Trending AI projects

A TypeScript library for orchestrating AI coding agents in isolated sandboxes, automating container management and git branch merging.

A recipe to serve the DeepSeek-V4.1-Flash-1017 model on two NVIDIA DGX Sparks using TensorFold, with quantisation, speculative decoding, and an OpenAI-compatible API.

A project context and engineering control plane that provides AI coding agents with durable project knowledge, governance, and risk-driven quality gates.

Compound Engineering is a cross‑platform plugin that equips AI coding agents with 36 skills to run a structured “brainstorm → plan → work → simplify → review → compound” loop, persisting learned insights for future AI‑driven development.

A privacy-first personal knowledge management system that uses block-based Markdown editing, two-way linking, and AI integration to help users organize and retrieve information.

A collection of design templates and visual rules that enable AI agents to generate consistent, high-quality data visualizations, slide decks, and reports.

vLLM‑Omni is an open‑source extension of the vLLM inference engine that adds fast, scalable serving for multimodal models (text, image, audio, video, and action). It supports both autoregressive and diffusion‑style architectures, provides a full‑duplex realtime API, works across GPUs, NPUs, and other accelerators, and ships ready‑to‑run recipes for many popular Hugging Face models. The project is production‑grade, Apache‑2.0 licensed, and backed by documentation, a Slack community, and recent technical reports.

A custom Swift and Metal inference engine that allows running the 26-billion-parameter Gemma 4 model on 8 GB Apple Silicon Macs by streaming model experts from disk.

A curated set of example plugins for OpenAI's Codex coding assistant, each with a required manifest and optional skill/configuration files, distributed via a marketplace catalog and covering tools like Figma, Notion, iOS/macOS/web app development, and Expo.

anarlog is an open‑source, local‑first meeting‑note app that records audio on your device, transcribes it with on‑device or user‑chosen STT models, and lets you generate AI summaries or chat using any LLM you control. All notes and recordings stay in a local SQLite database (exportable to Markdown) unless you opt‑in to encrypted CloudSync. Built with Tauri (Rust backend) and React/TypeScript, it runs on macOS, Windows, Linux, and has mobile/watch companions. MIT‑licensed for the community code, with optional commercial enterprise components.

An open-source data middle platform for enterprise data governance and development, featuring ETL, metadata management, and AI-powered Text2SQL data Q&A.

An official MCP server from Hex-Rays that connects AI agents to the IDA disassembler, letting them reverse engineer binaries and interact with IDA databases programmatically.

Lap is an open‑source, cross‑platform desktop photo manager that keeps all your images on your own disk. It offers folder‑first organization, local AI search (text‑prompt, similarity, face clustering) via ONNX Runtime, bulk editing, collage creation, and support for 60+ image, RAW and video formats. Built with Tauri + Rust, Vue, and SQLite, it runs on macOS, Windows and Linux and is released under GPL‑3.0.

A GitHub CLI extension that enables developers to define and run AI-powered automation workflows within GitHub Actions, supporting tasks like issue triage and code review.

DSPy is a framework that allows developers to build modular AI systems using Python code instead of manual prompt engineering, automatically optimizing prompts and model weights for better performance.

nono is a lightweight, cross‑platform runtime that sandbox‑isolates AI agents and the tools they call. It enforces per‑tool file, network and credential policies, verifies assets with Sigstore, logs tamper‑evident audits, and supports rollback. Installable via a single binary (curl, Homebrew, Nix, etc.), it runs unprivileged on laptops, containers or CI, and offers Rust/Python bindings for integration.

NVIDIA‑Omniverse usd‑brep adds true CAD‑style B‑Rep geometry support to Pixar’s USD format. It ships a C++ library, Python bindings, a USD schema plugin, and a small GUI for inspecting models. Build with Visual Studio on Windows or build‑essential on Linux, then use the provided `usd.heal_file` utility or the Python API to work with solid models inside USD scenes.

tiny-llm is a hands-on course for systems engineers to understand LLM inference end to end. You build a Qwen3 model from scratch using MLX arrays, implement operators yourself, and progress from basic matmul to a mini-vLLM and a coding agent — all on Apple silicon.

An AI-powered cross-platform e-book reader that supports multiple formats and provides AI-driven summaries, mind maps, and translation tools.

AI Companion Kit is a collection of modular tools that give a chat AI human-like memory, typing behavior, and the ability to control a smartphone and post to QQ Zone on its own. It solves the two common shortcomings of AI assistants: they feel robotic and they can't interact with the real world beyond text.

claude‑blog is a Claude Code skill suite that automates full‑lifecycle blog creation – writing, SEO, schema, multilingual publishing, audio, and quality‑gate checks – via a `/blog` command set backed by a 5‑gate delivery contract.

A toolkit of specialized AI Agent instructions (Skills) for business diagnosis, content creation, and knowledge management, based on a structured library of professional methodologies.

A free, organized hub for learning generative AI — research, courses, interview prep, and tools — structured around your goal (use, build, understand, or interview prep) with beginner-to-advanced levels for each path.

kvcached is a Python library/daemon that adds virtual‑memory‑style, on‑demand KV‑cache allocation to LLM serving engines (SGLang, vLLM). It lets multiple models share a GPU, puts idle models to sleep, and reduces time‑to‑first‑token by up to 28×, saving GPU costs.

A TypeScript-based runtime for coding agents that adds a supervisory layer to ensure tasks are completed with verifiable evidence rather than just model-generated claims.

An automated pipeline that turns a single AI-generated character still into a full set of game-ready, keyed, and timed sprite animations for engines like Godot and Aseprite.

An open-source data framework for building agentic applications by connecting LLMs to private data sources through ingestion, indexing, and retrieval.

A framework and toolset for AI agents to eliminate "AI-style" Chinese writing by enforcing human-like rhythms, factual substance, and structural checks.

A community-maintained catalogue of 6,000+ Claude Code mods that scans GitHub, records each mod's access permissions, and provides a searchable directory and a mod-finder assistant.

A multi-agent framework that scales test-time compute through parallel strategy exploration and structured noise injection to solve complex problems without getting stuck in cognitive loops.

GEOFlow is an open‑source Laravel platform that unifies knowledge‑base management, AI‑generated SEO content, automated quality checks, human review, multi‑site publishing and analytics for enterprise websites. It runs on PHP 8.3+, PostgreSQL (pgvector), Redis and Docker, and is released under AGPL‑3.0 with optional commercial licensing.

A theme plugin for DeepSeek Harness that replicates the visual style and interaction experience of the Claude Code desktop application, including animated status characters.

An MCP bridge that connects AI assistants to the TradingView Desktop app via Chrome DevTools Protocol for AI-assisted chart analysis and Pine Script development.

A self-hosted gateway for managing and distributing access to multiple Codex AI accounts, featuring account grouping, usage tracking, and a centralized management dashboard.

uniTerm is a cross‑platform desktop terminal that supports 30+ protocols (SSH, RDP, SFTP, databases, Kubernetes, Docker, etc.) and includes an AI Agent capable of autonomously planning and executing multi‑turn shell commands via Anthropic or OpenAI compatible models. Built with Go (Wails v3) and Vue 3, it runs on Windows, macOS, Linux and Android, offers split‑pane workspaces, cloud sync, theming, and is released under Apache 2.0.

Activepieces is an extensible, open-source automation platform (Zapier replacement) where every integration is a TypeScript npm package and also automatically exposed as an MCP server for LLM tools like Claude Desktop and Cursor. It combines a no-code builder with AI-first features such as agent support, human-in-the-loop steps, and an AI SDK, while remaining fully self-hostable.

SoloMD is a local‑first markdown editor with a built‑in AI agent panel, MCP server, and scheduled “recipes”. It lets you chat with, query, and have LLMs edit your notes, while keeping all data, keys, and git history on your machine. Supports 14 AI providers, Ollama, AutoGit sandboxing, optional encrypted Git sync, and runs on macOS, Windows, Linux, iPad and Android.

LocalMiniDrama is a local-first AI short drama creation tool that takes a story synopsis through script, characters, scenes, storyboards, images, and video generation, finally composing full episodes — all on your own machine with your own AI API keys and no cloud upload.

A collection of opinionated Oxlint rules for TypeScript and JavaScript that reject low-signal patterns and enforce stricter type safety and code efficiency.

3Dviz Pro Max is an AI‑agent skill that lets large language models turn a natural‑language description into a complete Three.js scene. It provides a ten‑step workflow, a catalog of 223 recipes, 22 proven Three.js kit blueprints, style defaults, and Python helpers that render the scene in headless Chromium, capture PNG frames, and feed the observations back to the model for refinement. Install by linking the `skills/3dviz-pro-max/` folder into Claude Code or Codex, then run the bundled examples (37 runnable demos) with Vite. The project is MIT‑licensed and focuses on reproducible, evidence‑backed 3D visualisation rather than claiming that every recipe is fully rendered.

Gufo is an open‑source, AMD‑Strix Halo‑optimised inference engine that serves large language, vision and audio models via an OpenAI‑compatible API, with containerised deployment and speculative decoding for high throughput.

A tool that extracts a complete design system from any live website, generating design tokens, Tailwind configs, and Figma variables for developers and designers.

A remote control system for AI agents that lets users dispatch tasks to agents running on their PC via a wearable voice-interface device and an iOS app.

An open-source multimodal AI creative workbench that combines an infinite canvas with AI image, video, and audio generation tools to streamline visual iteration.

A 9B vision model and toolset for extracting structured JSON from PDFs and images using schema-constrained decoding to guarantee valid output.

An open-source Python library for quickly building and sharing web applications for machine learning models and APIs without requiring front-end development experience.

A set of 9 AI agent skills for Claude Code and Codex that automate Instagram content creation, humanize AI text, and handle media scheduling via Publora.

A tool that turns short voiceover lines into editorial halftone paper-collage B-roll animations using a three-stage approval workflow and Gemini Omni Flash.