User repository on GitHub
先回｜消息优先级助手 Windows + Android双端工具 JEV规则引擎 + AI自动判定消息优先级
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
CommunityJev Demos & Experiments
Jev Xianhui Windows
XianHui for Windows screenshots the WeChat window, reads the other side's messages with local RapidOCR, and asks Jev which one has to be answered now, then shows up to five threads as a colour-coded overlay that jumps you to the right chat app. It ranks, it never replies.
- Stars
- 1
- Language
- Python
- Published by
- Community
- Author
- jeffyuysw
- Last verified
- Oct 2026
Works forJev for Classification
Used inJev for Software & DevTools
- Languages
- Python
Highlights
- Three tiers, reply now in red, soon in yellow and can wait in grey, with at most five entries shown at once and a click switching to the right app.
- It only judges: no reply text is ever generated, and the overlay just tells you which thread to look at first.
- OCR runs locally through RapidOCR and screenshots stay in memory; message text is sent only at the moment a judgment is made.
- One judgment takes roughly a second and about $0.00004, since Jev only answers which to pick, how much and yes or no.
Watch out
The WeChat window must stay visible and unminimized: Windows exposes no system-level message notification interface, so a minimized or closed window produces a blank capture, and the README calls this a platform constraint rather than a bug. Parsing depends on the relative position of left and right message bubbles, so a significant WeChat layout change means recalibrating capture/parser.py. OCR reads text only, so images and stickers return nothing; the ranked list is cleared on restart by design; and the OCR model takes a second or two to load on first start. Message text does leave your machine at judgment time, through the single POST in core/jev_client.py that reaches Jev via OpenRouter. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
No, and that is the hard constraint. Windows has no system-level message notification interface, so the app captures the WeChat window and recognizes it locally; a minimized or closed window has nothing rendered, which means a blank screenshot. The README is explicit that this is an inherent limitation of the Windows screenshot mechanism rather than a defect, and the overlay warns you when it happens. The Android build avoids the problem by reading system notifications instead.
It reads only chat content on your own screen, from your own account, and never writes to WeChat. There is no injection, no hooking, no database decryption, no auto-send and no modification of WeChat data. It produces a ranking and nothing else, since no reply text is generated, so clicking an entry only takes you to that conversation. Use it on devices you own or have been authorized for.
The message text, at the moment a judgment is made. OCR itself is local through RapidOCR, screenshots never touch disk or logs, the API key is stored in %APPDATA%\JevPriority\config.json, and the README states there is no statistics, no telemetry and no reporting. The single POST in core/jev_client.py is the only network exit, which you can confirm by capturing packets yourself. The judging criteria and tier rules are shared with the Android build.
Reasonably brittle. Bubble parsing relies on the relative position of left and right messages, so a large layout change means recalibrating capture/parser.py. Images and sticker content read as nothing, because OCR handles text only. The ranked list is also not persisted, so restarting clears it, deliberately, to avoid leaving chat traces. Autostart at boot is available without administrator rights, and the app needs Python 3.10 or newer to run from source.
Download XianHui-Setup.exe and double-click to install, open the program, paste an OpenRouter key, keep the WeChat chat window visible, then click start monitoring. To run from source, use Python 3.10 or newer with a venv, run pip install -r requirements.txt and then python main.py.
At most five entries appear, colour-coded into three tiers: reply now in red, soon in yellow and can wait in grey. Clicking an entry switches to the matching chat app, auto-detecting WeChat, QQ, DingTalk and similar.
No, deliberately. The list is cleared on restart to avoid leaving chat traces behind. Autostart at boot can be enabled from settings without administrator rights.
More Jev resources like this
46GitHub stars
Open 150M encoder that reproduces the typed noul, choice, and score interface in one forward pass, trainable on a free GPU and runnable locally for nothing.
Jev Demos & Experiments#community#jevbest#catalog-import
941GitHub stars
Library that adds schema-guaranteed typed outputs to existing autoregressive LLMs without changing weights, served through SGLang or a hosted API.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
TypeLLM/TypeLLMPython
TypeLLM/TypeLLMPython55GitHub stars
Toolkit that makes a model behind an OpenAI-compatible endpoint or running in-process answer typed choice, score, and yes/no questions from next-token probabilities in Jev API shape.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
fidecastro/jevifyPython
fidecastro/jevifyPython164GitHub stars
End-to-end toolkit that synthesizes decision data, trains and calibrates candidate scorers, evaluates fixed splits, and serves Jev-compatible inference for interactive web decisions.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
zwliJay/jev-forgePython
zwliJay/jev-forgePython63GitHub stars
Claude Code kit that guards PreToolUse calls with a free pre-filter plus one Jev question for ambiguous cases, and adds sub-agent sizing, rechecks, search, daemon speedups, and monitoring.
Jev Demos & Experiments#community#awesomejev#catalog-import
Community
jonathanavis96/jev-kitPython
jonathanavis96/jev-kitPython70GitHub stars
Self-hosted proxy that records typed decisions, distils them into small heads on a frozen encoder, and answers locally when confident.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
bladedevoff/stuntdPython
bladedevoff/stuntdPythonFrom the Jev guides
Original write-ups on what this entry is for.
A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.
Jev Demos & Experiments#gaming#agents#control
Read article
From the Jev community
Posts from builders shipping with Jev right now.