Has anyone tried this approach with the new automations and using it to keep kicking off an agent for long running tasks to maintain orchestration? If a conversation dies or stalls or is way off track, it could fix it / create a new one. Seems pretty straightforward.

It's like having an agent perform the ralph wiggum loop instead of yourself, seems cool

Hmm maybe if we let the model know the automation server is there and available, it would use that instead of a local cron job when it's doing its own orchestration. 🤔