Dynamic Agent Composition with Reinforcement Learning

[peymanrahi]

What feature would you like to be added?

Implement a system for dynamically composing and optimizing agent workflows using reinforcement learning (RL) techniques

(this feature can be integrated into Existing Framework by Extending the 'AgentWorkflow' class )

Why is this needed?

This feature would enhance AutoGen's ability to adapt to complex, multi-step tasks by automatically discovering and refining effective agent combinations.

[rysweet]

@peymanrahi can you be more specific about what you are missing? you can add/remove agents using the RegisterAgentRequest.