Building Multi-Agent AI Interaction
Let several AIs play different roles, debating and collaborating to produce interesting exchanges or complete complex tasks.
Translated by AI
💡
Multi-agent AI interaction is a fascinating topic. This approach lets several AIs play different roles, debating and collaborating with each other to produce interesting exchanges or complete complex tasks.
Example: the Venomous Critic and the Flattering Critic
Download the example workflow below.
Shared via Baidu Netdisk: L3-AI多智能体案例:毒舌和彩虹屁艺术评论家.json
Link: https://pan.baidu.com/s/1IzQvgZxbHlf7F1_36d09_Q?pwd=8hfn Access code: 8hfn
You can set up two AIs to play different roles; by talking to each other, they keep improving themselves. Try the workflow below.

This is the interface of the self-evolving AI agent setup.

Before starting a new conversation, set a few options:
- In “Starting Dialog”, set
is_reloadtotrue. - In both “API LLM general link” nodes, set
is_memorytodisableto clear the conversation history from the previous topic.

After the first round of conversation, switch those options to the opposite settings so the conversation can continue with the previous context.
Going further
💡
Could you connect this conversation to the earlier AI agent that can generate images?