Skip to content

Releases: LessUp/yolo-toys

v3.0.0

16 Apr 02:14

Choose a tag to compare

🎯 YOLO-Toys v3.0.0

English

This is a major architectural refactor release with significant improvements in code maintainability, extensibility, and test coverage.

🚀 Highlights

  • Strategy Pattern Architecture — Model handlers are now modular and extensible
  • Pydantic Settings — Unified environment configuration management
  • Modern FastAPI Lifespan — Replaced deprecated on_event with context managers
  • Test Coverage — Increased from 6 to 16 test cases
  • Documentation Overhaul — Comprehensive technical docs and contributing guides

✨ New Features

  • Multi-model support: YOLO (detect/segment/pose), DETR, OWL-ViT, Grounding DINO, BLIP (Caption/VQA)
  • WebSocket real-time streaming with dynamic config updates
  • FP16 acceleration for NVIDIA GPUs
  • Auto device selection (CUDA > MPS > CPU)

🔧 Changes

  • model_manager.py: 814 lines → ~100 lines (strategy pattern)
  • main.py: 410 lines → 91 lines (route extraction)
  • Route separation into app/routes.py
  • Structured logging replaces print() statements

📚 Documentation

  • CLAUDE.md — Complete Claude Code development guide
  • CONTRIBUTING.md — Development workflow and conventions
  • SECURITY.md — Security policy and reporting
  • CODE_OF_CONDUCT.md — Contributor Covenant 2.1
  • docs/README.md — 8-chapter technical documentation

中文

这是一个重大架构重构版本,显著提升了代码可维护性、可扩展性和测试覆盖率。

🚀 亮点

  • 策略模式架构 — 模型处理器模块化,易于扩展
  • Pydantic Settings — 统一环境变量配置管理
  • 现代 FastAPI Lifespan — 使用上下文管理器替代已弃用的 on_event
  • 测试覆盖率 — 从 6 个增加到 16 个测试用例
  • 文档全面重构 — 完整的技术文档和贡献指南

✨ 新功能

  • 多模型支持:YOLO (检测/分割/姿态)、DETR、OWL-ViT、Grounding DINO、BLIP (图像描述/视觉问答)
  • WebSocket 实时流式传输,支持动态配置更新
  • NVIDIA GPU FP16 加速
  • 自动设备选择 (CUDA > MPS > CPU)

🔧 变更

  • model_manager.py: 814 行 → ~100 行 (策略模式)
  • main.py: 410 行 → 91 行 (路由提取)
  • 路由分离到 app/routes.py
  • 结构化日志替代 print() 语句

📚 文档

  • CLAUDE.md — 完整的 Claude Code 开发指南
  • CONTRIBUTING.md — 开发流程和规范
  • SECURITY.md — 安全策略和报告流程
  • CODE_OF_CONDUCT.md — Contributor Covenant 2.1
  • docs/README.md — 8 章技术文档

📦 Installation

# Clone and install
git clone https://github.com/LessUp/yolo-toys.git
cd yolo-toys
pip install -r requirements.txt

# Run
python -m uvicorn app.main:app --host 0.0.0.0 --port 8000

🐳 Docker

docker-compose up --build -d

Full Changelog: v2.0.0...v3.0.0