All-Hands-AI
[Bug]: After max iterations are hit, they do not reset on a user message
Is there an existing issue for the same bug?
- [X] I have checked the troubleshooting document at https://docs.all-hands.dev/modules/usage/troubleshooting
- [X] I have checked the existing issues.
Describe the bug
I had a long and productive OpenDevin session, and hit the maximum number of iterations 100. I wanted to continue the session by providing additional input, but it seems that max_iterations is not extended on additional user input.
It would be nice to be able to continue long sessions.
Current OpenDevin version
commit 5e6fd58f56840f20fc619349428092b7ddae45fc
Installation and Configuration
make build; make run
Model and Agent
- Model: Claude 3.5 Sonnet
- Agent: CodeActAgent
Operating System
No response
Reproduction Steps
- Run OpenDevin for a long time
- Type "continue" when you get a max iterations error
Logs, Errors, Screenshots, and Additional Context
Related to https://github.com/OpenDevin/OpenDevin/issues/2710 , but a little different.
It actually does continue (for one more step), as shown in the screenshot below:
You just need to keep hitting the "continue" button.
Some UI design might be needed if we want to give users more flexibility, e.g. resetting the limit, increasing the limit exponentially, etc.
Yeah, my expectation would be that after the user gave input it would reset. But I'm not sure what others think.
Now that I think about it, max_iterations is much more useful when running headless and with user confirmation fully disabled, than in interactive UI runs. Particularly when the user will have the ability to stop or pause easily, it might become unnecessary to restrict iterations? Alternatively, a reset sounds good to me.
max_iterations is still a little bit useful to me in interactive ui runs because I will sometimes start one running, then go to the grocery store or something, and then come back. But I do think it'd be good to be able to continue for another 100 turns or so if you send a new message.
