either This source or All sources. Several channels can be active at once — email to a shared inbox, Telegram to the on-call chat — and each can be enabled or disabled without deleting it.
To set up Telegram, create a bot with @BotFather, copy the token, add the bot to the target chat, and use that chat’s id.
What triggers an alert
Section titled “What triggers an alert”| Trigger | When |
|---|---|
| Delivery exhausted | Every retry in the schedule failed for one event. |
| Destination unhealthy | The last 5 attempts to that endpoint all failed. |
| Circuit opened | 5 consecutive retryable failures paused the endpoint. |
All three are deduplicated per destination and kind within a 15-minute window, so a broken endpoint sends you one message, not one per queued event.
There is no alert for a single failed attempt. A 500 that the next retry fixes
is not worth waking anyone up for.
What an alert does not do
Section titled “What an alert does not do”Alerts are a notification, not a control. Nothing is dropped or disabled when one fires: deliveries keep retrying on the schedule, and events keep being accepted. The recovery path is the dead-letter queue once the endpoint is healthy.