Trending AI projects

A distributed search and analytics engine and vector database that enables near real-time search and analysis over massive datasets for AI and monitoring applications.

A C++/CUDA inference engine optimized for the NVIDIA RTX 5090, providing maximum performance for Qwen3.5 Dense and MoE models across text, image, and video.

Seedance 2.0 is an open‑source “Skill OS” that turns a short story idea into a shot‑by‑shot storyboard and precise text‑to‑video prompts, handling moderation and multilingual documentation. Users feed the generated prompts to any external video generation service, paying only for the rendered clips.

Open‑source multimodal LLMs (MiniCPM‑V 4.7 and MiniCPM‑o 4.5) that run on‑device, support vision‑language (images, video, docs) and full‑duplex audio‑visual interaction, with MoE inference optimised for phones.

iPolloWork is a local‑first, Electron‑based workbench that unifies multiple AI‑agent runtimes (OpenCode, Codex, DeepSeek Harness) into one collaborative UI for teams. It lets you manage agents, plugins, tasks and editable outputs (code, docs, designs, video) from a single workspace, with optional cloud integration for identity and hosted workers. Installers are provided for macOS, Windows and Linux; the source can be built with Node 22+, pnpm and Bun. The project is released under a source‑available license that permits personal use but requires permission for larger or commercial deployments.

A realtime voice runtime that enables AI agents to maintain continuous, full-duplex conversations while simultaneously executing background tasks and tool calls.

Decepticon is an open‑source autonomous red‑team platform that uses LLMs, a sandboxed Kali Docker environment, and specialist agents (BloodHound, Sliver, Ghidra, etc.) to run realistic attack chains. It provides a terminal CLI and optional web dashboard, stores findings in a Neo4j knowledge graph, supports multiple LLM providers with tiered fallback, and can be used as a Python SDK. Installation is Docker‑compose‑based; the project is Apache‑2.0 licensed and includes extensive documentation and a community Discord.

A local, private AI Agent that builds a deep psychological profile of the user to provide personalized, cross-platform content recommendations from sites like Bilibili, YouTube, and X.

An open-source SDK for building and running AI agents in Python and TypeScript, providing a model-agnostic framework for managing agent loops, tools, and memory.

An AI-powered exam preparation tool that turns your course materials into an interactive, grounded study tutor that tracks progress, quizzes you on your own content, and cites every source.

zero2Agent is a free, Chinese-language, MIT-licensed tutorial repo that teaches programmers how to engineer production-grade AI agents — covering core concepts, real frameworks, hand-written agent runtimes, interview preparation, and a complete final project — all delivered as 133 Jekyll articles on GitHub Pages.

pua is a multi‑platform skill that injects corporate‑style “performance‑improvement‑plan” rhetoric and a 7‑point debugging checklist into AI coding assistants (Claude Code, OpenAI Codex, Cursor, Kiro, etc.). It auto‑triggers when the model shows lazy patterns (retries, blaming the user, passive waiting) or can be invoked manually with `/pua`. The skill forces the model to exhaust solutions, verify results, and act proactively, offering language‑specific packs (Chinese, English, Japanese). Benchmarks on a handful of bugs show higher fix and verification counts, though speed gains are modest. Installation is via the respective tool’s plugin/skill system (e.g., `npx skills add tanweai/pua --skill pua-en` for Claude Code).

AdCraft is an open‑source, AI‑orchestrated platform that turns a textual brief into a fully editable advertising video. It chains specialized AI agents (creative director, scriptwriter, character designer, storyboarder, video director, sound director, etc.) to generate scripts, product renders, characters, scenes, storyboard images, video clips, music, and a final composited timeline. Users can interact via a visual canvas or natural‑language chat, edit any step, reuse assets from a unified library, and even clone existing ads to create new variants. The system supports many image/video/audio generation models and provides ready‑made character and scene catalogs.

A local, non-commercial research demo that lets you control a 3D virtual character via Chinese chat and simple motion commands using a compact text model and a small motion model. It runs on your own machine with no external APIs.

An Android screen translation tool that overlays translated text on games, manga, and videos, supporting both offline and cloud-based OCR and translation engines.

SenseVoice is a multilingual speech foundation model that provides high-accuracy speech recognition, emotion detection, and audio event identification with low-latency inference.

StarNet is a local‑first desktop app that lets you build and run multiple AI agents in a pixel‑art space‑station UI. Each room represents a bounded team with its own memory, permissions and budget. Agents use real LLM providers (OpenAI, Anthropic, Google, OpenRouter, or a local Ollama model) and can call tools, send messages to chat platforms, run scheduled recipes, and output real files. The visual layout is the live workflow, and all state (transcripts, spend logs, memory) is persisted on disk with secrets kept in the OS keychain.

Synthadoc is an open‑source, local‑first wiki engine that ingests PDFs, docs, spreadsheets, web pages, images, videos and LLM chat logs, uses an LLM to synthesize them into markdown pages with automatic wikilinks, citation provenance, contradiction detection, and a 5‑state lifecycle. It offers CLI, HTTP, Obsidian plugin and web UI, supports many LLM providers, and includes agentic ingest, lint and maintenance pipelines for autonomous self‑optimization.

OpenBot is a platform that turns low-cost smartphone hardware into the brain of an autonomous robot, enabling advanced features like real-time navigation and person following.

An integration-free development framework that allows calling remote methods across different languages and runtimes as if they were local dependencies, with built-in MCP support for AI agents.

A self-hosted AI tool that uses sonic analysis to organize music libraries and generate groove-aware playlists without relying on metadata.

Continue is an open-source coding agent available as a CLI, VS Code extension, and JetBrains plugin to assist developers with programming.

Isaac Sim is NVIDIA’s GPU‑accelerated robotics simulator built on Omniverse, supporting realistic physics, RTX sensor rendering, ROS 2 integration, synthetic data generation, and a reinforcement‑learning framework (Isaac Lab). It runs on Windows or Ubuntu with an RTX GPU and can be built from source, packaged as a binary, Python wheel, or Docker image.

A locally-hosted AI companion system integrating multi-modal chat, smart camera surveillance, and RAG-based memory to create a proactive personal assistant.

SkillHub is a self-hosted, enterprise-grade registry platform for publishing, discovering, and governing reusable AI agent skills within an organization.

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.

gitai is an open‑source, server‑centric version‑control system modeled on Anthropic’s Scion design. It moves Git history queries, merges, conflict resolution and verification to a stateless server backed by a relational DB and object storage, enabling thousands of short‑lived AI coding agents to fetch only needed objects, land changes concurrently via compare‑and‑swap, and automatically resolve conflicts with any LLM provider. The project includes a CLI (`graft`), a Go SDK, optional caching daemons, and full documentation. It is at version 0.1.0, MIT‑licensed, and ready for community contributions.

MaaFramework is a C++‑20, cross‑platform automation framework that drives applications via image‑recognition. It provides low‑code pipelines, language bindings (Python, .NET, Node, Go, Rust), and a rich ecosystem of GUIs, debuggers, and CLI tools for building game bots, UI tests, or AI‑agent environments.

better-douyin is a Rust/Tauri desktop client for Douyin (TikTok) offering browsing, downloading, multi‑account support, and AI‑assisted interactions. The public repo only contains the UI and a mock backend; the full functional binary is distributed via GitHub Releases under a non‑commercial license.

A full-stack AI creation toolkit for content creators that transforms platform trends and data into finished posts for WeChat, Xiaohongshu, and video platforms.

A Windows desktop assistant for League of Legends ARAM mode that provides real-time champion selection advice and in-game Augment recommendations via screen recognition and LCU data.

dsh‑genui is a TypeScript plug‑in for DeepSeek Harness that turns specially‑formatted JSON fences (`dsh‑ui`) emitted by the LLM into live, interactive UI components (tables, charts, plots, quizzes, media, etc.) directly inside the chat. It streams UI incrementally, handles most interactions locally, and can send user actions back to the model for a closed‑loop conversation.

Official multi‑language SDKs (Python, TypeScript/Node, Go, .NET, Java, Rust) that let applications embed the GitHub Copilot agent runtime. Handles Copilot CLI lifecycle, offers planning, tool calls, custom agents/skills, and supports GitHub auth or BYOK.

Elmo is an open-source, self-hosted platform that monitors how AI answer engines like ChatGPT, Perplexity, Gemini, and Google AI Overviews mention, cite, and describe your brand, helping you benchmark competitors and grow visibility in AI answers.

An open-source platform for building enterprise-grade AI agents that integrates RAG pipelines and agentic workflows to create intelligent knowledge bases and customer service tools.

Open‑source AI platform combining the Mojo systems language and the MAX Python framework for model authoring, GPU kernels, and OpenAI‑compatible serving.

AnyJev lets you ask an LLM a typed question (choice, yes/no, or score) and get back a decision with a probability distribution over options, without parsing. It offers a training-free mode that corrects order and label bias with zero labels, and self-distilled Tacit models that give one-pass decisions with optional adaptive reasoning.

An AI-assisted workflow and toolset for converting raster logos into clean, animated SVG logos with a focus on precision fitting and HTML-based motion delivery.

last30days‑skill‑cn is a Python‑based command‑line tool that aggregates the most discussed topics from eight major Chinese internet platforms (and optionally Hacker News, GitHub, Reddit) over the past 30 days. Designed for AI agents with CLI access, it can be installed and run via a single prompt, producing concise, source‑linked results in Markdown, JSON or offline HTML.

Inspect is a Python framework for evaluating large language models, offering built-in components for prompt engineering, tool usage, multi-turn dialog, and model-graded evaluations, along with over 200 pre-built evaluations.

TanStack AI is a TypeScript‑first, provider‑agnostic SDK that lets you build streaming chat, tool‑calling agents, structured‑output flows, multimodal generation (image, audio, video), realtime voice, and sandboxed code‑execution agents. It offers composable activities, type‑safe contracts (via Zod/Valibot/JSON Schema), adapters for many LLM providers (OpenAI, Anthropic, Gemini, Ollama, etc.), and framework‑native hooks for React, Vue, Solid, Svelte, and Preact. Tree‑shakable and observable, it’s aimed at developers who want a unified, type‑safe way to integrate generative AI into web and server apps.

Pulse is a self-hosted infrastructure monitoring platform for Proxmox, Docker, and Kubernetes that uses AI-driven analysis to identify system issues and provides an MCP adapter for AI agent integration.

whatsapp‑mcp is a Cloudflare‑Workers‑based server that lets an AI agent read, search and summarize your personal WhatsApp chats. It links to your phone via the standard WhatsApp Web QR flow, stores messages locally in SQLite, and exposes read‑only MCP tools for agents. The project is open‑source (MIT) and intended for a single user’s account only.

A Python‑only skill that lets AI coding assistants fetch US‑centric market data (quotes, K‑lines, fundamentals, options Greeks, short‑interest, SEC filings, macro rates, FX, Taiwan stocks, energy data) from 17 public sources. It provides 50+ endpoints across 16 logical layers, requires only the `requests` library, and includes a compliance tier table to guide commercial use.

Dormice is a self‑hosted sandbox platform for AI agents. It offers permanent, idempotent sandboxes that freeze to a few megabytes when idle, can be archived to S3, and expose an E2B‑compatible API. Install with a single script, manage via a tiny web console or the `dor` CLI, and use the native TypeScript SDK (or the official `e2b` SDK) to run commands, move files, and control lifecycle. Ideal for teams that want cheap, long‑lived agent sandboxes on their own hardware.

X‑AnyLabeling is a cross‑platform desktop application for AI‑assisted annotation of images, video, text, and 3D point clouds. It bundles dozens of state‑of‑the‑art vision models (YOLO, SAM, Grounding‑DINO, etc.) and supports local or remote inference, offering tools for detection, segmentation, OCR, pose, VQA, and more, with export to common dataset formats.

Cloudflare Computer is a Cloudflare‑hosted virtual filesystem backed by a Durable Object (SQLite) that can be accessed via three pluggable backends: a full Linux container (FUSE mount), a lightweight Bash shell, or a JavaScript worker. It lets AI agents and edge services run tools, store intermediate files, and share state across invocations, with stable core APIs and experimental container/worker backends.

Occamy-1.0 is an open-source, compact agentic language model (35B total, 3B active) designed for long-horizon co-work tasks like tool use, file manipulation, and multi-step workflows. It achieves leading results on co-work benchmarks while keeping inference efficient.