search

hoanganh27042001 / llm-multi-agents

0 stars 0 forks source link

Hi! very nice repo #1

Open NikolayTV opened 6 months ago

NikolayTV commented 6 months ago

Since you tried to implement both with autogen and langgraph can you share experience on which one is better?

hoanganh27042001 commented 6 months ago

Hi Sir, thank you for your interest in my project. Based on my experience, I prefer LangGraph over Autogen when implementing multi-agent systems. I think that Autogen was challenging to control its output, while LangGraph offered better precision and control over the cycles the language model is performing. Additionally, LangGraph was praised for its ability to enable users to manage the interactions and workflows of multiple agents effectively. Therefore, LangGraph seems to be a better choice for me over Autogen for building a big multi-agent system due to its precision, control, and ease of use in managing agent interactions and workflows.

NikolayTV commented 6 months ago

I spent few days with autogen and also found it extremely difficult to customize multi-agent flow.
I did not want to look on langgraph originally because I though that it is another pile of stuff... but after digging deeper I think now this is the tool I want to use

NikolayTV commented 6 months ago
image

i want to make something like this. Tool to help human to create some routine based on some pre-configured scenario that Coordinator will try to follow, and calling different specialized more or less expensive agents

hoanganh27042001 commented 6 months ago

Your project is so interesting. The Coordinator can take the role of "supervisor" (an LLM) to help delegate tasks. Here is a helpful notebook to start with supervisor agent I found with Langgraph.