Skip to content

狗头军师 Jev Chat

Chat copilot for Mac, Windows, Android and an iOS keyboard that reads chats, judges with Jev or DeepSeek and drafts replies.

ProjectprototypeMIT

Overview

狗头军师 Jev Chat (Goutou Junshi Jev Chat) is a Chinese chat copilot spun off from the author's earlier Goutou Junshi project. It reads the chat window, analyses the conversation and relationship, and drafts replies, with sending always left to the user. It offers a Mac source preview, a Windows preview zip, an Android debug APK (which cannot capture WeChat) and a new iOS keyboard preview; the Mac window module draws on jev-chat-jarvis-mac.

The iOS keyboard includes local Chinese pinyin typing with about 65,000 entries that works without network access, screenshot-based chat recognition via local Apple Vision OCR or cloud vision models after preview and confirmation, and an analysis step where DeepSeek or Jev judges and DeepSeek or OpenRouter writes replies. Analysis keeps possible intents, confidence, facts versus guesses versus unknowns, next steps, stop conditions, ranked candidates with reasons and costs, and a make-it-sound-more-like-me option; percentages are model estimates, not success rates. The README reports 225 offline checks passing for iOS and notes Windows and Android still need real-device testing. The main body is MIT licensed.

What the decision model decides

TypeQuestionAnswers
ChoiceWhat are the possible intents behind this message?(inferred)—
ChoiceWhich candidate reply ranks highest for the relationship goal?(inferred)—

How it works

  1. Capture chat text by screenshot OCR or screen reading
  2. User verifies the extracted text
  3. Jev or DeepSeek analyses intent and relationship
  4. A model drafts replies; candidates ranked
  5. User copies or fills a reply and sends manually

Strengths

  • Multi-platform previews
  • Separates facts, guesses and unknowns
  • Offline typing keyboard on iOS

Caveats

  • Preview builds; several platforms untested on devices
  • Android cannot read WeChat
  • Jev is one of two judgment options

At a glance

Type
Project
Status
prototype
Language
Python, Swift, Kotlin
Runtime
Mac/Windows overlay, Android app, iOS keyboard
Has code
yes
Has tests
yes
License
MIT

Use cases and integrations

How curated lists describe it

Frequently asked questions

What does Jev decide?

When selected, it judges intent and helps rank reply candidates.

Does it send messages?

No, sending is always the user's decision.

Which platforms are supported?

Mac, Windows and Android previews plus an iOS keyboard source preview.

Related projects

  • Windows WeChat reply assistant: local offline OCR reads the chat, Jev judges intent and tension and ranks three drafted replies to fill in.

    ProjectproductionPython
  • AstrBot plugin where Jev decides whether the bot should join a group conversation; the main model still writes the reply.

    ProjectproductionPython
  • iOS custom keyboard that shows a copied message's intent, risk level and candidate replies inside any chat app, filling but never sending.

    AppbetaSwift
  • Windows desktop chat assistant that OCRs a selected chat region, has Jev analyse the conversation, and ranks drafted replies for copying.

    ProjectbetaPython, TypeScript
  • macOS chat overlay that reads the current chat window and shows the incoming message's intent, risk and candidate replies, read-only.

    ProjectbetaPython
  • Unofficial fork of jev-chat-windows: WeChat overlay with local OCR and drafted replies, adding optional auto-send and a knowledge base.

    HookbetaPython