Automatically continue when usage resets

[Neokil]

Problem

Right now if your tokens are used up the workflow is in a failure state and needs to be manually restarted. This should happen automatically.

Solution

We should be able to determine when the usage-reset is happening and automatically schedule the step to be reexecuted once the usage-limit is reset.
This should also be shown in the WebUI with a "scheduled to X because the usage-limit was reached" message.

[irmorteza]

Hi, I'd like to work on this issue.

My approach for backend:

• Add quota_monitor.go — a background daemon responsible for tracking quotaReached state. It runs on a 15-minute interval and when quota is reached, checks the LLM with a minimal request (like Ping) to detect when the quota has reset.
• In jobs.go, workers check the quotaReached flag before picking up each job — if true, they pause and wait until it becomes false.
• If a running job hits a quota error, it sets quotaReached = true and re-queues itself back into the job queue with status reset to "queued".
• For the WebUI: show a "usage limit reached, retrying automatically" message on affected tickets. Open to discussion on the best API design for this.

Does this direction make sense?

[Neokil]

That sounds like a good way to do it. I had actually hoped that the tools like codex or claude would maybe expose that information somehow, but they seem to want to hide it until you reach the limit.