Trending AI projects

A C++ edge AI engine that enables the deployment of video analytics and VLMs across Sophon, Rockchip, and x86 platforms via a visual orchestration layer.

A free, hands-on, Chinese-language (with English version) open-source tutorial that teaches developers how to build RAG (retrieval-augmented generation) applications — from the core concepts, through data preparation, indexing, retrieval, and evaluation, all the way to two full practice projects.

Code‑Codex is a Windows add‑on for the Codex/ChatGPT Desktop client that adds a local file‑tree, preview/editor pane, and a plug‑in marketplace for extra previews and visual effects. It ships as an installer or can be built from source with Rust and Node, and is MIT‑licensed.

Codex++ is a companion management tool for the official OpenAI Codex/ChatGPT desktop app. It adds provider switching (official login, API, or mixed), per-model context settings, session management, request retry on overload, UI enhancements (usage status bar, typing effects, extensions), multi-instance launching, and sub-agent collaboration — all configured through a separate window and applied when you launch the official app. It doesn't patch the official app's files. Windows/macOS, built with Rust + Tauri, AGPL-3.0.

An AI-powered automated short-video engine that generates scripts, visuals, and voiceovers from a single topic to create complete videos without manual editing.

OpenSandbox provides isolated, secure environments on your own infrastructure for AI agents to execute code, run commands, and operate browsers or desktops.

The Arena skill runs a tournament of up to 100 Claude sub‑agents, each given a unique reasoning‑mode/workflow/strategy “card”. Agents attack, defend, and are judged on a rubric until one answer survives, giving you a potentially stronger response. Install it in Claude Code, invoke via `/arena` (options for agents, quick mode, seed, wave), and the tool orchestrates everything with pure‑Python `bracket.py`. Free, MIT‑licensed; you only pay Claude token usage.

Gentle‑AI is a deterministic CLI wrapper that adds memory, structured delegation, and risk‑aware review to existing code‑generation agents, enabling trustworthy AI‑assisted development.

A Codex Skill that helps AI agents generate unique, hand-drawn, minimalist illustrations for Chinese-language content by turning abstract concepts into visual metaphors featuring a character named Xiaohei.

Cherry Studio is a free, cross‑platform desktop client that aggregates dozens of LLM providers (cloud, web, and local) into one UI, offering pre‑built AI assistants, rich document handling, and productivity tools, all under an AGPL‑3.0 license.

Cognee is an open‑source Python library that turns text, code, and other documents into a searchable knowledge graph, giving AI agents persistent, long‑term memory. It works locally with tiny extraction and embedding models, but can plug in OpenAI/Ollama LLMs for answer generation. Features include ingestion (`remember`), retrieval (`recall`), feedback (`improve`), and deletion (`forget`), plus SDKs for Python, TypeScript, Rust, a REST API, UI, Docker deployment, and plugins for Claude Code, Codex, etc. Ideal for building a “Company Brain” or adding memory to autonomous agents.

An educational proof-of-concept for an AI-powered hedge fund that uses LLM agents to simulate trading decisions through paper trading and backtesting.

A self-hosted gateway that turns Tencent CodeBuddy accounts into an OpenAI-compatible API, featuring multi-account management, automated daily tasks, and a web-based dashboard.

The      ...

An MCP server that enables AI assistants to interact with Home Assistant, allowing users to control devices and manage automations using natural language.

A Claude‑generated music‑video project. The repo contains AI‑written storyboard, animation code (p5.js), and a Node script to render frames with headless Chrome and ffmpeg, turning the song *I’m Upping My P(doom)* into an MP4.

Synaplan is an open-source, self-hosted AI platform (PHP/Symfony) that lets you chat with any model, search your documents, embed a chat widget, and see the cost of every answer. It runs with one command, supports hundreds of models, and integrates with web, mobile, desktop, Outlook, WhatsApp, email, and more. The README also compares it to Open WebUI, LibreChat, and Dify, and lists related ecosystem projects.

A high-performance C/C++ port of OpenAI's Whisper model for efficient, offline, on-device speech-to-text transcription across diverse hardware.

An opinionated, batteries-included agent harness built on LangGraph that provides out-of-the-box planning, context management, and task delegation for long-horizon work.

Onyx is a self-hosted knowledge and context layer that connects to over 50 applications to provide AI agents with secure, relevant access to internal company data. It enables advanced features like deep research, custom agents, and external actions while maintaining strict data sovereignty.

CraftStation is a Windows desktop app that lets you visually compose AI agents by pairing LLMs with their runtimes (harnesses). You can save any Model × Harness combo as a reusable Recipe, manage multiple provider accounts with quota monitoring, and use built‑in tools like browser automation and desktop actions. The app ships as a single installer/portable binary and is Apache‑2.0 licensed.

Agent Substrate is a Kubernetes‑native runtime that multiplexes millions of stateful “agent” sandboxes (actors) onto a small pool of worker pods, offering sub‑500 ms suspend/resume, full memory‑disk snapshots, and strong isolation via gVisor or micro‑VMs. It provides a control‑plane API, node supervisor, and Envoy‑based routing, enabling high‑density, low‑latency execution of AI agents (LangChain, Claude Code, etc.) and any stateful container workload.

OpenSRE is an open‑source framework for AI‑driven Site Reliability Engineering agents. It connects to 60+ observability, cloud, and incident‑management tools, fetches relevant context, runs a tool‑calling reasoning loop, and returns evidence‑backed diagnoses or remediation actions. Users can interact via an interactive shell, a headless CLI, or a Python API, and deploy the system via a managed cloud gateway, self‑hosted container, or EC2 AMI. The project is in public‑alpha (v0.1) and licensed under Apache 2.0.

Pi Web is a local browser interface for the pi coding agent. It reuses pi's session files and configuration, letting you browse and resume conversations, run agent turns, manage models and resources, and inspect project files entirely from a web page.

Zotero PDF2zh is a Zotero plug‑in (with a Python server) that translates PDFs to Chinese using LLM APIs while keeping formulas and layout. It offers bilingual side‑by‑side PDFs, crop‑and‑compare for double‑column papers, batch translation, a web dashboard for progress, and supports many LLM services (OpenAI‑compatible, DeepSeek, SiliconFlow, etc.). Install the plug‑in, run `server.py` (via uv or Conda), configure an API key, then right‑click PDFs in Zotero to translate.

Modly is an open‑source desktop app that runs AI models locally on your GPU to convert a single photo into a 3D mesh. It works on Windows, Linux and Apple‑Silicon macOS, supports NVIDIA, AMD (ROCm) and Apple GPUs, and offers an extensible plugin system for adding new 3D generation models. A graphical workflow editor and a Python CLI let you use it interactively or automate it.

SUPRA-Tex is a 3D generation system that creates relightable, PBR-textured assets from a single image by using surface-aligned observations and an Interleaved Flow Transformer to preserve fine visual details.

obsidian‑second‑brain is an open‑source tool that gives LLM agents persistent memory inside an Obsidian vault. It introduces a shared logbook (Crew) and 48 slash‑commands for saving meeting notes, ingesting files/URLs, auto‑reconciling contradictions, synthesising patterns, and pulling live research. The vault rewrites itself, tracks fact timestamps (OKM), and runs scheduled agents to keep knowledge fresh. Supports Claude Code, Grok Bot, Codex, Gemini, and other models, and installs via a single curl script.

A CLI tool that provides a verification layer for AI coding agents, allowing them to test their generated code against live applications and automatically fix bugs based on failure reports.

Papermorph is an agentic tool that automates the conversion of static PDFs into interactive, animated web books featuring narration, diagrams, and quizzes.

An Agent Skill that guides AI agents to generate simple, cute, and commercially viable IP mascot characters with consistent composition and style.

OpenRig is an open-source orchestration layer that turns AI coding agents (Claude Code, Codex) into a persistent, coordinated team defined in YAML and launched with one command, with a lead agent coordinating specialists and a TUI dashboard for monitoring.

A ComfyUI‑based recipe for generating NSFW images with Qwen‑Image 2.1, an NSFW LoRA, and a 136‑pose library. Provides scripts for text‑to‑image, pose‑preserving edits, and exact pose selection, plus a quick‑start command that reproduces a reference Civitai showcase image.

social-auto-upload is a multi-platform publishing tool that offers a unified CLI and agent skills for content creators and operators, supporting video, article, and image-text uploads to platforms like Douyin, Bilibili, Xiaohongshu, and more.

A context-engineering and spec-driven development framework that prevents AI coding agents from degrading due to context bloat by using a disciplined phase loop and fresh-context subagents.

Langflow is an open‑source visual builder for designing, testing, and deploying LLM‑driven agents and workflows, offering a web UI, Python‑editable components, API/MCP deployment, and Docker/desktop installers.

Presenton is an open‑source, self‑hosted application that generates fully editable PowerPoint decks using LLMs. It supports many AI providers (OpenAI, Anthropic, Gemini, Ollama, etc.), runs via Docker, a cross‑platform Electron desktop app, or in the cloud, and offers a drag‑and‑drop editor, API, and enterprise deployment options.

Orb is a macOS and iOS client that unifies access to local, private‑cloud, and provider‑hosted AI coding agents. It works with a self‑hosted Rust backend (sandboxed.sh) to store projects, missions, and coordinate remote execution, offering rich, resumable conversations and cross‑device sync.

Timestral provides inference‑only Python code for Mistral AI’s family of transformer‑based time‑series forecasting models (5M‑209M parameters). It supports multivariate probabilistic forecasts with covariates, returns 21 quantiles per step, and runs on CPU or GPU. Models are hosted on Hugging Face; the repo is Apache‑2.0 licensed, weights under a Modified MIT license.

MCP for Unity is an MIT‑licensed Unity package that connects LLMs (Claude, Codex, Gemini, local models, etc.) to the Unity Editor via the Model Context Protocol. It lets you create scenes, edit C# scripts, manage assets, run tests, and build games using natural‑language prompts from any MCP client. Install via Unity’s Package Manager, configure detected AI clients, and start issuing commands like “add a cube with a Rigidbody.” The project includes 50 ready‑made tool actions, multi‑instance routing, remote server support, and extensive documentation, with community channels on Discord and GitHub.

liyupi/ai-guide is a free, open-source AI knowledge base that collects tool guides, tutorials, reviews, and industry news. It aims to reduce the AI information gap and help anyone, from beginners to developers, learn practical AI skills and build monetisable products.

A Node.js CLI and markdown‑based spec (`DESIGN.md`) for defining design tokens (colors, typography, spacing, component styles) that AI coding agents can read. It validates files, compares versions, and exports tokens to Tailwind or W3C token formats.

Plannotator is an open‑source, locally‑run browser (and terminal) UI that lets you annotate plans, specs, code diffs and HTML artifacts generated by AI coding agents (Claude Code, Codex, Copilot CLI, Gemini, etc.). Annotations are sent back to the agent as structured feedback, enabling a human‑in‑the‑loop review step. It supports Git/GitHub/GitLab, integrates with VS Code, Obsidian, Bear, and offers encrypted link sharing. Installation is a one‑liner script that auto‑detects agents; all data stays local unless you explicitly share it.

Sentry is a Python library that monitors LLM‑agent execution, detects runtime failures, and supplies hard or soft repairs drawn from a learned playbook. It improves benchmark performance (≈37 % average gain) while keeping token usage low. The repo currently holds the paper and a minimal codebase; full release is pending.

A browser automation CLI that converts websites into deterministic interfaces for humans and AI agents, allowing them to operate websites using logged-in Chrome sessions.

Mobile MCP is an MCP server that lets AI agents (Claude, Gemini, etc.) control iOS/Android simulators, emulators, and real devices through a unified set of tools — tapping, typing, swiping, app management, and screen inspection via accessibility trees.

An AI-oriented quantitative investment platform that provides a full ML pipeline for data processing, model training, and back-testing in quantitative trading.

A minimal AI software engineering agent that uses only bash commands to solve coding tasks, providing a high-performance, hackable alternative to complex agent frameworks.