What exactly leaves my machine when I run an analysis?

Only redacted text and placeholders leave the machine; session names become generic labels, IDs and timestamps are dropped, and animated emoji become a placeholder. The draft path only runs when its switch is on.

How are the API key and local chat history protected?

Yes. The TypeSafe key comes only from an environment variable, the draft key is encrypted with Windows DPAPI or an override variable, and local history is AES-GCM encrypted under the current user.

Does rapid retriggering mix up two different chats?

Analysis runs in the background and only the newest request is kept, stale results cannot overwrite the current chat, and history is isolated by database session and message IDs. OCR is only a title fallback.

Will it auto-send a suggested reply into WeChat?

Open the target chat, press the hotkey, pick a 100 to 250 message window, and read the ranked candidates. Nothing is pasted or sent automatically, including the explicit do-not-reply option.

What platform and runtime does the tool require?

Windows with the WeChat desktop client, Python 3.12, and Git. Setup runs through setup.ps1, or you can download a verified Windows ZIP from Releases.

How many messages does it analyze, and how is it triggered?

Press Ctrl+Alt+J, then choose 100, 150, 200, or 250 recent messages. It reads the text and animated-emoji records from the local database for the conversation in the foreground.

Which drafting model does it recommend, and where?

DeepSeek, defaulting to DeepSeek-V4.1-Flash (deepseek-flash), against https://api.deepseek.com/chat/completions in non-thinking mode. The drafting key is encrypted with Windows DPAPI and the feature is off by default.

CommunityJev Tools & CLIs

Wechat Jev Assistant

Windows desktop aid that reads the local WeChat database for the foreground chat, strips identifiers, and asks Jev for stage, intent, tension, suggested moves, and a reply review.

See highlights

Open github.com

Stars: 19
Language: Python
Published by: Community
Author: yushen100
Last verified: Oct 2026

yushen100/wechat-jev-assistant

User repository on GitHub

View on GitHub

Windows 微信对话分析助手：本地读取、脱敏、TypeSafe Jev 判断与加密历史

Forks: 1
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Python

Highlights

Reads text plus animated-emoji placeholders from the local read-only database; images, voice, video, and files never upload.
Optional draft replies default off: up to three texts plus a do-not-reply choice, ranked by Jev with color coding.
Local anonymization replaces chat and member names with generic labels and strips phone, mail, ID, and card patterns.
Needs Windows, desktop WeChat, Python 3.12, and an env key; history stays in an encrypted on-machine database.

Watch out

Desensitization is automatic and cannot catch every identity clue, so only analyze sessions you are allowed to process. 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.

More Jev resources like this

jevclip

Video tool where Jev scores subtitle segments and code assembles a short highlights cut plus a de-watered full cut with timestamped summaries.