Human in the loop

There is not. However, I think you could make a tool that calls on the user for input? But then it depends on the agent calling that tool

Otherwise, subclass is the way to go

got it, thanks @Logan M !

now i'm looking at whether i can subclass in a stateless web server.

is there an obvious way to serialize the agent (e.g. the BaseReasoningSteps) and then reconstruct the agent after the user approves?

i can't see an obvious way to initialize an agent with partially complete reasoning

ooo yea serializing like that will be a little tough 😅 tbh, the react agent class probably needs to be designed from the bottom up with human in the loop in mind 🤔

Tbh that would be an amazing PR if you wanted to tackle that

got it, thanks!!

i'm hacking on it now, i'll let you know if anything comes of it.