OpenUI support

I think using OpenUI or any other ways to let the agents to output generative ui between the generic text output would be a nice feature to have deeper information rich and still glanceable conversations.