Skip to content

Jev directory

Work with Jev

A local-first classifier for Feishu and WeCom work messages: Jev judges whether each message matters to you and the app sorts them into urgent, to-do, worth reading and skip, building a cross-group to-do list. The demo runs without a key.
Stars
4
Language
JavaScript
License
MIT
Reading time
3 min
Published by
Community
Author
adkid-zephyr
Last verified
Oct 2026

Highlights

  • Four buckets - urgent, to-do, worth reading and skip - plus manual correction, cross-group to-dos, drag ordering and completion with undo.
  • npm ci and npm start serve the interface at http://127.0.0.1:4173; the built-in demo uses clearly labeled preset results with no registration or key.
  • Real classification needs a TypeSafe API key entered in settings; the app lets you check messages before sending, and calls consume your API quota.
  • Feishu reads selected groups the user authorizes via the official CLI; the WeCom bot receives private chats and @-mentions, not full group history.
  • Reading and calling the model are separate steps - no auto-replies or task execution - and batch size is selectable from 1 to 200, default 50.

Quickstart

bash
git clone https://github.com/adkid-zephyr/work-with-jev.git
cd work-with-jev
npm ci
npm start

Watch out

Copy reviewed against the repo README on 2026-10-15. The primary README is in Chinese with a linked English version; the four-bucket examples are scenario illustrations, not an accuracy evaluation, and the README warns that probabilities are not correctness guarantees and default thresholds need testing on your own messages.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai