[v0.4] High-level interface for Mixture of Agents

[MarkYav]

Thank you a lot for the very detailed docs regarding Mixture of Agents in the autogen-core package!

Do we (or will we) have a high-level API for this in the autogen-agentchat package (like RoundRobinGroupChat)?

[ekzhu]

This could be big. Though we are still waiting to see some solid application scenarios. Before that, anyone is welcome to create a community package that extends agentchat with a mixture of agents team.

See community packages: https://microsoft.github.io/autogen/dev/user-guide/extensions-user-guide/index.html

We can share and index.

[MarkYav]

@ekzhu Is there a guide on how to make a nested chat for AutoGen v0.4?

[ekzhu]

The official design for nested chat is very much in flight at the moment. You can take a look at the implementation of SocietyOfMind agent: https://microsoft.github.io/autogen/dev/reference/python/autogen_agentchat/autogen_agentchat.agents.html#autogen_agentchat.agents.SocietyOfMindAgent

The general idea is to put a team inside an agent. You can also implement a different agent with a nested team.