OthersMidscenePWMidscene
ByteDance's open-source AI vision-driven UI automation framework automates interface interactions on any platform using natural language to describe operational intents, without the need to write selectors.
Runnable, documented projects worth studying and extending.
OthersMidscenePWByteDance's open-source AI vision-driven UI automation framework automates interface interactions on any platform using natural language to describe operational intents, without the need to write selectors.
OthersEvoMap/evolverPWAn AI Agent self-evolving engine based on gene expression programming (GEP). Through genomic evolution protocols, agents can autonomously evolve skill trees and optimize behavioral strategies. This week's GitHub hot list project: 4.3K Star.
Othersraphaelmansuy/edgequakePWA high-performance GraphRAG engine written in Rust, inspired by LightRAG. Transform documents into intelligent knowledge graphs to achieve higher-quality retrieval and enhanced generation. 1.9K Star, a rising star in the Rust ecosystem.
Othersz-lab/dflashPWDFlash: Lightning inference decoding technology based on block diffusion, significantly accelerating LLM inference speed. The new generation inference acceleration solution has 1.8K Stars this week.
OthersCowAgentPWA super AI assistant framework based on large models supports proactive thinking, task planning, skill creation and execution, and long-term memory. Supports integration across multiple platforms including WeChat, Feishu, DingTalk, and WeChat, with options for various models including OpenAI, Claude, Gemini, and DeepSeek. 43K+ Stars。
OthersQwen CodePWTongyi Qianwen is an open-source terminal AI coding assistant, similar to an open-source alternative to Claude Code. Supports file editing, shell command execution, code search, and more, and can connect to various LLM backends. This week, GitHub received 23.5k stars.
OthersEinoPWByteDance's CloudWeGo team is a Go language LLM application development framework that offers agent orchestration, RAG, tool invocation, and other capabilities, making it suitable for building production-grade AI applications in Go. This week, GitHub received 10.7k stars.
OthersT3 CodePWTheo (t3.gg) launched a TypeScript coding agent tool focused on AI-assisted development for full-stack TypeScript projects. Today's GitHub trending list: 9.5k stars.
OthersOpenAI Agents JSPWOpenAI officially released a lightweight multi-agent workflow framework (JavaScript/TypeScript version) that supports voice agent and tool call orchestration
OthersSpark - 3D Gaussian Splatting for THREE.jsPWA high-performance 3D Gaussian sputtering renderer for THREE.js, utilizing WebGL to achieve real-time 3DGS scene rendering, suitable for web-based 3D reconstruction and visualization
OthersDFlashPWA lightning inference decoding framework based on block diffusion models significantly accelerates LLM inference speed, a cutting-edge inference acceleration research project by the z-lab team
OthersGenericAgentPWThe self-evolving Agent framework automatically grows a skill tree starting from 3,300 lines of seed code, achieving complete system control capability while reducing token consumption by 6 times. It showcases a new paradigm for agents to learn and accumulate skills independently.
OthersVoiceBoxPWAn open-source speech synthesis studio supporting multilingual voice generation, voice cloning, and creative voice design. Provides an intuitive Studio interface, ideal for developers and content creators to quickly build voice applications.
OthersEvolverPWBased on GEP (Gene Expression Programming), the AI Agent self-evolution engine enables agents to independently discover and optimize strategies through genomic evolution protocols, evomap.ai production
OthersDFlashPWDFlash is a lightning inference decoding method based on block diffusion that can significantly accelerate LLM inference speed. Unlike traditional autoregressive inference decoding, DFlash uses a diffusion model to generate multiple token blocks in parallel, achieving more efficient inference acceleration.
OthersMOSS-TTSPWMOSS-TTS is developed by Fudan University MOSI. An open-source family of speech synthesis models from the AI and OpenMOSS team, supporting high-fidelity multilingual speech generation, multi-speaker dialogues, voice/character design, ambient sound effects, and real-time streaming TTS, covering complex and realistic scenarios.
OtherscompressOPWcompressO is a 100% free and open-source video/image compression tool that supports compressing any video and image to a minimum. It offers desktop clients for Mac, Windows, and Linux, with a clean and user-friendly interface.
OthersMakiPWAn efficient Rust AI coding agent focused on fast code generation and modification, with extremely low resource usage
OthersVoxCPM2PWTsinghua OpenBMB's open-source tokenizer-free TTS model supports multilingual speech synthesis, creative voice design, and high-fidelity voice cloning
OthersEvoMap EvolverPWAn AI Agent self-evolving engine based on gene expression programming (GEP). Through genomic evolution protocols, Agents can evolve autonomously like living organisms in strategies and capabilities, representing an innovative exploration in the direction of autonomous evolution.
OthersRalphPWThe autonomous AI Agent loop executor automatically plans tasks, writes code, and runs tests based on PRD documentation, executing loops until all requirements are completed. Turning AI coding from auxiliary to autonomous delivery.
OthersVoiceBoxPWAn open-source speech synthesis studio that supports unified management, comparative testing, and one-click deployment of multiple TTS models. Provides voice cloning, style transfer, and batch generation.
OthersKronosPWFinancial market language foundational model, specifically designed for financial time-series data. Supports price forecasting, market sentiment analysis, and trading signal generation.
AI and Agentsandrej-karpathy-skillsPWA CLAUDE.md best practices document, based on Andrej Karpathy's observations on LLM coding pitfalls. It helps Claude Code avoid common issues like over-engineering and illusion imports, significantly improving AI coding quality.