s4h1lk

Very cool. Curious if the results are better for certain domains vs others and if there's a sweet spot in terms of number of turns in the convos. What have you seen deploying this in the wild so far?

I had chosen AutoGen over a number of other multi-agent frameworks out there because of some notable features and also the fact that it seemed to have MS's muscle behind it with a lot of activity with folks from a diverse set of backgrounds. That said, what has u folded in the last few days both from the two main others of the original paper (now branching off with AG2), and the MS team (autogen-chat et al) has put all notions of stability in question.

In the meantime, while there were turf wars being waged in the Autogen arena, CrewAI has been building itself it to much more reliable and fully featured alternative. They have even removed LangChain as a dependency. The close association with LangChain had been a thorn in their side, IMO for some side and a reason for me to opt not to use.

Sad to see these event in the OSS community. I'll be looking at CrewAI for my next projects.