Webhooks

Live
Receive signed terminal Prediction events with retries and delivery logs.
Open in ChatGPT
(opens in new tab)Last verified: 2026-08-11

Register an endpoint

Call POST /webhooks/endpoints with a public HTTPS URL. Save the returned signing secret once. Polling through GET /predictions/:id remains available and is the recovery path when delivery is delayed.

Verify and deduplicate

Read the raw body, parse x-modelrush-signature as t=TIMESTAMP,v1=HMAC, reject stale timestamps, and compare HMAC-SHA256 in constant time. Deduplicate by x-modelrush-event-id; handlers must be idempotent because delivery is at least once.

Retries and logs

ModelRush uses a 10-second timeout and exponential backoff for up to eight attempts, then marks the delivery dead-letter. Read logs at GET /webhooks/deliveries; retry failed or dead-letter events with POST /webhooks/deliveries/:id/retry. Manual retry never executes the model task again.
PreviousUpload media
NextErrors, retries, and limits