AI / ML
94 个项目机器学习框架、数据工具，以及驱动现代 AI 应用的库。
在自有硬件上使用 VoiceStudio 实现本地化 ElevenLabs 级语音合成
VoiceStudio 将 ElevenLabs 级别的语音合成、克隆和转录功能带到您的自有硬件上——无需云端、无需订阅，只有开源的力量。
How to Run Cyclic AI Agents in Java
LangGraph4j brings cyclic AI agent orchestration to Java, with native Spring AI integration, state persistence, and a visual studio for debugging complex graphs.
Apache Maka Turns AI Agent Work into a Verifiable Event Log
Apache Maka brings event sourcing to AI agents, recording every action to an append-only SQLite log for full auditability and reproducibility.
AI Tools for Economists and Data Analysts in One Repository
OpenEcon's awesome-ai-for-economists repository brings together AI tools, MCP servers, and econometric libraries tailored for rigorous data work.
如何在配备 4 GB 显存的笔记本电脑上微调 8B 神经网络
使用 Soup 在消费级硬件上微调 8B 模型的实用指南——一款将分层流式处理和自动化流水线引入小显存 GPU 的 CLI 工具。
How to Create a 3D Traffic Projection from a Street Camera and Satellite Image
A practical approach to creating 3D traffic projections by combining street camera footage with satellite imagery, using Python and neural networks.
How to Confuse Vulnerability Scanners with Krawl
Krawl is a deception server that pretends to be a vulnerable web app, feeding scanners infinite trap pages and wasting their resources.
车牌识别：以每秒三千帧的速度
了解 fast-Plate-ocr，这是一款轻量级车牌识别库，通过紧凑的 CCT 模型和 ONNX Runtime 可实现高达每秒 14,000 次车牌识别的速度。
无需手动解析 DevTools，将任何网站转换为现成的 API 客户端
无需手动解析 DevTools，将任何网站转换为现成的 API 客户端。reverse-api-engineer 项目可自动化整个流程。
How to Check MCP Servers for Hidden Threats and Backdoors
Open-source mcp-scanner audits MCP servers for hidden threats using YARA, LLM-as-judge, and Docker sandboxing.
SkillsGate 为 AI Agent 技能和提示词带来秩序
SkillsGate 就像一个 AI 助手技能包管理器，让你一键同步 Cursor、Claude Code、Cline 等 20 多种工具的配置。
How to Stop Mindlessly Copying AI-Generated Code and Start Understanding It
AntiVibe forces AI to explain code architecture instead of just writing it. Adapts to your skill level and helps you understand what you're deploying.
如何让两个神经网络在 VS Code 中直接辩论你的架构
厌倦了手动切换 AI 模型来审查代码？认识一下 Mysti，这款 VS Code 扩展可以协调多个 CLI 助手协同工作、辩论架构，并以团队形式发现 bug。
使用 UniFace 库在单个 Python 包中构建完整人脸分析流程
UniFace 将人脸检测、对齐、嵌入、分割、头部姿态、视线追踪等功能整合到一个基于 ONNX 的 Python 接口中。