Trending AI projects

An open-source machine learning framework for building contextual AI assistants and agents that automate text and voice-based conversations.

An AI-powered platform for long-form fiction writing that ensures plot continuity and structural integrity through a structured pipeline from outline to prose.

A tool for automatic 2D motion capture and biomechanical analysis that computes joint trajectories and angles from a single video or webcam.

EventCatalog is a source‑available web tool for documenting event‑driven software architectures. It lets you model domains, services, messages and schemas, offers AI‑powered natural‑language queries, searchable schema exploration, versioning, and visual flow diagrams. Install with a one‑line npx command and run the Astro/React UI locally or host it yourself.

A real-time stream editing pipeline for FLUX.2 that transforms webcam or video feeds with low latency and interactive prompt updates on consumer GPUs.

RepoPrompt CE is an open‑source macOS app that helps you collect and review the exact parts of a codebase (files, diffs, CodeMaps) that an LLM‑based coding agent needs, then hands that curated context to the agent via a built‑in MCP server. It also lets you launch and coordinate CLI coding agents from the UI, supporting multi‑repo workspaces and token‑budget‑aware prompt building.

MolmoSpaces is a large-scale ecosystem for robot manipulation and navigation that provides unified tools for data generation, benchmarking, and simulation across platforms like MuJoCo and Isaac.

An extension for the pi CLI that manages multiple AI service subscriptions, providing automatic rate-limit rotation and intelligent failover across different accounts and providers.

An open-source observability platform that provides zero-code, full-stack monitoring and distributed tracing for cloud-native and AI applications using eBPF.

The NVIDIA k8s‑device‑plugin is an official DaemonSet that implements Kubernetes’ device‑plugin API for NVIDIA GPUs. It discovers GPUs, reports health, and makes them schedulable via the `nvidia.com/gpu` resource (with optional MIG and shared‑GPU modes). Install the NVIDIA driver and container toolkit on each node, deploy the plugin (static manifest or Helm), then request GPUs in pod specs. Configuration is possible via flags, env‑vars, or a YAML file, covering MIG strategy, device‑list delivery, shared‑GPU (time‑slicing or MPS), and more. The plugin is beta, lacks full health‑check/cleanup, and only the official NVIDIA version is supported.

A full-stack development platform for AIoT and HMI products that integrates hardware abstraction, multi-runtime app support, and AI agent capabilities for ESP devices.

Open Agent Email is a self‑hosted service that gives each AI agent a real email address and mailbox. It handles inbound SMTP, stores messages, and lets agents read, reply, forward, schedule, or request human approval before sending. Accessible via REST, Python/TypeScript SDKs, CLI, WebSocket, webhooks, and MCP. Features include per‑agent inboxes, thread view, SPF/DKIM/DMARC validation, prompt‑injection screening, approval gate, any SMTP relay, delivery tracking, signed webhooks, and a Next.js dashboard. Built with Python/FastAPI, SQLite/PostgreSQL, and Docker; Apache‑2.0 licensed.

A modular, production-grade SDK for connecting AI agents to WeChat via the iLink protocol, supporting Node.js, Python, Go, and Rust.

Judge0 is an open-source online code execution system that compiles and runs untrusted code in a sandbox, supporting 90+ languages via a simple HTTP JSON API. It is designed for building applications like AI agents, e-learning platforms, online judges, and IDEs that need secure, scalable code execution.

An enterprise-grade AIOps platform that uses multi-agent AI collaboration to automate the full cycle of alert diagnosis, human-approved repair, and verification.

A privacy-preserving human pose estimation system that uses WiFi Channel State Information (CSI) and machine learning to detect poses without cameras.

A Chinese-language version of the CLIP model trained on 200 million image-text pairs for cross-modal retrieval and zero-shot image classification.

A local AI video production workbench that converts text scripts and reference audio into whiteboard animations and dynamic infographics.

JoyAI-Echo is a research-focused toolkit for generating long-form, coherent audio-visual content and interactive world models that maintain narrative and environmental continuity.

A self-published, continuously updated guide to deep learning with a private learner's narrative — moving from basic neural networks, PyTorch, CNNs, and RNNs, through Transformers and FlashAttention, into building GPT-2 from scratch, LLM training engineering, and modern inference frameworks (vLLM, SGLang).

BFM-Zero is a promptable foundation model for humanoid robot control that uses unsupervised reinforcement learning to enable versatile behaviors like motion tracking and goal-reaching from a single trained policy.

EmbodiedJev is a browser-based workbench for testing and visualizing how AI models control robotic arms in simulated environments, supporting both local and cloud-based models.

Refero is an AI agent skill and MCP connection that forces design-focused LLMs to research real-world app screens and user flows before generating UI/UX code.

A collection of distilled LoRA checkpoints for MiniMax-H3 that enables fast, few-step video and audio generation via Diffusers and ComfyUI.

A Codex skill that transforms travel photos into collectible digital memory sticker cards.

LeoAI is an open‑source AI‑augmented red‑team platform. It provides a web UI for host inventory, interactive shells, file/database/network tools, and a LangChain4j‑based LLM assistant that can plan and execute multi‑step post‑exploitation tasks. The backend is Spring Boot 3.5 (Java 17+), the data store is embedded SQLite, and agents run either as Java components or a single‑file PHP 5.6+ HTTP service. Install via a single JAR or Docker, configure OpenAI‑compatible models, and use the built‑in Skills or custom plugins for automated recon, credential dumping, persistence, and reporting.

A Python‑based Agent Skill that lets LLM agents interact with Kaggle: fetch competition info, summarize top write‑ups, search discussions, download and reproduce notebooks, submit solutions, and upload datasets—all via natural‑language commands.

Hippo‑Memory is a Node‑JS CLI/library that gives coding agents a shared, local SQLite‑backed memory store with decay‑based reinforcement, markdown mirrors, BM25 search, optional embeddings, and hooks for Claude Code, Codex, Cursor, OpenCode, Pi and more.

Dapr Agents is a Python framework that makes it easy to build reliable, scalable AI agents using LLMs. It leverages Dapr's built-in messaging, state, and durable workflow capabilities to guarantee your agent tasks complete — even across failures — while keeping costs low through scale-to-zero virtual actors. It auto-discovers tools from MCP servers, connects to 50+ data sources, and runs natively on Kubernetes. Currently in development for Python, with stable release expected Q4 2025.

An AI-powered penetration testing tool that uses a multi-agent architecture to simulate human security experts in discovering vulnerabilities in web apps and APIs.

NextPlaid is a local-first multi-vector search database with a REST API; ColGREP is a semantic code-search tool built on it — a single Rust binary that indexes your codebase locally and answers natural-language queries (optionally via a small embedded agent model), combining regex filtering with semantic ranking. NextPlaid solves the storage cost of multi-vector embeddings (hundreds per document) with quantization, memory-mapped indices, incremental updates, and SQL metadata pre-filtering, all CPU-friendly. Both tools are Apache-2.0 licensed, use ONNX-exported ColBERT models, and keep all data on your machine.

AI Maestro is a self-hosted peer-mesh platform that orchestrates AI agents across multiple machines, giving them persistent memory, agent-to-agent messaging, shared task boards, and secrets management. It solves the problem of running many agents in separate terminals with no coordination, and works with existing agents like Claude Code, Codex, and Grok Build.

A spatial development workspace that organizes AI coding agents, terminals, and notes on an infinite 2D canvas to maintain visible context.

A curated, open collection of Agent Skills for scientific research, covering bioinformatics, chemistry, data analysis, literature search, and more. It gathers high-quality, science-relevant skills from official and community sources into a clone-ready repository that works across multiple AI coding agents.

This repository contains the source code, build scripts, and technical examples for the book "Learning Deep Representations of Data Distributions," enabling users to compile the text or reproduce its figures.

A curated directory of frameworks, tools, and resources for developers building AI agents that can browse, scrape, and interact with the web.

A curated repository of research papers and resources on robot manipulation, organized into a systematic taxonomy that bridges high-level planning and low-level learning-based action models.

A Chinese-language, community-maintained guide collecting ChatGPT and AI resources — from prompts and apps to model ecosystems, development guides, and frontier topics — with a Telegram community and open contribution model.

BeatAI is a technical publication providing deep-dive guides and architectural insights for developers building and scaling agentic AI systems.

A serving stack that runs GLM-5.3-Flash (125 GB) on a single RTX 3090 24 GB by splitting its 12,096 routed experts across VRAM, RAM, and NVMe, delivering up to ~28 tok/s decode with near-exact output.

SimUtil is a unified toolkit for managing Android and iOS devices that enables developers to stream, control, and automate testing across multiple emulators and physical devices, including support for AI agents via MCP.

A configuration template for Claude Code that provides specialized agents and knowledge skills to help developers build production-ready, cloud-native Spring Boot 4.x applications.

...

A tool that converts human motion capture data from BVH files into joint animations for humanoid robots by adapting proportions and solving inverse kinematics.

A long-running AI agent loop that automates software development by iteratively executing a task list within Docker Sandboxes until completion.

RL-Kernel is a high-performance infrastructure library that ensures numerical consistency between rollout and training engines in RL post-training, preventing log probability discrepancies across different hardware backends.

Distill is a terminal-based coding agent that reduces token costs and development time by orchestrating a hierarchy of models to handle planning, execution, and utility tasks efficiently.

OmniChar Studio is an open‑source Python application that lets you build a portable `.char` file describing a character’s face, body, clothing and voice. The same character can be used with many generative models (FLUX, Krea, MiniMax H3, LTX‑2.5, etc.) for image or video generation, with optional LoRA training. It provides a web UI, a CLI installer that auto‑selects the correct CUDA/PyTorch wheel, low‑VRAM and CPU profiles, and a ComfyUI node for workflow integration. Licensed under GPL‑v3 and supported on NVIDIA, Apple Silicon, AMD ROCm, and CPU‑only setups.