Activepieces 10-minute flow timeout and long Sume video jobs

Do not hold one Activepieces run open while a Sume job renders. Submit with the default async mode, which returns 202 with a job id immediately, then pause the flow and continue when the result arrives. The Activepieces Cloud flow run timeout is 10 minutes of active execution, and paused time does not count against it.

Activepieces limits are from its limits page and release notes; Sume behavior is from Jobs and results, read 2026-09-30.

What exactly is the Activepieces limit?

The limits page lists 10 min on Cloud for both the flow run and a single action run (AP_FLOW_TIMEOUT_SECONDS, default 600 self-hosted). It says the timeout counts only active execution time, and that flows paused by Wait for Approval or Delay do not count against it.

Why not poll inside the run?

Polling keeps the run active, so every minute counts toward the 10. Sume's docs say to use exponential backoff and stop on completed, failed or canceled, and to honor next_poll_after_seconds when it is present. A loop that sleeps inside the run burns the active budget doing nothing.

What should the flow do instead?

Submit with mode: "webhook" and a webhook_url that resumes a paused step, so the run stops counting time while Sume works. Sume sends one terminal event (job.completed, job.failed or job.canceled) and the docs say to verify its signature and keep polling status_url as a backup. See the waitpoint resume pattern.

What if the run times out anyway?

Treat the timeout as a lost connection, not a failed render: the job keeps its id on Sume. Store the id at submit time so a new run can read it with GET /v1/jobs/{job_id}/status.