The chat decision assistant: before you reply, Jev reads the chat, judges intent and risk, and drafts replies you fill in with one tap. You press send. Android · Windows · macOS · iOS | 聊天辅助决策工具:先看懂对方再回复,发不发由你。

★ 8,316⑂ 1,327👁 14⑃ 34● 51▤ 2◈ 0
KotlinMITCreated 2026-09-21chatjevs.com
8,316
stars
+8+100+1,023
1,327
forks
14
watchers
34
prs
51
issues
2
releases
0
discussions

README

jev-chat

 The chat decision assistant

Read them first, then reply. Before you answer, Jev works out what the other person really means, how risky the moment is and how to respond, then drafts replies you can fill in with one tap. Whether to send is always up to you.





 🌐 Official website: chatjevs.com

English | 简体中文 | Tiếng Việt | Releases

Download · Products · How It Works · Privacy · Community



 ❤️Sponsors

 Interested in sponsoring the project?


Show or hide sponsors




Thanks to Bocha for sponsoring this project! Bocha is a search engine for AI, giving your applications access to information from across the web with clean, accurate, high-quality results. Its services include the Web Search API, Bocha Jev API, and other search and model APIs. open.bocha.cn



Thanks to Xiaoyou Store for sponsoring this project! Xiaoyou Store sells digital products and account services, with a selection available to users of this project. Visit the store.



Thanks to Vytal for sponsoring this project! Vytal is an AI video workflow platform with reusable workflows for batch production, making video creation more accessible to content creators, training providers, and small teams. Visit Vytal.





 Why jev-chat?

When a message lands, the hard part is not typing. It is working out what the other person really means: are they upset, testing you, or just chatting? Most AI tools skip that step and go straight to writing a reply.

jev-chat judges first. Its Jev judgment model reads the conversation and works out the other person's intent, the risk in the moment and the best way to respond. Only then does it draft replies, and it fills the one you pick into the message box.

- Judge before writing — intent, risk, what they need and the best action come before any draft
- Checked, ranked replies — every draft is checked against that judgment and ranked, so the best fit comes first
- Works where you chat — WhatsApp, QQ, Feishu and X on Android; chat windows on Windows and macOS; any app through the iOS keyboard
- You press send — jev-chat only fills the input box and never sends on its own
- Your keys, no server of ours — requests go only to the model provider you configure, with your own key
- Open source — MIT licensed, five apps across Android, Windows, macOS and iOS

 Products

| Platform | Product | How you use it | Status |
| :--- | :--- | :--- | :--- |
| Android | Jev for WhatsApp (global edition) | Overlay on WhatsApp chats in English | Preview v0.1.0 |
| Android | Jev Chat Assistant (Chinese edition) | Overlay on QQ, Feishu, X and WhatsApp | v1.7 |
| Windows | Jev for Windows | Sits beside your chat window and reads it with screenshots and on-device OCR | Released |
| macOS | Jev for macOS | Overlay that reads the chat on screen and judges with a local model | Released (Apple Silicon) |
| iOS | Jev Keyboard | A custom keyboard: copy a message and see intent, risk and replies on the keyboard | Source only, build it yourself |

Both Android editions 

Top issues

Latest release

v1.7 — 支持 WhatsApp,分析结果改为常驻卡片 v1.7 2026-10-05
新增
- 支持 WhatsApp 一对一聊天(测试中):读取聊天、判断、候选回复、填入整条链路。群聊不读;WhatsApp 里不截屏。读取规则来自 @smgonthebeat(73)。
- 分析结果改成一张常驻小卡片:收起时只显示风险等级和对方意图,点标题展开看完整分析和候选回复,再点收起。卡片可以拖到屏幕任意位置,每个 App 分别记住位置。
- 候选回复下方新增「换一组」:判断不变,只重新生成一组候选,不会和上一组重复。
- 设置里可以选候选回复条数(1 / 2 / 3 条);只要 1 条时不做排序,出得更快。
- 判断接口新增「Vercel」「OpenCode Zen」两个预设。

改进
- 只有对方发来新消息才重新分析。上下翻聊天记录不会清空卡片、不会重新分析,也不会弹来弹去;更新时旧结果先留着,标题旁显示已等待的秒数。
- 候选回复的语言跟随对方:对方说英文就用英文回复。
- 判断和回复各自最多等 30 秒,超时就停下并给出「重试」。
- 无障碍服务的注册名称改为 Jev 自己的名称。
- 截屏识别更流畅(感谢 @ahxoeh);兼容更多中转网关(感谢 @xuancanhit99)。

修复
- 候选回复偶尔一直显示「生成中」。
- QQ 群聊上下滚动时分析面板卡成空白。

已知限制
- 从 v1.4 或更早版本升级后,需要到系统设置里重新打开一次 Jev 助手的无障碍开

Snapshot: 2026-11-10T09:35:25Z · Data license Apache-2.0 · JSON