Cloi – 在您的终端中运行的本地调试代理
大家好！我是Gabriel Cha，刚刚被哥伦比亚大学录取的学生，想和大家分享我在过去两周内开发的项目：Cloi——一个完全本地的调试代理，可以直接在你的终端中运行。链接：https://github.com/cloi-ai/cloi
我创建Cloi是因为市面上的每个AI编码工具都需要API密钥、订阅服务，并且要求将你的整个代码库上传到他们的服务器。而Cloi则完全在你的机器上运行。没有云端，没有API密钥，没有订阅服务，数据不会离开你的系统。
这个技术非常简单：它捕捉你的错误上下文，局部启动Ollama，生成针对性的修复方案，并且——仅在你明确允许的情况下——直接将补丁应用到你的文件中。你可以在任何更改发生之前审查每个差异。我们默认使用Phi-4（在其体积下表现非常出色），但你可以切换到任何已安装的Ollama模型。你的机器，你的模型，你的规则。
安装非常简单：$ npm install -g @cloi-ai/cloi
顺便提一下，Cloi是免费的。如果你对实现感兴趣或想请求新功能，可以在这里提出：https://github.com/cloi-ai/cloi
自发布以来，我们在短短5天内就获得了202颗星星！！！如果你觉得这个工具有用，我会非常感激你的点赞。试试看，告诉我们你的想法！
来自Gabriel Cha和Mingyou Kim
查看原文
Helloo HN! I’m Gabriel Cha and an incoming student @ Coluimbia and just wanted to share what I've been building past 2 weeks: Cloi - a completely local debugging agent that runs right in your terminal. https://github.com/cloi-ai/cloi<p>I created Cloi because every AI coding tool wants API keys, subscriptions, and your entire codebase uploaded to their servers. Cloi, however, runs entirely on your machine. No cloud, no API keys, no subscriptions, no data leaving your system.<p>The tech is brutally simple: it captures your error context, spins up Ollama locally, generates targeted fixes, and - only with your explicit permission - applies patches directly to your files. You review every diff before anything changes.We ship with Phi-4 out of the box (crazy good for its size), but you can swap to any Ollama model you've got installed. Your machine, your models, your rules.<p>Installation is just: $ npm install -g @cloi-ai/cloi<p>BTW Cloi is free, if you're interested in the implementation or want to request new features, you can do so here: https://github.com/cloi-ai/cloi<p>We've hit 202 stars in just 5 days since release!!!
I'd really appreciate your star if you find this useful.
Try it out and let us know what you think!
From, Gabriel Cha & Mingyou Kim