feature request: a persistent queue

[xinjiguaike]

In my case, we need to run the project in multi instances, but current queue is in memory, the queue can't be shared between instances; so a persistent queue(like celery...) would be very helpful, then this project would be suitable for production.
Any play for this ?

[hiSandog]

A persistent queue would help production deployments, but it may be worth scoping the contract before choosing Celery/Redis/etc. The tricky parts are job idempotency, image artifact storage, retry behavior after worker crashes, and whether multiple API instances are allowed to pick up the same job. If those semantics are written down first, the backend can be swapped without changing the public job API later.