Hi!
I am developing a project that involves human-in-the-loop. For a design_roadmap_task, I set the human_input = True
. But I do not know how to run this without the CLI (running in the command line with crewai run
). In my main.py
file, I initiated the whole process by response = MyCrew().crew().kickoff(inputs=inputs)
, and print(str(response))
. Is there other ways to pass in the human feedback?
I don’t get it. What’s your idea? How would you like to pass in the human feedback?
You can set
human_input=True
on the task
@matt He did this. He wants to run it via the CLI?
He said he does not know how to run it without the cli
It’s worth pointing out that while it can be done using Crewai as a chatbot is not the intended purpose, it’s built for agentic workflows.
Human_input attribute to help you refine your responses
You’re right, my bad.
so any idea of how to run this without the cli ? i think it will be good for long flows to ask if we want to proceed or not for example
Are there any updates on this?
Hi.
Training - CrewAI might be useful to help give feedback to your agents
If you need chat I’d look at linking to Slack / Google etc and then wait for a response. Then you can not expose your command line to end users
I had the same issue. Check out https://youtu.be/pODI1SWTVeo?si=xrmdg9xtL0wGRyOx for instructions on how to do this with Panel. I also used Process.sequential for the crew so that I could have human input on multiple tasks and only proceed to the next one once I was good with the previous.