HN 提问：有没有支持多大语言模型协同工作的编排框架？
1 分•作者: ch3coohlink•3 天前
我的总体看法是：我觉得 Gemini 在高层次的代码重构方面非常出色，但在编写实际代码时却充满了 bug。另一方面，GPT/Claude 在编码方面表现优异，但说到重构，它们往往只进行小修小补。它们喜欢为了维护向后兼容性而加入不必要的防御性编程，或者最终写出冗长混乱的代码。
我的想法是互补它们的优势：让 Gemini 提供方向性/架构性的想法，然后让 GPT/Claude 来讨论和实现这些想法（事实上，我一直都是手动这么做的，效果相当不错）。
所以我的问题是：是否有任何 Agent 框架能够有效地自动化这种协作？
我清楚现有的“子 Agent”功能，但在我的经验中，AI 并不总是选择调用它们。此外，当子 Agent 工作时，主模型通常只是在那里空闲着。这感觉不像真正的协作，更像是一种防止外部任务污染主上下文的机制——至少，这是我的感觉。
查看原文
Here is my general take: I feel Gemini is excellent at high-level refactoring but riddled with bugs when writing actual code. On the other hand, GPT/Claude excel at coding, but when it comes to refactoring, they tend to stick to minor patches. They love throwing in unnecessary defensive programming just to maintain backward compatibility, or ending up with verbose spaghetti code.<p>My idea is to complement their strengths: let Gemini provide the directional/architectural ideas, and then have GPT/Claude discuss and implement them (in fact, I do this manually all the time, and the results are pretty good).<p>So my question is: Are there any Agent frameworks that can automate this kind of collaboration effectively?<p>I'm well aware of the existing "subagent" features, but in my experience, the AI doesn't always choose to invoke them. Moreover, while the subagent is working, the main model usually just sits there idling. It feels less like actual collaboration and more like a mechanism to prevent outsourced tasks from polluting the main context—at least, that's how it feels to me.