duckegg0623-create/jev-wechat-live
User repository on GitHub
用 TypeSafe Jev 实时解读微信消息的桌面浮层 —— 未完成的实验项目，判定准确率不达标，附完整踩坑记录
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jev Wechat Live
An abandoned WeChat overlay experiment published with its post-mortem: it reads WeChat's local encrypted database and asks Jev to show emotion, subtext, intent and a suggested reply. The pipeline works, but the README says judgment accuracy fell short.
- Stars
- 5
- Language
- Python
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- duckegg0623-create
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- Python
- Capabilities
- Open sourceRequires API key
Highlights
- An unfinished experiment published with its post-mortem: the decrypt-and-overlay pipeline works, but the README says Jev's subtext reading fell short.
- Reads WeChat's local encrypted database directly: session.db polled every 200 ms, message databases decrypted once, about 0.45 s for 80 MB, then WAL patches.
- Shows probability bars for subtext, emotion, tone score, true intent, relationship state, expectation and a suggested reply; roughly $0.0001 per judgment.
- KNOWN_ISSUES.md documents 11 real messages item by item, the A/B experiments, the failed fixes and the diagnosis, and asks someone to take over.
- Key extraction is out of scope: it needs all_keys.json exported by ylytdeng/wechat-decrypt because WeChat 4.1+ no longer caches keys in memory.
Quickstart
bash
pip install httpx pywin32
cd wechat-live
python live.py --dryWatch out
Copy reviewed against the Chinese README on 2026-10-15. The README is explicit that prediction accuracy did not meet the bar after prompt changes, added context and A/B comparisons, and it publishes no numeric accuracy. The README also mixes its API story — jev-chat is described as a Jev client while setup asks for an OpenRouter key — so check jev-chat/config.example.json.
Questions about this entry
Short answers drawn from the linked project or page.
Partly, by design honesty: the README says the pipeline — decrypting the database, merging message fragments, the overlay and debouncing — all works, but Jev's reading of subtext fell short of the accuracy bar. The author tried prompt changes, more context and A/B comparisons without fixing it, and published the project unfinished.
It reads WeChat's local encrypted database directly: polls session.db every 200 ms by mtime, decrypts about 0.5 MB in roughly 9 ms, and decrypts the message databases once at startup, about 0.45 s for 80 MB, then keeps them current through WAL patches, using a local_id cursor and real_sender_id to tell who sent each message.
For every new message: subtext, emotion state, tone closeness score, true intent, relationship state, what the other side expects, and a suggested reply, each as probabilities or a score. The README reports about one second end to end and roughly $0.0001 per judgment.
Keys exported by ylytdeng/wechat-decrypt as all_keys.json — the project does not extract keys itself — plus pip install httpx pywin32, and a config.json in both jev-chat (API key) and wechat-live (your wxid, the db_storage path, the keys file). The README notes WeChat 4.1+ no longer caches keys in memory.
The README is mixed: its table describes jev-chat as a Jev API client with a 7-question instant interpretation set, while the dependency table says httpx calls the OpenRouter Decisions API and setup asks for an OpenRouter key from openrouter.ai/keys. Check jev-chat/config.example.json for what it actually targets.
The author published it with the full post-mortem: KNOWN_ISSUES.md documents 11 real messages item by item, the A/B experiment data and the suspected causes, and the README says it is released in the hope that someone picks up the second half.
Effectively it is Windows software: it follows Weixin.exe's window with pywin32 (win32gui), reads the WeChat database directory, and ships a run_live.bat launcher. The README documents no macOS or Linux path.
More Jev resources like this
A Windows fork of the jev-chat WeChat assistant that fixes DPI-unaware screen capture on scaled displays and adds an optional key-free local laya judge — with the author's own measurements showing that local engine is not yet usable for Chinese chat.
Jev Tools & CLIs#community#catalog-import#logicrw
An Ollama-style local runner for System One decision models from Hugging Face: pull, serve and query them behind the same /v1/systemone API as TypeSafe's Jev, with a terminal model manager and a browser playground.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
nvkudva/ollajevPython
nvkudva/ollajevPythonA local app that reads a few lines of notes about your product and lets Jev pick its way to a finished landing page: Jev never writes text, it only chooses between options, and code builds the page from those picks.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
sagapranav/jev-pagesHTML
sagapranav/jev-pagesHTMLToolkit, knowledge base and local dashboard for building with Jev via OpenRouter Alpha Decisions or the TypeSafe Direct API: a zero-dependency SQLite request log, a localhost REST server and Python CLI demos.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
rileycarney/jevtoolsHTML
rileycarney/jevtoolsHTMLA Chrome extension that reorders the Gmail message list on screen only - unread first, critical messages on top - using Jev to score each row's urgency and criticality and to answer one boolean question per custom label.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
iamomiid/jev-inboxTypeScript
iamomiid/jev-inboxTypeScriptModeration for communities and apps, powered by Jev: every message gets a probability for spam, scam, harassment, nsfw, doxxing and more, plus rules in plain English. Thresholds and actions are yours, and flag-only is the default.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
ohernandezdev/jevmodPython
ohernandezdev/jevmodPythonFrom 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.