Hello, thanks for putting up a great repo to experiment with MoA. I have made a few changes which you might find helpful:

• Fixed the display of agent replies, now they are properly delimited
• Updated Readme focusing on how to update the default list of models - at the beginning I was not clear if I should enter model names to pick from somewhere in the UI
• Made the selection smaller and picked models that are easier to run on a consumer PC (I manager of have all 3 proposers to run OK on a single RTX 4090)
• Changed agents system prompts for agents, removed priming which made the results better for small models