展示HN：PrivaKit – 通过WebGPU实现的零上传AI浏览器工作区
嗨，HN，我是这个项目的开发者。
我意识到在敏感工作流程中使用云端AI API——比如转录董事会会议、光学字符识别（OCR）雇佣合同或处理身份证照片——存在巨大的隐私风险。因此，我使用transformers.js、Whisper和WebGPU构建了一个客户端工作空间。所有操作都在本地进行。您可以在初始模型加载后启用飞行模式，系统仍然可以完美地转录和提取文本。
为了保持透明，我撰写了一份技术审计，详细说明了数据的流动（或者说，数据并没有流动）。我的唯一后端是位于新加坡的一台小型2核节点，运行自托管的Plausible分析：[<a href="https://gist.github.com/ygx2378/3275b333504c6a9def50ef531b548a26" rel="nofollow">https://gist.github.com/ygx2378/3275b333504c6a9def50ef531b54...</a>]
我仍在学习基于浏览器的内存管理，因此非常希望您能反馈一下在您特定GPU上模型的加载情况！
查看原文
Hi HN, I'm the builder.
I realized that using cloud AI APIs for sensitive workflows—like transcribing board meetings, OCRing employment contracts, or cleaning up ID photos—is a massive privacy liability.
So I built a client-side workspace using transformers.js, Whisper, and WebGPU. Everything runs locally. You can turn on Airplane Mode after the initial model load, and it still transcribes and extracts text perfectly.
To keep myself honest, I wrote a technical audit of how the data flows (or rather, doesn't flow). My only backend is a tiny 2-core node in Singapore running self-hosted Plausible analytics: [<a href="https://gist.github.com/ygx2378/3275b333504c6a9def50ef531b548a26" rel="nofollow">https://gist.github.com/ygx2378/3275b333504c6a9def50ef531b54...</a>]
I'm still learning the ropes of browser-based memory management, so I'd love your feedback on how the models load on your specific GPUs!