Trending AI projects

A reproducible training and evaluation pipeline for long-horizon shopping agents using SFT and GRPO within the ShopSimulator environment.

Loghub is a free collection of real system log datasets for AI-driven log analytics research. It includes 20 datasets across six categories (distributed systems, supercomputers, OSes, mobile, server apps, standalone software), with metadata like line counts, sizes, and labeling status. Logs are intentionally unsanitized. Usage requires citing the Loghub paper and referencing the repo URL.

A Linux voice-input plugin for the Fcitx5 input framework, supporting offline and cloud speech recognition, LLM-based post-processing, and voice-controlled text editing.

MiniOneRec is an open‑source Python framework for building generative recommender systems. It converts items into compact semantic IDs (SIDs) via RQ‑VAE/Kmeans quantisation, fine‑tunes a large language model on user histories (SFT), and then refines it with GRPO‑style reinforcement learning. The repo provides full data pipelines, multi‑GPU training scripts, evaluation utilities, and supports various LLM back‑ends.

A high-performance neural network inference framework designed for efficient deployment on mobile, embedded, and desktop devices with no third-party runtime dependencies.

MagenticLite is an agentic application from Microsoft AI Frontiers that automates browser and local file system tasks using efficient, small-scale AI models.

AI Work Assistant is a Windows desktop tool that manages multiple accounts for Trae, WorkBuddy, CodeBuddy, Doubao, and Qoder. It automates daily check-ins, switches login states, and provides an OpenAI/Anthropic/Codex-compatible API gateway that routes requests across account pools.

JevBench is a benchmarking framework for AI decision models that evaluates performance across intelligence, calibration, speed, and cost to provide a standardized ranking of system capabilities.

Comma is an open-source, persistent agent that automates multi-step tasks across your devices and applications by breaking goals into autonomous, fault-tolerant execution loops.

A project that repackages the official Claude Desktop for various Linux distributions, adding a custom launcher for environment compatibility and a diagnostic tool for troubleshooting.

A low-level orchestration framework for building controllable, stateful AI agents with support for durable execution, long-term memory, and human-in-the-loop oversight.

Mortiflix is a local motion design studio that uses Claude to produce videos step by step, with you approving each stage. It replaces one-shot AI video tools with a structured, reviewable workflow where every session stops at a gate for your feedback, and it learns your taste so future videos get better.

An open-source, low-cost XYZ motion platform that provides submicron precision for microscopy and alignment tasks using stepper motors and closed-loop control.

The ChatGPT Orange Book is an open-source guide that helps developers and power users integrate ChatGPT and Codex into their software development workflows and real-world projects.

A curated directory of open-source tools, frameworks, and educational resources designed to help developers and security professionals secure AI systems against emerging threats.

A curated directory of official, trusted MCP servers that lets AI assistants connect to external tools and data reliably, avoiding unofficial or abandoned projects.

AIPerf is a comprehensive benchmarking tool for generative AI models that measures performance metrics like latency and throughput across various inference servers and traffic scenarios.

An open-source, cloud-native distributed SQL database that provides horizontal scalability, high availability, and MySQL compatibility.

Alook is a collaboration platform that gives local AI coding agents a persistent identity and shared workspace, allowing teams to interact with them from any device while keeping the agents running securely on their own machines.

A local observability tool for AI coding agents that visualizes project progress and activity through a structural map or a 3D kitchen-themed interface.

An autonomous orchestrator that keeps coding agents running in a loop, making incremental git-committed changes toward a goal while the user is away.

A feature-rich local e-book reader that uses AI to provide intelligent text formatting, plot analysis, character visualization, and multi-voice text-to-speech.

An ultra-fast, multi-threaded in-memory data store that is fully compatible with Redis and Memcached APIs but scales vertically to deliver significantly higher throughput and lower resource usage.

A high-performance, Rust-based framework for real-time robotics and AI applications that uses zero-copy shared memory and Apache Arrow to minimize latency.

A GPU-accelerated data curation framework for building repeatable pipelines to load, filter, deduplicate, and transform massive multimodal datasets for AI training.

Symposium is a Cargo subcommand that scans your Rust workspace's dependency graph and installs the appropriate skills, hooks, and MCP servers into AI coding agents, giving them project-specific guidance. It supports multiple agents with a single configuration and lets crate authors ship extensions for their libraries.

EasyVoice 是一个开源的文本转语音工具，能将长篇小说或大段文字一键生成带字幕的音频，支持多角色配音、流式播放和 AI 推荐，完全免费且无字数限制。

An open-source, JavaScript-driven desktop robot platform based on M5Stack that allows users to create and customize robot behaviors via browser-based tools and MODs.

An open-weights model and inference engine for real-time music generation, optimized for streaming audio on Apple Silicon.

A framework for structured classification and scoring using open-source LLMs by reading next-token logits instead of generating text, reducing latency and ensuring schema adherence.

Firefox-Reverse is a specialized browser that integrates an AI agent and isolated fingerprint management to automate the reverse-engineering of complex web security parameters like signatures and tokens.

An inference engine for Hugging Face-compatible language and multimodal models that provides continuous batching, paged KV-cache management, and distributed serving.

fx0 is an open‑source Vercel‑hosted app that turns a textual page description into a single‑file HTML design using the libfx LLM‑agent framework. It features durable sessions via Vercel Queues, optional free‑trial prompts, a set of design tools (web fetch, search, edit, sub‑agents), and versioned design outputs displayed in a sandboxed preview.

Goldie is a CLI tool that automates the creation of App Store and Google Play screenshots and preview videos by replaying UI flows on simulators/emulators, framing captures with device bezels, applying themes, and outputting store‑ready assets. It integrates with LLM‑powered coding agents via a ready‑made skill, letting agents generate and tweak the assets through natural‑language commands.

A CUDA-accelerated library for the rasterization of 3D Gaussians, providing a faster and more memory-efficient way to render radiance fields compared to the official implementation.

A webcam-based AI motion capture application for animating 3D avatars (VRM/MMD) in real-time for VTubing and XR purposes.

A security auditor for AI agent configurations that scans Claude Code setups for secrets, permission misconfigurations, and prompt injection vectors.

A CLI that runs AI coding agents in an OS sandbox (bubblewrap/Landlock/seccomp on Linux, sandbox-exec on macOS), scoped to prevent accidents — not to stop attackers. The project directory is read-write; the host is hidden by default, with opt-in flags for credentials, network, GPU, display, etc. Includes a strict `--lockdown` mode and a "phantom credential" proxy that keeps real API keys out of the sandbox. A mature, well-documented Rust project — not a list, tutorial, or demo.

An MCP server that gives AI assistants real-time access to 85+ financial and crypto data sources with AI-analyzed impact scores and trading signals.

An automated translation pipeline for Galgames that uses LLMs and prompt engineering to create high-quality embedded translation patches.

The repository is a genuine open‑source platform for AI‑driven quantitative trading, offering data pipelines, alpha mining, diverse strategy examples (including RL, DL, GNN, and LLM‑based models), back‑testing tools, live‑trading simulations, and low‑code assistant utilities such as an Excel stock monitor.

A personal AI assistant with a sophisticated eight-type memory system that learns user behavior and operates across multiple apps and devices.

A self-hosted, private knowledge base and task management platform that features AI RAG capabilities, MCP support for AI agents, and multi-device synchronization.

Open‑XiaoAI Bridge is a server‑side bridge that lets a Xiaomi XiaoAI speaker talk to external AI services (OpenAI‑compatible APIs, OpenClaw agents, QwenPaw, XiaoZhi). It provides wake‑word detection, VAD, optional offline ASR, multi‑agent routing, continuous dialogue, and a Docker‑ready HTTP API for remote playback. Configuration is done via `config.py` and environment variables; TTS can use the speaker’s native engine, Doubao cloud, OpenAI TTS, or a local MLX‑Audio model.

A curated library of 169 JSON‑based visual‑style prompts for text‑to‑image models. Users copy a `style.json`, fill in a few variables, paste it into a multimodal LLM (e.g., ChatGPT‑4‑Vision, Gemini‑3‑Pro Image) and get a consistent, stylized image. The repo includes previews, copy‑ready prompts, and a validator script.

PyTorch Tutorials is a repository of learning materials published as web docs. It accepts feature requests and bug reports, uses Python files converted to HTML with auto-generated notebooks, organizes content by difficulty level, and offers GPU-based full builds or quick no-plot previews.

gpt-image2-ppt-skills is a Python skill for multimodal AI agents that uses OpenAI’s gpt‑image‑2 model to generate high‑quality 16:9 slide images and package them into a .pptx file. It supports a visual‑first mode and an optional editable mode that reconstructs slides into native PowerPoint objects. Users can create new decks, clone existing .pptx templates, or edit specific slide elements via natural‑language prompts. Installation is done by asking a supported agent to install the skill or by running a provided script; an OpenAI API key and a local PowerPoint/Keynote/LibreOffice renderer are required. The project is Apache‑2.0 licensed.

A set of markdown "output styles" for Claude Code and other AI agents that force concise, answer-first, scannable responses. A benchmark shows it cuts output length ~43% while keeping coding task pass rates identical (97%).