Open Source
Explore the latest AI open-source projects from GitHub and HuggingFace.
Explore the latest AI open-source projects from GitHub and HuggingFace.
Explore the latest AI open-source projects from GitHub and HuggingFace. Discover projects across categories like LLM, Vision, Audio, and more.
352 projects
lsdefine
GenericAgent is a 3,300-line open-source autonomous agent that grows a personal skill tree from every solved task—each new capability reduces future token consumption by up to 6x, supporting Claude, Gemini, Kimi, and MiniMax.
sgl-project
SGLang is a high-performance open-source serving framework for LLMs with RadixAttention prefix caching, zero-overhead CPU scheduling, and prefill-decode disaggregation — deployed across 400,000+ GPUs at xAI, LinkedIn, and Cursor.
agentscope-ai
AgentScope is an open-source Python multi-agent framework from Alibaba's research team, featuring OpenTelemetry observability, A2A and MCP protocol support, realtime voice agents, and human-in-the-loop interruption—designed for trustworthy production AI systems.
Google ADK is an open-source Python framework for building, evaluating, and deploying AI agents with code-first design, multi-agent composition, MCP tool integration, and native Vertex AI deployment support.
Magika is a Google-built open-source deep learning tool that detects 200+ file content types with ~99% accuracy in ~5ms on CPU, trained on 100 million samples and deployed in Gmail, Drive, and Safe Browsing.
obra
Superpowers is a 154,000-star MIT-licensed agentic methodology framework available across all major AI coding platforms, enforcing a six-stage development pipeline with mandatory TDD, git-worktree isolation, and two-stage code review to produce reliable, production-ready code.
Tracer-Cloud
OpenSRE is an open-source framework for deploying AI agents that autonomously investigate production incidents, correlating signals across 40+ observability, infrastructure, and incident management integrations with full LLM provider flexibility.
hugohe3
AI-powered PPTX generator that converts any document into natively editable PowerPoint presentations using real DrawingML shapes, supporting multiple LLM backends including Claude, GPT, Gemini, and Kimi.
xming521
WeClone is an open-source one-stop solution for creating an AI digital twin from personal chat history. Fine-tunes LLMs using LLaMA Factory with Microsoft Presidio PII filtering, multimodal (text + image) training, and deployment to Telegram, Discord, Slack, and WeChat.
Open-LLM-VTuber
Open-LLM-VTuber lets you run a customizable, voice-interactive AI companion entirely on your own hardware — no cloud required — pairing major LLMs with animated Live2D avatars, real-time speech recognition, and hands-free conversation.
NVIDIA
PersonaPlex is NVIDIA's open-source real-time speech-to-speech system that enables full-duplex conversation with text-defined personas and audio voice conditioning, built on the Moshi architecture with a 7B Helium LLM backbone.
bytedance
Protenix is ByteDance's fully open-source biomolecular structure prediction system that outperforms AlphaFold3 across diverse benchmark sets. The v2 release (April 2026) scales to 464M parameters with full training pipeline access under Apache 2.0.