jev-chat-windows-laya
Fork of the jev-chat Windows app adding a key-free local LAYA judgment backend and a DPI-aware screen capture fix.
Overview
jev-chat-windows-laya, documented in Chinese with English summaries, is a fork of an MIT jev-chat project that changes only the Windows desktop version. Its main fix addresses high-DPI screens: at 150% or 175% scaling the app captured the wrong screen region because Win32 rectangles were in logical coordinates while screenshots were in physical pixels, which could send unrelated screen content such as browser or editor text to the configured TypeSafe or DeepSeek APIs. Declaring the process DPI-aware unifies coordinates, and the upstream unit tests (14/14) pass.
It also adds an optional local LAYA judgment engine via JEV_BACKEND=laya, so the app can run without a TypeSafe key (which was on a waitlist) and without chat text leaving the machine; the README warns its Chinese accuracy is insufficient and it is not enabled by default. A UI Automation criterion that can mistake internal window names for chat text is documented but not fixed. Each change's rationale and measurements are in FORK-NOTES.md. It is MIT licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What is the intent of the incoming chat message? (via Jev or local LAYA)(inferred) | — |
How it works
- Select the chat region (DPI-correct capture)
- OCR the visible messages
- Judge with TypeSafe Jev or local LAYA
- Show judgment and suggested replies
Strengths
- Fixes a privacy-relevant capture bug on scaled displays
- Optional fully local judgment with LAYA
- Documented evidence for each change
Caveats
- LAYA's Chinese accuracy is insufficient, per the README
- Windows only
Getting started
`git clone https://github.com/ZJemYoung/jev-chat-windows-laya.git`, run `windows\install.ps1` and `windows\start.ps1`; set `JEV_BACKEND=laya` to use the local engine.At a glance
- Type
- Project
- Status
- beta
- Language
- Python
- Runtime
- Windows desktop app
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
This project integrates Jev to provide structured decisions for its workflow. See the repository for implementation details.
logicrw/awesome-jev-projects
Frequently asked questions
What bug does this fork fix?
Wrong-region screen capture on high-DPI displays, which could send unrelated content to APIs.
Can it run without a TypeSafe key?
Yes, with the local LAYA backend, though Chinese accuracy is weaker.
What else changed?
Only the Windows desktop app; Android and question sets are untouched.
Related projects
Local browser agent that picks each step's operation and element with a Jev-shaped model such as a fine-tuned Laya, no cloud needed.
MCP serverbetaPythonOpen 2B browser-agent decision model fine-tuned from MiniCPM5 that serves the System One protocol for jev-ultrafast.
ProjectresearchPythonRuns jev-browser entirely on a local Jev-style engine (Qwen2.5 1.5B decisions, 0.5B typing) on an RTX 2060 SUPER, with no cloud API.
BenchmarkprototypePython, TypeScriptMangaba AI's local, calibrated typed-decision model for Portuguese text, trained to optimize the Brier score; proprietary since v5.
ProjectproductionPython- blink★ 22
Open one-pass typed-decision model with a dependency-free C99 runtime that also builds to a 66 KB WebAssembly module.
ServicebetaC - Foq★ 11
Local open alternative to Jev using a 1.58-bit ternary 8B model to answer booleans, choices, scores and Pydantic extraction in ~25 ms.
ProjectbetaPython