Great video Andy. This new Autogen Agent Class will be really important for applications where multiple levels of agent discussions are needed that do not mix with the others. Watching your video has gotten me excited to get working on my use case. I am thinking about where I have many agents each assigned to a specific knowledge base. The discussion between a managing agent and those knowledge base agents needs to remain on a level below the primary agent conversation that relays a summarization of the discussed knowledge to the user proxy. It is a simple example, but in my case my knowledge base agents can encompass large sets of data and there can be multiple of the knowledge base agents involved depending on the extent of the user proxy input.
Thank you! Yeah that's the exact use case for the new agent type. I've already tried using it for a side project of my own that sounds just like what you're describing. Results have been mixed so far, but as usual, the real trick is going to be with clever prompting. Let me know how it goes, or if I could look into covering any other topics that might help!
Excellent video. Would be interesting to see if it is possible to have a "society-of-mind" in a "society-of-mind", almost like a thought inside a thought.
Thank you! I haven't experimented with it too much, but I really don't see why not. I don't think there's anything in place that would explicitly prevent someone from doing that.
Excellent video, very clearly explained. I can immediately see the utility of this in a project I am working on.
Thank you, I'm glad it was helpful!
Great video Andy. This new Autogen Agent Class will be really important for applications where multiple levels of agent discussions are needed that do not mix with the others.
Watching your video has gotten me excited to get working on my use case.
I am thinking about where I have many agents each assigned to a specific knowledge base. The discussion between a managing agent and those knowledge base agents needs to remain on a level below the primary agent conversation that relays a summarization of the discussed knowledge to the user proxy. It is a simple example, but in my case my knowledge base agents can encompass large sets of data and there can be multiple of the knowledge base agents involved depending on the extent of the user proxy input.
Thank you!
Yeah that's the exact use case for the new agent type. I've already tried using it for a side project of my own that sounds just like what you're describing. Results have been mixed so far, but as usual, the real trick is going to be with clever prompting. Let me know how it goes, or if I could look into covering any other topics that might help!
Excellent video. Would be interesting to see if it is possible to have a "society-of-mind" in a "society-of-mind", almost like a thought inside a thought.
Thank you! I haven't experimented with it too much, but I really don't see why not. I don't think there's anything in place that would explicitly prevent someone from doing that.
Phenomenal