Crush Monitor
Local chat analyser that labels each WeChat message's emotion and intent with Jev and grades how your replies landed.
Overview
Crush Monitor is a small local tool, documented in Chinese with an English README, that analyses a chat with a crush or partner. It renders a WeChat-style interface and shows two label rows under each message: the top three of 12 emotion classes and the top three of 35 intent classes. It also shows an overall affection-signal score and rates your own replies on a scale from SSS to D with suggestions for the next step.
The author chose Jev because the task is per-sentence judgment rather than long-form generation, and emotion and intent can be evaluated in parallel in one request. Jev is recommended, reachable via TypeSafe, OpenCode Console, Vercel AI Gateway or OpenRouter, with OpenAI Decisions and ordinary LLMs such as DeepSeek or Qwen as alternatives. The README is candid that the AI does not know the real relationship and results are for fun and reference.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which emotion does this message express? | — |
| Choice | What is the intent of this message? | — |
| Score | How well did this reply land?(inferred) | SSS · S · A · B · C · D |
How it works
- Paste a chat record into the local web app
- Messages are batched and sent to Jev (or another provider)
- Top emotion and intent labels with probabilities are shown per message
- Affection score and reply ratings are computed; results persist locally
Strengths
- Runs locally with your own key
- Parallel emotion and intent judgment in one request
- Supports several providers including Jev through multiple gateways
Caveats
- No accuracy benchmark for this scenario
- Interpretations are for entertainment and reference
- Chat content is sent to the chosen provider
Getting started
Run it locally, then choose a platform and enter your key in the web page; no config file edits are needed.At a glance
- Type
- Project
- Status
- prototype
- Language
- JavaScript
- Runtime
- local web app
- Has code
- yes
Use cases and integrations
How curated lists describe it
Reads a WeChat conversation and labels each message with emotion and intent, rating how your own replies landed. Local, with your own key.
MrJev/awesome-jev
Frequently asked questions
What does it label?
Each message's top emotions (of 12) and intents (of 35), plus a rating of your replies.
Do I need Jev?
Jev is recommended, but DeepSeek, Qwen or OpenAI Decisions can be used instead.
Is the analysis accurate?
The README says there is no accuracy ranking for this scenario and results are for reference.
Related projects
DeepSeek Harness plugin that classifies each user message's emotion, urgency and department with Jev and injects it as context.
PluginbetaTypeScriptZero-dependency Android app that judges the visible WeChat message's intent, emotion, urgency and reply posture with Jev.
AppbetaJavaAndroid WeChat assistant where Jev judges the chat and scores the relationship, and a chat model drafts three replies.
AppbetaKotlinWeChat plugin that classifies a long-pressed message's intent, emotion, urgency and suggested reply posture with Jev.
PluginbetaJava- jev-chat-windows★ 802
Windows WeChat reply assistant: local offline OCR reads the chat, Jev judges intent and tension and ranks three drafted replies to fill in.
ProjectproductionPython Android chat co-pilot that reads the active chat via Accessibility and returns typed verdicts on intent, emotion and risk; never replies.
AppbetaKotlin