请问HN：如何在生产环境中管理多个AI代理？
我正在构建一个多智能体系统，但对一些问题有些困惑。
- 如何运行多个可能会接触到相同数据的智能体？
- 如何制定关于智能体可以/不可以做什么的政策？
- 有没有人成功地将不同的框架（如LangGraph、CrewAI和自定义智能体）混合在一起？<p>你们在系统中是如何处理这些问题的？
查看原文
I'm building a multi-agent system and I'm a bit confused about a few things.
- How to run multiple agents that might touch the same data?
- How do one have policies about what agents can/can't do?
- Has someone successfully mixed different frameworks (LangGraph, CrewAI and custom agents)?<p>How do you handle this in the system?