Sume webhooks plus a sweeper: recover jobs whose callback never came

A Sume job webhook can fail to reach you and the job will still finish. Delivery is retried up to 10 attempts total with a 10 second timeout each, and ten refused attempts leave a failed delivery but a job in its real terminal state. The docs say to keep status_url polling available for exactly this, and a sweeper is that polling, done once in a while rather than in a tight loop.

The pattern

Store every job id with a status column on submit. The webhook handler marks a row terminal. A scheduled sweeper selects rows still non-terminal after your usual job time, reads their status once, and marks them. Redeliver is a second option: POST /v1/jobs/{job_id}/webhook/redeliver re-sends the real terminal event with a fresh signature and does not use one of the automatic attempts.

import asyncio, os
import httpx

async def sweep(open_job_ids: list[str]) -> dict[str, str]:
    headers = {"x-api-key": os.environ["SUME_API_KEY"]}
    found: dict[str, str] = {}
    async with httpx.AsyncClient(base_url="https://api.sume.com", headers=headers) as c:
        for job_id in open_job_ids:
            r = await c.get(f"/v1/jobs/{job_id}/status")
            r.raise_for_status()
            body = r.json()
            if body.get("terminal"):
                found[job_id] = body.get("sume_status", "unknown")
    return found

async def main() -> None:
    print(await sweep(["job_123"]))

asyncio.run(main())

Make handling idempotent

Both paths can report the same job, so key your handler on job_id and ignore a second report. Use the same handler function for the webhook and the sweeper.

What not to do

Do not resubmit the paid request because a callback is missing; that creates a second paid job. Read the status instead.