camel-ai
[Feature Request] Support human on the loop
Required prerequisites
- [x] I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
- [ ] Consider asking first in a Discussion.
Motivation
CAMEL has supported human in the loop, this approach integrates humans directly into the agent process to refine results continuously, while human-on-the-loop systems use human oversight to correct results after they are produced, and even go back to earlier steps and then rerun (with changes) from there this feature need some systematic design
Solution
No response
Alternatives
No response
Additional context
No response
If it's not on the front-end page but in the command line, perhaps triggering human-in-the-loop by pressing certain keys?
I have some ideas:
The "human" only acts as a substitute for the "user" to provide instructions.
Rollback (undo) should be achievable by deleting memory. After rolling back, the "human" should again issue the instruction.
Rollback might not be very intuitive to present to the user in a terminal-based code execution environment. Should we display a pop-up selection window after choosing to rollback?
what do you think @Wendong-Fan
hmm should we maybe discuss a design for this human in the loop first? what components is this for?
hmm should we maybe discuss a design for this human in the loop first? what components is this for?
i think It can first be used for Roleplaying