andura.ioProduct / Public code
Long jobs outlive the request. Give them a runner.
Runpin is a self-hosted durable job runner. Enqueue over HTTP, get a 202 back, and a worker with leases, heartbeats, and retry backoff finishes the job on your own machine.
What it is
Queue work you can actually walk away from.
Long work outlives the request that started it. A long AI job or a batch import hits the platform timeout, so you either hold the connection open or hand-roll a queue with retries.
I kept writing the same queue — claim, retry, cron, status page — inside every project. I wanted long-running work to finish somewhere I control, with the queue and the logs in one place.
A durable job runner you host yourself. Enqueue over HTTP and get a 202 back; a worker claims the job with a lease, heartbeats while it runs, reclaims stale leases, and retries with backoff until max attempts. Cron schedules run in UTC, and an operator UI reads the same SQLite state.