A from-scratch x86_64 operating system where the only interface is natural language, allowing an LLM to manage hardware, write drivers, and create apps via 64 kernel syscalls.

DynamicVLA is a Vision-Language-Action model and dataset designed to enable robots to manipulate objects that are moving or changing state.

A collection of specifications and tools for 360-degree video and spatial audio to ensure immersive media is correctly recognized and played back.

A community-driven platform for C++ heterogeneous programming that enables a single binary to run across CPUs and GPUs from all major vendors.

A collection of production-grade robotics knowledge modules (SKILL.md files) that provide AI coding agents with the domain expertise needed to generate reliable, safe, and professional ROS1/ROS2 software.

A high-performance C++-based batched environment pool for reinforcement learning that significantly accelerates experience collection across various simulators.

A technical SEO auditing framework that turns Google's official documentation into an AI skill, allowing coding agents to audit live servers and fix technical SEO gaps.

rclip is a local, terminal‑native image search tool that uses OpenCLIP embeddings to let you find photos by natural‑language description, example image, or weighted combinations, with an optional interactive UI and support for many image formats.

A Python library for creating tabular synthetic data using machine learning to emulate real-world patterns while maintaining privacy.

An AI-native database designed for LLM applications that provides high-speed hybrid search across dense and sparse embeddings, tensors, and full-text data.

A lightweight and highly accurate wake word engine for building always-listening voice-enabled applications across IoT, mobile, and web platforms.

An open-source web application for AI model security assessments that uses probes to test for vulnerabilities and jailbreaks before deployment.

An AI software engineer that automatically picks up tickets from GitHub, Jira, or Linear, implements the code, and opens pull requests for review.

An AI-powered agent and MCP server that allows AI IDEs to autonomously design, deploy, and troubleshoot Home Assistant automations and dashboards using natural language.

An AI-powered e-ink photo frame that uses vision models to score and caption memories, displaying the best photo from "today in history" daily.

Hy3 is a 295B-parameter MoE model from Tencent designed for high-performance reasoning, stable tool-calling, and reduced hallucinations in productivity and agentic tasks.

A Claude/Codex skill that lets an LLM draft, review, and (optionally) post social‑media comments on Facebook, Instagram and Threads via a controlled Chrome session. It learns a brand’s voice, keeps a cryptographically‑signed ledger of every action, and includes extensive safety checks and test suites. Live posting is off by default; the repo supplies the automation framework, not the private content.

An open-source data catalog that provides a context layer for AI agents and teams to discover and understand data assets like tables, APIs, and queues.

A Python package for text normalization and inverse text normalization, used to prepare text for speech processing systems.

A modular, local-first AI agent framework designed for token efficiency and personal life management, supporting multiple interfaces and sandboxed shell execution.

A VS Code extension that integrates various domestic Chinese LLM providers and custom API-compatible models into GitHub Copilot Chat.

A lightweight, Kubernetes-native LLM inference platform that provides modular control and data planes for scalable, cost-efficient model serving.

An agent skill that enables AI coding agents to control BaoCut for automated video transcription, subtitle translation, and timeline editing via natural language.

An open-source marketplace repository for AI agent skills, providing a secure distribution system and CLI for adding capabilities to tools like Claude Code and Codex.

A monitoring and observability dashboard for AI coding agents that tracks tool calls, costs, and session trajectories across 32 different runtimes without requiring code changes.

An AI-powered novel writing system that ensures long-term narrative consistency by managing story elements as structured data and state snapshots rather than relying on LLM context windows.

A multi-agent simulation framework that uses LLMs to create interactive story worlds, featuring the ability to generate entire simulated environments from a single natural language prompt.

A cycle-level simulator for LLM serving infrastructure that models heterogeneous hardware and parallelism strategies to optimize inference performance.

A collection of modular agent skills for Claude Code and Codex that provide journalists and researchers with specialized workflows for source verification, records requests, and professional reporting.

A set of rules and scripts that enable AI Agents to configure, build, flash, and debug TI MSPM0 microcontrollers by directly interacting with SysConfig and hardware toolchains.

A collection of official and external plugins for GitHub Copilot that extend its capabilities through skills, hooks, and Model Context Protocol (MCP) servers.

A unified framework for high-fidelity egocentric and robotics data collection designed to streamline human demonstration data for robot learning.

A TUI-based model manager and server launcher that auto-detects inference backends and local model files to simplify serving LLMs.

Vui is an open‑source, real‑time voice‑assistant stack that stitches together speech‑to‑text, a local LLM, and a streaming TTS model (Vui Nano, 300 M parameters). It offers a WebRTC/WebSocket pipeline with turn‑taking, barge‑in, OpenAI Realtime‑API compatibility, a one‑shot voice‑note HTTP endpoint, pluggable ASR (faster‑whisper or Moonshine) and LLM back‑ends (Ollama, vLLM, OpenAI‑compatible), tool routing, voice cloning, and optional Claude‑style task delegation. Install via a one‑liner script or Docker‑compose; supports Linux + NVIDIA GPU, macOS + MLX, and mobile access via Cloudflare Tunnel or Tailscale.

A cross‑platform demo for the 4 B‑parameter Bonsai diffusion model, offering a web‑based studio and CLI for image generation on macOS (MLX) or NVIDIA GPUs (gemlite + HQQ).

A high-performance Settlers of Catan simulator and AI player designed for running large-scale simulations and training reinforcement learning agents.

Hypersim is a large synthetic indoor dataset (≈77 k HDR images, 1.9 TB) with per‑pixel geometry, semantics, and material channels, plus a V‑Ray‑based toolkit for generating and editing similar data. It’s designed for training and evaluating scene‑understanding models (depth, normals, segmentation, intrinsic image decomposition) and includes a ready‑made train/val/test split.

A standalone C++ implementation of the TRELLIS image-to-3D pipeline using GGML, enabling the generation of textured 3D models from images without Python at runtime.

An automated machine learning toolkit that serves as a drop-in replacement for scikit-learn estimators to automate model selection and hyperparameter tuning.

An AI-powered system that extracts Subject-Predicate-Object triplets from unstructured text to create interactive, queryable knowledge graphs.

A ComfyUI plugin for the Nunchaku inference engine that enables efficient 4-bit quantized model inference for image generation and editing models.

A large-scale RGB-D dataset of indoor scenes captured via mobile LiDAR, providing ground truth for 3D object detection and depth upsampling.

A collection of Model Context Protocol (MCP) servers that enable LLMs to query Neo4j databases, manage Aura cloud instances, and maintain persistent knowledge graph memory.

A foundation model for joint segmentation, detection, and recognition of biomedical objects across nine imaging modalities, supporting both 2D and 3D volumetric inference.

A state-of-the-art 3D promptable segmentation framework for medical images that allows users to refine volumetric masks using points, scribbles, and lasso prompts.

Merlin is a 3D vision-language foundation model and dataset for computed tomography that integrates CT scans with electronic health records and radiology reports for medical analysis.

A ComfyUI node that dynamically adjusts reserved VRAM in real-time to prevent OOM errors and optimize GPU memory usage.

EgoDex is a large-scale dataset and benchmark of egocentric videos paired with 3D pose annotations, designed to train AI and robots in dexterous tabletop manipulation.