Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

any way of easier "debugging" what agents are saying in the run? #220

Open
dakipro opened this issue Feb 5, 2024 · 6 comments
Open

any way of easier "debugging" what agents are saying in the run? #220

dakipro opened this issue Feb 5, 2024 · 6 comments

Comments

@dakipro
Copy link

dakipro commented Feb 5, 2024

Hi,
love the script!
I wonder if there is (and if not if it could be added) some way of visualizing what agents are thinking and doing.
With a lot of text and instructions, it is very cumbersome to follow what was happening in the console

Thanks!

@joaomdmoura
Copy link
Owner

I hear you, is that a specific style you would like these to be on, a dashboard maybe of sorts?

@dakipro
Copy link
Author

dakipro commented Feb 8, 2024

it would be easiest to use something existing I guess. I like the "gamification" and visualization that devchat has, but it is also difficult to customize (the advantage of crewAI). How does LangChain do that? It is already used in crewAI...

but a simple parser, or a simple HTML with accordions to review the chain after the fact would be helpful. Especially now with several processes. Who said what and to whom. Simple but useful :)

nice job with the project btw, thank you for making it. I've tried them all, and crewai is so far the easiest and goes straight to the point, at least for me for video content creation of 3min youtube videos (if you want to make an example on that usecase - it would be awesome :) )

@ZmeiGorynych
Copy link

I'd love to see that example ;)

@JaDenis
Copy link

JaDenis commented Feb 11, 2024

I agree, langsmith-like logs would be perfect, but I’d be 100% happy with a json :)
My interest lies in collecting prompts agents are sending to each other. These prompts are often crazy good and could be useful to collect.

@ZmeiGorynych
Copy link

If #246 is accepted, you could just use eg a Langchain agent and LangSmith :)

@Codeshark-NET
Copy link

I hear you, is that a specific style you would like these to be on, a dashboard maybe of sorts?

Also feel this would be quite worthwhile for debugging and iterating on the agents. 👍 Promptfoo is not agent focused, but it did solve that problem very well for me in the past with traditional prompt engineering, perhaps some inspiration there. I'm happy to provide some assistance with the frontend part too. 🚀

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

5 participants