Ask HN:有人在生产环境中使用 OpenAI 的 Agent SDK 吗?
1 分•作者: sukit•6 个月前
我正在探索构建一个简单的基于 Agent 的应用程序,并发现了 OpenAI 的 Agent SDK:https://openai.github.io/openai-agents-js/。据我所知,它封装了许多功能——比如 agent 循环、函数调用以及与 OpenAI MCP 服务器的集成——与使用普通的 OpenAI SDK 相比,这可能会为我节省大量工作。<p>但是,我很好奇:<p>Agent SDK 是否过于抽象或难以调试?<p>有人在实际的生产应用程序中使用过它吗?<p>如果为了获得更多控制权和透明度,我直接在普通的 OpenAI SDK 上自己实现逻辑,会不会更好?<p>感谢任何见解。
查看原文
I'm exploring building a simple agent-based app and came across OpenAI's Agent SDK: https://openai.github.io/openai-agents-js/ . From what I understand, it wraps a lot of functionality — like the agent loop, function calling, and integration with the OpenAI MCP server — which could potentially save me a lot of work compared to using the plain OpenAI SDK.<p>However, I'm wondering:<p>Is the Agent SDK too abstracted or hard to debug?<p>Has anyone actually used it in a real production app yet?<p>Would I be better off just implementing the logic myself on top of the plain OpenAI SDK for more control and transparency?<p>Appreciate any insights.