FerryCorleone/crush-monitor

What it solves

This project provides a tool to analyze text-based chat logs between you and a romantic interest or partner. It helps users interpret the emotional tone and intent behind messages, evaluates the quality of their own responses, and offers suggestions for improvement.

How it works

It is a local web application that processes chat logs (from WeChat, QQ, WhatsApp, etc.) by sending them to an AI model of your choice. You provide your own API key for services like Jev, OpenAI, DeepSeek, or others. The tool parses the chat history, displays the conversation in a familiar chat-bubble interface, and annotates messages with emotional and intent labels, a "likability" score, and a grade for your own replies.

Who it’s for

It is designed for individuals who want an AI-assisted perspective on their personal romantic communications. It is intended for users comfortable with running a local Node.js application and managing their own API keys.

Highlights

• Multi-Model Support: Compatible with specialized classification models like Jev and Liquid AI, as well as general-purpose LLMs like DeepSeek, Qwen, and OpenAI.
• Detailed Analysis: Provides granular insights including 12 types of emotions and 35 types of intent for each message.
• Response Grading: Evaluates your own messages on a scale from SSS to D and provides actionable advice for future replies.
• Privacy-Focused: Runs locally on your machine; chat history and analysis are stored in your browser's local database, not on a remote server.
• Visual Interface: Reconstructs chat logs into a readable format with analysis results displayed directly beneath the relevant messages.