Trending AI projects

Osaurus is a native macOS app that lets you run private AI agents locally. Agents have their own memory, can execute code in an isolated Linux VM or macOS sandbox, call tools (file system, browser, Apple apps, image generation), and communicate securely via encrypted channels. The platform works with local models (MLX, Apple Foundation Models, Liquid AI) or cloud APIs (OpenAI, Anthropic, etc.), and includes privacy‑filtering, cryptographic identities, plugin support, voice input, and system‑wide shortcuts. Install via Homebrew or a .dmg; requires macOS 15.5+ on Apple Silicon.

A centralized repository of open-source tutorials, guides, and learning materials designed to help developers learn and implement GPU accelerated computing.

A deployment recipe for running DeepSeek-V4-Flash-Vision-Exp on two-node DGX Spark clusters, featuring 1M-token context support, speculative decoding, and native image processing.

Postiz Agent provides plugin manifests that let AI agents (ChatGPT, Claude, Gemini, etc.) call the Postiz CLI to discover social‑media integrations, upload media, schedule posts, and fetch analytics across 28+ platforms.

A free, wiki-style directory of AI tools that cost nothing to use, split by whether signup is required, covering chat, media, voice, IDE, and API categories — with a clear warning that free services may log your data or vanish anytime.

Box is a security-hardened Android application that enables users to run powerful AI models, including LLMs and image generators, entirely offline using local hardware acceleration.

A framework for building LLM-powered applications that simplifies the integration of models, data sources, and components through a modular, interoperable architecture.

NaturalVoiceSAPIAdapter is a Windows SAPI 5 text‑to‑speech engine that lets any program use Microsoft’s neural (natural) voices—from Windows 11 Narrator, Edge’s Read Aloud, or Azure Speech Service—by installing a small DLL and optional Azure key.

director is an AI‑agent skill that guides a language model through the full video‑production pipeline—script, visual style, characters, voice, shot‑by‑shot prompts—and uses LibTV or 即梦 CLI tools to generate ready‑to‑edit video clips. It supports three modes (Animated Explainer, Storytime Animation, Cinematic Drama) and is installed as a skill directory for agents like Codex.

Free hands-on course (from DataTalks.Club) teaching disciplined AI-assisted software development — using coding agents/assistants to plan, build, test, deploy, observe, and audit real apps, with a 5-module curriculum and a capstone project. Certificate only via the live cohort (starts Aug 31, 2026).

Oxigraph is a high-performance, standards-compliant graph database and toolkit for RDF data and SPARQL queries, available as a Rust library, Python package, and JavaScript module.

gflow-cli is a Python CLI (and optional MCP server) that automates Google Flow’s Veo video‑generation and Imagen image‑generation via a real Chrome session. It supports batch jobs, reusable characters, prompt‑rewriting tools, and can pipe results to cloud storage. Designed for developers and AI‑agent integration, it runs in alpha and requires a headed Playwright‑Chromium browser.

RTAB-Map is a robotic mapping and localization library and standalone application used to build maps of unknown environments.

OpenLLM is a Python tool that lets you run open‑source LLMs (e.g., Llama 3, Qwen 2.5, Mistral) locally and expose them through OpenAI‑compatible REST APIs. With a single command you can start a GPU‑accelerated server, use a built‑in web chat UI, and deploy the service via Docker or Kubernetes. It supports a catalog of models, custom model repositories, and integrates with the high‑performance vLLM backend.

A tool that enables DLSS 5 neural rendering in games that lack native DLSS support by simulating the necessary API calls using ReShade data.

LongMemory is a local‑first, SQLite‑backed memory engine for LLM‑driven agents. It stores immutable, temporally‑aware graph nodes, offers strict/historical/associative/world‑grounded recall, and includes governance, multilingual support, and many integrations (CLI, HTTP API, VS Code, n8n, etc.). Deployable via Docker, cloud platforms, or locally.

A comprehensive JavaScript reverse engineering toolkit for AI agents to analyze Web APIs, bypass anti-scraping signatures, and restore encrypted logic in Node.js or Python.

This project is the companion resource for a 66-page survey paper on "harness self-evolution" — the process by which an AI agent improves the software around its model (prompts, memory, workflows, tools) using past execution experience, without retraining the model itself. It provides a definition, taxonomy, categorized literature list, related surveys, evaluation and theory branches, and background precursors.

AI-FW is a self-hosted gateway that sits between AI agents and the models they call, inspecting every prompt and response, enforcing governance rules, and producing regulator-ready audit logs. It supports fail-closed operation, agent identity, cost-saving caches, and integration with tools like Claude Code, Cursor, and M365 Copilot.

Easy-RL (also known as the Mushroom Book) is an open-source Chinese tutorial for reinforcement learning that compiles classic lecture notes into a structured book with chapters, exercises, code notebooks, and paper readings. It aims to make RL theory and practice accessible to beginners, with free online reading, PDF download, and a printed edition.

Iris is a pure C inference pipeline for generating images from text prompts using open-weight diffusion transformer models like FLUX.2, FLUX.3, and Z-Image-Turbo. It runs with zero external dependencies, supports GPU acceleration via Metal, CUDA, and ROCm, and can operate in low-memory environments using memory-mapped weights.

A video direction skill for Dreamina/Jimeng Seedance 2.5 that transforms creative ideas and reference images into structured video scripts and optimized prompts.

MinusPod is a self‑hosted Docker service that transcribes podcasts, uses an LLM to spot ad segments, cuts them out, and serves ad‑free feeds. It supports local Whisper‑based transcription, multiple LLM providers, pattern learning from user corrections, and a web UI for review.

A read-only, agentless diagnostic tool for PostgreSQL that provides health reports, performance regression tracking, and AI-powered explanations of database metrics.

FlashVSR is a diffusion‑based video super‑resolution system that runs at about 17 fps for 768 × 1408 video on an A100 GPU. It achieves this speed through a three‑stage distillation pipeline, locality‑constrained sparse attention, and a tiny conditional decoder, and is trained on a new 120 k video dataset.

A collection of 27 AI agent skills for Codex and Claude Code to help scientists write, revise, and manage rebuttals for Nature-series life science and computational biology papers.

AI Content Studio is an open‑source platform that automates the whole video‑creation workflow. A supervisory set of AI agents researches a topic, writes and QA‑checks a script, then a multimodal pipeline (Whisper transcription, OpenCV scene analysis, FFmpeg processing) produces a fully‑styled, loudness‑normalized video with captions, B‑roll and branding. It supports style‑cloning from reference clips, Urdu TTS, and 24/7 auto‑pilot campaigns, and runs via a FastAPI backend + Next.js UI.

ZLUDA is a drop-in replacement for CUDA that enables unmodified CUDA applications to run on non-NVIDIA GPUs with near-native performance.

RuVector is a Rust-native vector database and persistent memory substrate for AI agents. It combines local semantic embeddings, vector retrieval, graph relationships, and explicit feedback learning to give agents durable, cross-session memory and fast local decisions, with SDKs for Node.js, Python, Rust, and browser WebAssembly.

A framework for self-referential, self-improving agents that can autonomously optimize their own code to solve any computable task.

A collection of Docker configurations and automation scripts for deploying high-performance vLLM inference on single or multi-node NVIDIA DGX Spark clusters.

OpenCompany is an open‑source desktop app that lets you create autonomous AI “employees” – small teams of LLM‑driven agents – that can read email, write code, post to social media, handle payments, etc. You define a job, drag specialist agents onto a visual canvas, connect them to tools (Gmail, GitHub, Stripe, browsers, Android emulator, …), and the employee runs locally (or in Docker) using any LLM you provide. State, memory, and skills are editable, and the system works without a subscription.

GPT‑Load is an open‑source Go service that acts as a single, self‑hosted gateway for many LLM APIs. It lets you configure multiple providers, manage API keys, schedule traffic, handle retries, and view unified logs and cost dashboards—all behind a familiar OpenAI‑compatible HTTP interface.

md2wechat is a Go‑based CLI that converts Markdown into WeChat public‑account HTML, offers advanced layout modules, image generation, multi‑account handling, and a JSON‑driven discovery interface for AI agents. It works in a free “prompt‑only” mode or a paid Professional API mode that returns final HTML.

ClawRouter is an open-source smart LLM router that analyzes each request across 15 dimensions and routes to the cheapest capable model in under 1ms, entirely locally. It supports free models, credit card top-ups, and USDC micropayments via the x402 protocol, letting autonomous agents use LLMs without accounts or API keys.

A Model-as-a-Service (MaaS) platform and library that provides a unified interface for discovering, inferring, and fine-tuning hundreds of state-of-the-art AI models across multiple domains.

An AI-powered tool that analyzes GitHub repositories or local codebases to automatically generate beginner-friendly tutorials and visualizations explaining how the code works.

A collection of simple PyTorch implementations of neural networks and related algorithms, each documented with clear explanations and rendered side-by-side on the companion website. It helps you understand deep learning papers by giving you minimal, readable code and notes that explain how it works.

A managed agent memory platform and ecosystem of integrations that provides AI agents with long-term memory and data ingestion pipelines.

LingBot‑World‑V2 is a research‑grade video‑generation model that creates infinite, interactive worlds from text prompts. It offers fast (real‑time) and high‑quality variants (14 B and 1.3 B), supports diverse actions, and uses a pilot‑director agent architecture. The repo provides inference code, model download links, and a quick‑start guide; deployment code is not released.

A collection of Claude Code “skills” that let the Anthropic Claude assistant review papers, grant proposals, pre‑analysis plans, and analysis code, turn papers into policy briefs or slide decks, and convert PDFs to markdown. Install as `SKILL.md` files, invoke with slash commands (e.g., `/review-paper`), and receive structured markdown reports or HTML outputs.

A browser extension that integrates a reader, media player, and AI tools to help users capture and organize learning notes from articles, videos, and podcasts directly into Obsidian.

ChatGPT2API is an open‑source self‑hosted service that turns the ChatGPT web UI into an OpenAI‑compatible REST API, with a Vue admin console for multi‑account, proxy, quota and asset management.

A Swift package that integrates Claude into Apple's Foundation Models framework, allowing developers to use Claude as a server-side model via native Apple APIs.

An AI-powered database client and SQL workspace that allows developers and analysts to generate, explain, and optimize SQL using natural language.

Mini Pupper 4 is an open-source quadruped robotics platform that enables developers to train neural network policies in high-fidelity simulations and deploy them directly to physical hardware using ROS 2 and micro-ROS.

A versatile AI-powered tool that translates books and documents into bilingual versions by leveraging various LLMs and translation services while preserving original file formatting.

docTR is a PyTorch-based library that provides an end-to-end pipeline for detecting and recognizing text within images and PDF documents.