Trending AI projects

A tool that teaches AI agents to interact with websites via HTTP requests instead of browser automation, reducing latency and token costs.

universal-modder is an AI‑driven framework that equips LLM coding agents with skills, a CLI and a knowledge base to discover, reverse‑engineer, create assets (via Fal), test and publish mods for virtually any PC game, while enforcing safety rules.

A comprehensive prompt-engineering library that provides a structured system of styles, layouts, and color palettes to help users generate consistent, professional-quality AI images for various creative and commercial applications.

A lightweight, high-performance headless browser engine written in Rust, designed as a drop-in replacement for Puppeteer and Playwright to power AI agents and web scrapers with minimal resource usage.

A coding-agent skill that enables the creation of stunning, single-file HTML presentations from scratch or by converting PowerPoint files, using a visual style discovery process.

Docling is an open‑source Python library/CLI that parses many document types (PDF, Word, slides, spreadsheets, HTML, e‑books, audio, video, email, etc.) into a unified, structured representation. It extracts layout, tables, code, formulas, images, and speech‑to‑text, then exports to Markdown, HTML, JSON, DocLang, DocTags, etc. Designed for generative‑AI workflows, it offers plug‑and‑play integrations with LangChain, LlamaIndex, Crew AI, Haystack, and supports local execution for privacy‑sensitive data.

byoungd/up 是一份持续更新的开放书稿，由韩先凯（离谱）撰写，从英语学习扩展到 AI、创业、生活恢复等主题。它提供中英双语书稿、EPUB/PDF 下载、可复制的工作表，以及一套从 5 分钟到 90 天的实践路径。核心理念是：定义问题→保存基线→完成最小任务→记录证据→延迟复测→决定迁移。AI 是辅助工具，判断和责任仍属于人。书稿分六部主线，覆盖英语、生活、AI、实践、长期改变和自我认知。正文采用 CC BY-NC 4.0，站点代码采用 MIT。

onetake is a Claude‑compatible AI skill that generates single‑take product launch or demo videos. It rebuilds UI in HTML, uses a library of measured motion primitives, adds realistic motion blur and room‑based sound, and automatically scores continuity with an oracle. Installable via a git clone, it works with Python, Node, and optional Whisper/TTS for narration. Non‑commercial use is free; commercial use requires a paid Pro license.

OpenMAIC is an open‑source platform that uses multiple AI agents to turn any topic or document into an interactive classroom. It auto‑generates slides, quizzes, simulations, and voice‑over, stores courses in PostgreSQL, and integrates with many LLM providers and agent workbenches. Install locally with Node 22, pnpm, and Docker‑Postgres, or deploy via Docker Compose for production.

DigUp is a macOS app that uses the open‑source EmbeddingGemma 2 multimodal model (run locally via llama.cpp) to index and semantically search the contents of pictures, PDFs, documents, audio, video and optionally source code. Queries are entered via a global shortcut; results open the file at the exact moment or line, all while keeping data on‑device.

An open-source AI gateway that provides a unified, OpenAI-compatible interface for over 100 LLM providers, simplifying integration and management for production applications.

A tool registry and proxy that allows AI agents to access a curated catalog of thousands of APIs and CLIs using a single token, eliminating the need for multiple provider subscriptions.

A cross-device open-source workbench that allows users to manage, monitor, and control AI agents running on their local work devices from any mobile or web client.

ARTEX is an open‑source, LLM‑driven autonomous penetration‑testing platform. It combines a Go backend, embedded Next.js UI, and PostgreSQL to maintain a global asset graph and per‑task exploration graph. Agents (planner + workers) iteratively generate intents, run real security tools, record every step, and feed results back into the graphs. The system offers rich dashboards, asset sync from ScopeSentry, approval gating, self‑updating binaries, and multiple deployment options (Docker, one‑click script, cross‑platform releases). Licensed under AGPL‑3.0, it is intended for local learning and research only.

Explained the project as a free 21-lesson course by Microsoft Cloud Advocates, covering its structure, topics, requirements, and audience — all grounded in the README.

A collection of CLI toolsets that enable AI agents to control WPS/Microsoft Office, Zotero, Adobe Illustrator, and Photoshop via Windows COM automation.

An offline-first knowledge and education server that bundles local AI, offline Wikipedia, and educational courses into a single managed Docker environment.

A utility that stores multiple Claude Code OAuth loginscheaply, switches between them on demand or automatically at a configurable usage threshold, and shows live usage dashboards for each account's 5-hour and 7-day rate-limit windows.

XiaoZhi ESP32 is an open‑source firmware that turns ESP32‑based boards into voice‑interactive AI chatbots. It captures audio, uses ESP‑SR for wake‑word detection, streams to cloud LLMs (Qwen/DeepSeek) via WebSocket or MQTT/UDP, and returns TTS responses. The device also runs the MCP protocol for controlling peripherals and smart‑home actions, supporting many boards, networking options, multilingual UI, speaker ID, and customizable assets.

KiviDB is a Rust‑based, Redis‑compatible in‑memory database that runs on every CPU core and adds built‑in full‑text and vector (ANN) search. It offers massive throughput (21 M GET/s on 32 cores) and sub‑ms latency, making it suitable for AI workloads that need fast key‑value access plus similarity search on embeddings. Optional Lua scripting, TLS, and standard Redis persistence are available, and it can be deployed via Docker, Linux packages, or pre‑built binaries.

Indra OCR is a 1 M‑parameter, CPU‑only OCR model that converts PDFs and images to markdown, plain text, or JSON with detailed layout boxes. It processes up to 49 pages / s (≈230 ms per page) and scores 62.1 on the OmniParseBench benchmark, outperforming Tesseract and PaddleOCR in speed and many accuracy metrics. The package provides a CLI, a Python API, and a local web viewer, and supports Latin, Cyrillic, Greek, Devanagari, Bengali, and Chinese OCR, plus any script via PDF text layers. Optional `--math` flag adds LaTeX‑style math extraction. The model is ~18 MB ONNX, licensed under a modified OpenRAIL‑M (free for research/startups), with code under Apache 2.0.

Omnira is an open‑source, self‑hostable platform for running AI agents on any LLM model and any machine. It provides a REST API, TypeScript SDK, and CLI to launch agents, stream events, and handle approvals. Features include model‑agnostic support, built‑in and custom tools, reusable skills, persistent memory, sub‑agents, multi‑tenant RBAC, scheduling, and integrations with Slack, Discord, GitHub, webhooks, and more. You can try it via Omnira Cloud or run it locally with Docker Compose; the Apache 2.0‑licensed code keeps all agent data in your own Postgres.

Sub2API is an AI API gateway platform built with Go and Vue that enables developers to manage and distribute subscription quotas across multiple AI model providers.

A continuously updated, MIT-licensed reference for production AI system design and AI engineering interviews — covering RAG, agents, model selection, evaluation, security, and 20 real-world case studies, plus a 147-question interview bank.

A vectorless, reasoning-based RAG engine that replaces vector databases and chunking with a hierarchical tree index to improve retrieval accuracy for long, complex professional documents.

Oil UI is an AI agent skill that generates and iterates on distinct UI design concepts for websites and apps, allowing users to compare and refine layouts based on specific brand needs.

Odysseus is a self‑hosted web app that combines chat, autonomous agents, research, document editing, email, notes, tasks and calendar into one AI‑powered workspace. It runs via Docker Compose, supports local or API language models, and is released under AGPL‑3.0.

A standardized, open format for packaging specialized knowledge and workflows into portable folders that AI agents can load on demand to gain domain expertise.

A curated directory of hosted Model Context Protocol (MCP) servers that allow you to connect AI agents to external tools via a URL without needing local installation or runtime management.

A high-fidelity document parsing tool that converts PDFs, images, and Office files into structured Markdown or LaTeX, featuring multi-tiered quality levels and stable locators for AI agents.

GVS5H introduces ledger-based zero-shot self-orchestration, a training-free technique where multiple instances of one language model decompose and solve coding problems via a shared filesystem. On LiveCodeBench hard tasks it lets open-weight models like Qwen3.8-27B jump from 66.8% to 92.4% accuracy, approaching or beating larger proprietary models at far lower cost.

dsh‑web is a modular plug‑in ecosystem that extends DeepSeek Harness’s web UI (and desktop client) with task automation, mobile remote control, web‑SSH, Git worktree isolation, token usage dashboards, visual skins and more, all packaged as npm modules for easy installation.

A plugin for AI agents that converts videos from URLs or local files into timestamped frames and transcripts for visual analysis.

Whisper is OpenAI’s open‑source, transformer‑based speech‑to‑text system that handles transcription, language identification, and translation across many languages. It ships as a pip‑installable package with six model sizes (tiny → large, plus a fast “turbo” variant). The CLI (`whisper …`) and a simple Python API (`whisper.load_model(...).transcribe(...)`) let you turn audio files into text with just a few commands. The project runs on Python 3.8‑3.11, PyTorch, and requires ffmpeg (and optionally Rust for the tokenizer). All code and model weights are MIT‑licensed.

Franco is an outbound sales agent that runs on your own machine and uses your existing Claude Code or Codex subscription instead of a model API bill. It finds or imports accounts, writes or sends approved email sequences, reviews replies, and records every fact with its source.

A Claude Code skill that rewrites English into Simplified Technical English (ASD‑STE100) to make agent‑to‑agent messages unambiguous. It flags structural STE violations, offers strict and relaxed modes, and can show diffs or rule explanations. Install via `npx skills` or by cloning, then ask it to simplify tool descriptions, error messages, or prompts. MIT‑licensed.

Answer me with HTML is a Node.js‑based skill that lets LLM agents turn a short Markdown draft into a full, offline‑viewable HTML page (or explainer video). By handling SVG, CSS, layout, and diagram generation itself, the model only writes ~1/8 of the tokens a hand‑written page would need, cutting cost and latency. It supports multiple agents (Claude Code, Codex, Cursor, DeepSeek Harness), offers three built‑in themes, a writing‑style checker, real‑code inclusion, and an “always‑on” mode that auto‑creates pages for every agent conclusion.

An open-source, general-purpose AI agent available as a desktop app, CLI, and API that handles coding, research, and automation tasks.

An open-source, pure-Rust reimplementation of Adobe After Effects for motion graphics and visual effects, featuring native Lottie support and an MCP server for AI agent control.

RAGFlow is an open‑source Retrieval‑Augmented Generation platform that adds agentic reasoning to LLMs. It ingests many document types, creates searchable embeddings, and supports multi‑step retrieval with configurable “thinking” depth. Features include template‑based chunking, knowledge compilation (wikis, graphs, mind‑maps), grounded citations, and plug‑in LLM/embedding models. Deploy locally via Docker or from source; a hosted cloud version is also offered. Licensed under Apache‑2.0 and backed by an active community.

PaddleOCR is an open‑source OCR and document‑AI toolkit that converts images or PDFs into structured, LLM‑ready text, tables and markdown/JSON. It offers multilingual scene‑text models (PP‑OCRv6) and lightweight vision‑language models (PaddleOCR‑VL, PP‑StructureV3) with state‑of‑the‑art accuracy, fast CPU/GPU inference, and deployment options ranging from Docker to browser SDKs. Integrated with RAG platforms like Dify and RAGFlow, it serves as a data‑engine for building AI agents and retrieval‑augmented generation pipelines.

AIHOT is an open‑source, Docker‑based framework that scrapes various feeds, uses LLMs to filter, score, translate and summarise items, clusters them into events, and ranks them by source diversity to produce daily/weekly/monthly news digests. All prompts, thresholds and selection logic are in the repo, making it easy to replace the demo AI‑news sources with any industry’s feeds and re‑brand the site.

An open-source AI-powered game studio that integrates design, asset generation, coding, and playtesting into a single canvas-based application.

A library of 33 interactive, isometric line-based UI figures that respond to pointer movement, designed for lightweight integration into web applications with support for AI-assisted generation.

Beacon is an open-source memory layer for AI coding agents that captures session history across multiple tools, allowing developers to reuse debugging patterns and workflows across different coding environments.

OmniVoice is a diffusion‑based, zero‑shot TTS model covering 600+ languages. It offers voice cloning, attribute‑driven voice design, and auto‑voice generation, with fast inference (RTF ≈ 0.025). Install via pip, load the model with `OmniVoice.from_pretrained`, and call `model.generate` (or use the provided CLI tools). Optional FlashInfer kernels give 2‑3× speedup. The repo includes training scripts, a Gradio demo, and a Colab notebook.

A comprehensive library of 398 modular skills and plugins that extend AI coding agents with domain-specific expertise in engineering, business, and compliance across 13 different platforms.

A free and open-source AI image upscaler that uses Real-ESRGAN and Vulkan to enlarge and enhance low-resolution images without losing quality.