用 Jev 做 Go 终端文档分类器 Docket
作者拿到 Jev 权限后做了个 Go 终端应用，对 PDF/图片判断类别、敏感度、紧急度和行动概率，开发测试成本不到 1 美元。
Monday came early. Got access to Jev from @typesafeai and built Docket, a Go terminal app for classifying documents. Give it a PDF or image and it tells you the category, sensitivity, urgency, and probability that you actually need to do something about it. OpenAI or OpenRouter handles text extraction, Jev makes the judgments, and Go applies explicit thresholds, shows guidance, and files a local copy. I also wanted the local-data story to be pretty boring in a good way: config.json only stores provider/model/path settings, SQLite stores metadata + Jev results, extracted document text is not persisted, and BYOK credentials live in the OS keychain. The Jev portion of development + testing cost me less than $1. Repo: https://github.com/glamboyosa/docket Marketing: https://docket.glamboyosa.xyz
这条是引用着另一条写的：x.com/glamboyosa/status/2101415401347313705