问HN：我正在准备我的第一次开源发布，你会使用这个吗？
嘿，HN，
我这几个月一直在构建一个AI编排框架，主要是为了我自己的项目，现在我考虑很快将其开源。我以前从未进行过开源发布，所以我希望能获得一些建议，帮助我以一种不会在GitHub上消失的方式进行发布。
简而言之，它的功能包括：
- 以清晰的结构协调多个AI代理的提示和输出。
- 允许你添加“防火墙”，在代理输出发送之前对其进行过滤或重写。
- 拥有一个分块系统，可以将大任务拆分为小任务并并行运行。
- 使用数据库进行持久化/并发（如果大家觉得太重，可以选择不使用）。
- 非常“即插即用”。创建一个新代理基本上就是创建一个文件夹并编写一个提示。
- 为“随性编码”而构建。工具和过滤器是最小的样板代码，方便你快速组合。
我很想听听HN社区的看法：
有没有社区或用例会对这样的东西感兴趣？
在发布开源时，是什么让你真正想尝试或贡献给一个新框架的？
对“随性编码”的强烈关注会被视为优点，还是只是显得草率？
目前还没有开源——只是想确保在发布时能够正确启动。
查看原文
Hey HN,<p>I’ve been building an AI orchestration framework for the past few months, mostly for my own projects and I’m thinking about open-sourcing it soon. I’ve never done an open-source release before, so I’m looking for advice on how to launch it in a way that doesn’t just vanish into the GitHub void.<p>What it does (in short):
- Orchestrates multiple AI agents with a clear structure for prompts and outputs.
- Lets you add “firewalls” to filter or rewrite agent outputs before they go anywhere.
- Has a chunking system for splitting big jobs into smaller tasks and running them in parallel.
- Uses a database for persistence/concurrency (optional if people think it’s too heavy).
- Very “plug-and-play”. Creating a new agent is basically making a folder and writing a prompt.
- Built for “vibe coding”. Tools and filters are minimal boilerplate so you can hack them together fast.<p>I’d love to know from the HN crowd:
Are there communities or use cases that would be interested in something like this?<p>When releasing open source, what makes you actually want to try or contribute to a new framework?<p>Would a strong focus on “vibe codability” be seen as a plus or just sloppy?<p>Not open source yet—just trying to make sure I launch it right when I do.