HSE Emotion Assistant

Multimodal emotion assistant with visual emotion analysis, Qwen realtime voice interaction, and local RAG knowledge base.

中文文档 | English Guide

Highlights

• Visual emotion analysis with local camera input
• Realtime voice assistant based on Qwen Omni / ASR / TTS models
• Local knowledge-base retrieval with embedding and rerank support
• Desktop GUI for text chat, voice mode, and environment setup

Quick Start

conda create -n hse_emotion python=3.10 -y
conda activate hse_emotion
pip install -r requirements_stable.txt
pip install -r requirements_llm_ui_v0.1.3.txt
copy .env.example .env
python -m hsemotion_ui

Before running the app, edit .env and set:

DASHSCOPE_API_KEY=your_key_here

Required Models

This project is currently configured around these DashScope / Qwen model IDs:

• qwen3.5-plus
• qwen3-omni-flash-realtime
• qwen3-asr-flash-realtime
• qwen3-tts-flash-realtime
• text-embedding-v4
• qwen3-rerank

Important Notes

• No API key is included in this repository
• You must provide your own DashScope / Qwen API key
• Runtime artifacts such as .env, logs, vector databases, and model caches are intentionally excluded

Documentation

• For full Chinese deployment instructions, see README_ZH.md
• For full English deployment instructions, see README_EN.md
• For environment setup with a small installer UI, run tools\run_env_installer.bat