Hello. I am trying Koog daily. I have a question:...

# koog-agentic-framework

[nhiroaki]

Hello. I am trying Koog daily.

I have a question:
how can I implement Human in the Loop on the server side? I don’t think readln() like the AskUser tool can be used on the server side. Should I implement it with SSE or something? Can the state be preserved when interrupted?

With langGraph’s interrupt, it seems that it can be resumed using Command.
https://langchain-ai.github.io/langgraph/how-tos/human_in_the_loop/add-human-in-the-loop/#resume-multiple-interrupts-with-one-invocation

[Sergei Pestrikov]

Hi! Sorry, I don't have any updates about human-in-the-loop support for now. I think we'll post updates to the issue.
If you have suggestions or requirements for your use case please let us know (here or in the issue).
Is see that in addition to Langgraph Interrupts there is also Human-in-the-loop Middleware in Langchain. I was wondering which approach is more convenient for your use cases.

[electrolobzik]

@Sergei Pestrikov I believe it would be nice to have both. The first is to control the flow, when you need some feedback from the user inside the graph, especially if it affects the following steps. The second for me looks more like a policy thing, which allows to control what is allowed and what is not, getting confirmation from the user.

[electrolobzik]

In my particular case the first is most important, since I am building chat-bots where agent can ask something or confirm something with the user.

[Sergei Pestrikov]

Thank you! I've noted it in the ticket.