Open-source repositories gaining traction right now.
316 repositories

The Epstein Island LLM with a built in inference tool

ASTER: Agentic Scaling with Tool-integrated Extended Reasoning

Paper and experiments for reasoning by repeatedly refining a latent state.

Video generation skill for OpenClaw. Text-to-video, image-to-video, and video editing via natural language using xAI's Grok Imagine API.

[ICLR 2026] ODE-GS: Latent ODEs for Dynamic Scene Extrapolation with 3D Gaussian Splatting

这是一个面向 AstrBot 的 Gemini 语音桥接插件。它会优先识别语音消息,将 silk/amr/wav/mp3 自动转换并转写为文本,再交由框架按正常会话链路继续回复,实现“语音≈自动打字”的体验。插件本身不抢人格回复,支持群聊白名单、失败策略(放行/拦截/提示)、简洁或详细输出、模型名自动清洗、语音来源与说话人信息附加等配置,适合复杂插件生态下稳定接入语音能力

Python Library for running SHARE (Compress multiple LoRA adapters into a shared subspace)


JAX implementation of https://arxiv.org/abs/2602.04770

The Slopism manifesto, generated through democratic collaboration between 5 AI models and 1 human curator. Every step documented.

Recursive reasoning engine for AI agents and vector databases. Powered by MIT's RLM.

Official implementation of "Cooperative Edge Caching with Large Language Model in Wireless Networks".

基于AI大模型的本地照片/视频智能整理工具

52 Layers 4B(0.6B Active) MoE | Nemotron-3 Style + Teon Optimizer + Mamba-2 SSM + FP8 Training on H100

Vision AI Pro is a sophisticated computer vision application that transforms how businesses and developers interact with object detection technology. Built on Ultralytics' state-of-the-art YOLOv8 architecture, this platform delivers enterprise-grade performance with an exceptionally polished user experience.

Official implementation of **HPSG (Human Preference and Success-based Grasping)**, a deep reinforcement learning framework for **end-to-end 4-DoF robotic grasping from RGB-D observations**.

molecule.ai

Real-time gesture detection system using YOLO-Pose to trigger multimedia responses and track performance analytics.

基于 edge-tts 的文本转语音工具,支持流式播放和MP3导出

AlphaZero-inspired chess engine with MCTS self-play, neural network training, and Lichess bot integration


claude Code subprocess / Anthropic api 기반 Agent


商业应用的Bot,目前OpenClaw和NanoBot都是用于个人的,不太支持多用户,基于多租户多用户重新修改Bot

DinoxAI 笔记 Claude Skill - 支持创建、查询、更新笔记和 AI 问答


UFLDv2的超图神经网络改进版本,嵌入HyperACE模块,通过超图卷积建模车道间的高阶交互关系

Claude Code Agent Skill for searching and reading Reddit via OAuth2 API

🤖💰 Agent Economy Demo — Wallet gives agents capital. TiDB Cloud Zero gives agents memory. x402 payments + TiDB Cloud Zero + multi-agent collaboration.

minimum system supporting skills

Persistent memory for AI coding agents

A cross platform AI exam paper maker, now come with Alevel papers.

A framework for building nexuses of agents -- where intelligence emerges from interaction, not isolation.



🏦 Intelligent Banking Challan Processing System - Transform handwritten banking forms into digital data using deep learning and computer vision. Automates extraction of digits from deposit slips, withdrawal forms, and fund transfer documents with industry-leading accuracy.

macOS computer vision mapping hand gestures to mouse + keyboard control.


Observation and evaluation framework for multi-agent AI systems. Run experiments across coordination topologies, measure behavior and performance.

A professional, AI-driven system for cryptocurrency market data acquisition, technical analysis, and trading signal generation, powered by large-scale models.

AI Powered ATS Resume Analyzer built with Streamlit

Full transcripts for all my practical AI tutorials and interviews on https://www.youtube.com/@PeterYangYT

An advanced agentic ML orchestrator automating the end-to-end data science lifecycle. Integrates LLM reasoning with AutoML to deliver production-ready models and diagnostic PDF reports via automated Gmail delivery. Final project for Orange Digital Center (ODC) Egypt (AI Agents & LLMs track).


OpenAgentTrace from LearnChain

LLM service router and scheduler for single-GPU serving with llama-server

Build Your Own OpenClaw from Scratch in Python — 11 chapters + tech blog

AI Llama client with tool calling support from scratch coding without needing to OpenAI, LangChain, LlamaIndex modules and so on

A Very User-Friendly Ai Agent, which gives full Privacy to the Users, Self Authentication for Privacy and security of the User Data.


AI-powered end to end chess scanner project using YOLOv8 to convert board images into FEN strings. Includes a FastAPI web app and Lichess integration. ♟️🤖

selective DCT FineTuning

Certify Intel - AI-driven Competitive Intelligence Platform for CERTIFY Health

OmniVoice is a Python-based automated voice command assistant that integrates speech recognition, text-to-speech, browser automation, and YouTube playback using Selenium. It supports dynamic command processing, multi-mode interaction, and real-time voice-controlled task execution.

RAG 智能检索系统 一个基于向量数据库的智能文档检索系统,支持 PDF 导入、向量嵌入和智能检索功能。 基于hello_agents学习

An advanced agentic ML System automating the end-to-end data science lifecycle. Integrates LLM reasoning with AutoML to deliver production-ready models and diagnostic PDF reports via automated Gmail delivery. Final project for Orange Digital Center (ODC) Egypt (AI Agents & LLMs track).

Smart home assistant completely offline using openWakeWord for wake word, Whisper for STT, Piper for TTS and Ollama for LLM model inference with tool calling

Fractional Fourier Transform Loss Function

TCFP (Tensor Core Floating Point) Hardware-accelerated precision-enhanced FP8 training formats

FinOS: AI-Driven Financial Orchestrator
Paste a github.com URL. Submissions are reviewed before they are tracked.