Retries & Replay

Delivery rules#

• Timeout 10 seconds. Success is any 2xx. Redirects are not followed.
• Failures retry at +1 m, +5 m, +30 m, +2 h, +12 h — six attempts in total — then the delivery is marked dead.
• A 410 Gone disables the endpoint immediately. 30 consecutive failures also disable it (disabled_reason: too_many_failures).
• Re-enable a disabled endpoint with PATCH { "status": "active" }; the failure counter resets.
• Deliveries and events are retained 30 days.

Delivery log#

Portal: Developers → Deliveries (admin) or the endpoint page on the developer dashboard. API:

curl "$API/webhooks/deliveries?status=dead" -H "Authorization: Bearer $KEY"
curl "$API/webhooks/deliveries?endpoint_id=we_…&event_type=exam.graded" -H "Authorization: Bearer $KEY"
curl $API/webhooks/deliveries/whd_… -H "Authorization: Bearer $KEY"     # payload + your response body

Each row shows status (pending, delivering, succeeded, failed, dead), attempt count, next attempt time, HTTP status, response time and the first 2 KB of your response.

Replay#

curl -X POST $API/webhooks/deliveries/whd_…/replay -H "Authorization: Bearer $KEY"

Replay creates a new delivery (replay_of points at the original) with a fresh signature timestamp. Use it after fixing a bug on your side. You can also replay an event to a different endpoint from the portal's event browser.

Polling fallback#

If you cannot receive webhooks at all:

curl "$API/events?since=2026-09-14T09:00:00Z&type=exam.*&limit=200" -H "Authorization: Bearer $KEY"

Persist the newest created_at you processed and pass it as since next time.

Operational tips#

• Respond 202 immediately and process from a queue.
• Alert when GET /webhooks/endpoints shows status: disabled or consecutive_failures > 5.
• Keep the endpoint idempotent on id; a replay or a retry after a timeout can deliver the same event twice.