recovery.incident_opened
A health probe found the creator's sign-in account unreachable, a project's bot refused by the platform, or a channel or group gone or no longer administered.
When this fires
The hourly probes check every bot, every channel and group a project sells, and the creator's own sign-in account. When one comes back banned, deleted, or stripped of the rights the bot needs, an incident opens in the recovery ledger and this event fires — once. A probe that finds the same broken thing on its next run reopens nothing and fires nothing; the incident stays open until it resolves. Undoing an account relink also opens an account incident (reason: relink_disputed) so support reviews the dispute.
Required ability
project-recovery:view — token must carry this at mint time to subscribe an endpoint to this event.
Payload
{
"id": "evt_01HX...",
"type": "recovery.incident_opened",
"created_at": "2026-09-12T03:19:04Z",
"api_version": "2026-05-01",
"project_id": "7f3d1c92-8b45-4e6a-9d21-5c8e0a4b6f13",
"data": {
"incident_id": "6f1e9b27-4c3a-4d58-9e02-b7a1c5d38f64",
"kind": "resources",
"connector": "telegram",
"project_id": "7f3d1c92-8b45-4e6a-9d21-5c8e0a4b6f13",
"resource_id": "b73c5f21-9d80-4a6e-8215-4f70ce13a9d6",
"reason": "chat_not_found",
"detected_at": "2026-09-12T03:19:04Z"
}
}Field reference
| Field | Type | Notes |
|---|---|---|
id | string ULID | Unique event id, prefixed evt_. Use for idempotent processing. |
type | string | Always recovery.incident_opened for this event. |
created_at | ISO 8601 timestamp | Server-side emission time. |
api_version | string | Webhook API contract version. |
project_id | string UUID | null | Project this event belongs to; null for an account incident. |
data.incident_id | string UUID | The incident; read it with GET /v1/recovery/incidents/{incident}. |
data.kind | string | account (the creator's sign-in account), bot (a project's installation) or resources (a channel or group). |
data.connector | string | null | The connector the broken thing lives on, such as telegram. |
data.project_id | string UUID | null | The project, for bot and resources incidents; null for account. |
data.resource_id | string UUID | null | The channel or group, for resources incidents; null otherwise. |
data.reason | string | null | The connector's code for what it saw (chat_not_found, bot_removed, …); the API's reason_label puts it into words. |
data.detected_at | ISO 8601 timestamp | When the probe saw it. |
Caveats
- One event per incident. The ledger opens one incident per subject and keeps it open, so a subject that stays broken for a week fires this once, not hourly.
- A restriction the platform reports as "still reachable" opens nothing: only a verdict that warrants a replacement counts. See How detection works.
Related events
recovery.incident_resolved— the same incident healing or being fixed.project.resource.status_changed— the health verdict that precedes a resources incident.- Recovery events overview — back to family overview.
How is this guide?