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
Cross-platform assistant that ranks pending messages by urgency with one Jev call per item, shown in a draggable overlay that snaps to the screen edge. Android reads system notifications via NotificationListenerService; Windows screenshots the WeChat window and OCRs it locally. It judges, never replies.
- Stars
- 1
- Language
- Kotlin
- License
- MIT
- Published by
- Community
- Author
- jeffyuysw
- Last verified
- Oct 2026
Works forJev for Classification
Used inJev for Software & DevTools
Highlights
- Four typed questions per message (urgency 1-9, reply now yes/no, category, reason), bucketed identically on both platforms into four tiers.
- One judgment costs about a second and roughly $0.00004, through an OpenRouter key kept in app-private storage and sent only with judgments.
- Android reads system notifications via the official NotificationListenerService; Windows has no such API, so it screenshots and OCRs locally.
- Judges only: permissions are INTERNET, SYSTEM_ALERT_WINDOW, POST_NOTIFICATIONS and FOREGROUND_SERVICE, and it never sends a reply.
Watch out
Setup is manual by design. Notification access and overlay permission can only be granted in Android system settings, no app can open them for you, and Chinese OEM ROMs freeze background processes, so Xiaomi and HyperOS need autostart plus an unrestricted battery policy; autostart itself is a vendor-private switch with no read API, which is why the in-app badge can only reflect the battery setting. Notifications carry only the opening fragment, so long messages are judged on their first line, and jumping to a conversation depends on each app's own contentIntent, with some devices needing a background pop-up permission. The Windows build lives in a separate repository and requires the WeChat window to stay visible rather than minimized. Feishu is listed as supported but untested on Android. Changing judgment behavior means editing jev/PriorityQuestions.kt and rebuilding, and the per-judgment cost figure is the author's own. 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.
Just notifications, through the official NotificationListenerService on Android. No root, no faked system services, no bypassing app security and no reading chat databases; the declared permissions are INTERNET, SYSTEM_ALERT_WINDOW, POST_NOTIFICATIONS and FOREGROUND_SERVICE. Message text leaves the device only for the moment of a judgment, in a single POST, and never reaches a log.
The README puts one judgment at about a second and around $0.00004, routed through an OpenRouter key stored in the app's private storage and sent only with judgment requests. No key is hardcoded in the repository. Judgment history is local SQLite capped at 2000 rows with one-tap clear, backups are disabled, and there is no analytics or telemetry.
Only after manual setup. Notification access and overlay permission can be granted only from Android system settings, and Xiaomi and HyperOS additionally require autostart and an unrestricted battery policy or the service is frozen and no messages arrive. Autostart is a vendor-private switch with no readable API, so the badge in the app can only tell you whether the battery setting is right.
One file on Android. jev/PriorityQuestions.kt holds the four questions (a 1-9 urgency score, a reply-now Noul, a category Choice and a reason) and the buckets, which are identical on Windows: reply now or 7 and above, 4 and above, below 4, or still judging. Changing it means rebuilding the app.
WeChat, QQ and Feishu, with Feishu listed but untested. It reads the notification shade through the official NotificationListenerService, so any app that posts a notification can be picked up, including in the background or on the lock screen. It needs Android 8.0 or newer.
Each message goes out in one request that asks urgency on a 1-9 score, a reply-now yes or no, a category choice such as work, personal or promotional, and a reason such as a deadline or the other side waiting. The four tiers are identical on both platforms.
Records live in a local SQLite database capped at 2000 rows with the oldest trimmed, clearable in one tap from the analysis page. It is never uploaded, system backup is disabled with allowBackup=false, and there are no analytics, tracking or reporting.
More Jev resources like this
15GitHub stars
Android accessibility app that reads visible WeChat single-chat text and asks Jev for intent, emotion, urgency, and posture, showing exactly 3 Toasts without writing replies or changing WeChat.
Jev Demos & Experiments#community#awesomejev#catalog-import
Community
Nisaka520/JevBystanderKotlin
Nisaka520/JevBystanderKotlin36GitHub stars
WeChat relationship assistant that combines screen reading, one 8-question Jev judgment with a strategy score, per-contact memory, and three chat-model replies behind a resident score bar.
Jev Demos & Experiments#community#awesomejev#catalog-import
CommunityNisaka520/JevGuideKotlin
Nisaka520/JevGuideKotlin7GitHub stars
Android noise gate that asks a model whether each notification or SMS is noise and suppresses only explicit flags. Uncertain paths fail open, verification codes never leave the device, and sender blacklists stay local.
Jev Demos & Experiments#community#awesomejev#catalog-import
Community
ufec/jev-block-android-adKotlin
ufec/jev-block-android-adKotlin46GitHub 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/TypeLLMPython98GitHub stars
Self-hosted Docker service and web UI that packages Laya structured-decision inference behind a compatible System One endpoint.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
1Panel-dev/laya-serverPython
1Panel-dev/laya-serverPythonFrom 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.