Project Library

ImageHash is a Python library designed for image fingerprinting, enabling the comparison of images to determine near-identical visual content. It supports multiple hashing algorithms, including Average, Perceptual, Difference, Wavelet, HSV Color, and Crop-Resistant hashing. Unlike cryptographic hashes, these algorithms produce similar outputs for visually similar images, making them ideal for tasks like duplicate detection or image similarity analysis. The library is built on PIL/Pillow, numpy, and scipy.fftpack, with easy installation via pip. Users can adjust hash sizes to enhance sensitivity to image details. A demo script, **find_similar_images**, demonstrates practical applications, such as identifying similar images in a directory. ImageHash is open-source, hosted on GitHub, and includes references to detailed algorithm explanations.

AB Download Manager is a free, open-source desktop application designed to enhance file download management and organization. It offers faster download speeds, queue and scheduler functionalities, and browser extensions for seamless integration. The app supports multiple platforms (Windows and Linux) and features a modern UI with customizable dark and light themes. Installation is straightforward, with options for Linux scripts, Winget, or Scoop for Windows. Users can report bugs, suggest features, and seek help through the Telegram community or GitHub Issues. The project is actively developed, with ongoing improvements and new features planned. Contributions and translations are welcome, fostering a collaborative environment for creating a better download manager.

Minisign is a straightforward and efficient tool designed for signing files and verifying signatures. It supports multiple build options, including compilation with Zig or CMake, and can be dynamically or statically linked with libsodium. Pre-built packages are available via Homebrew, Scoop, Chocolatey, and Docker, ensuring easy installation across platforms. Minisign emphasizes simplicity and interoperability, offering deterministic signatures by default, though it supports non-deterministic variants. It also has a wide range of implementations in various programming languages (e.g., Zig, Go, Rust, Python, PHP) and libraries for integration into diverse applications. This tool is ideal for developers seeking a lightweight, cross-platform solution for file integrity verification and signing.

The Ping Clock is a real-time internet connection monitoring tool designed for video calls and network diagnostics. It visually displays ping times using dual clock hands: the outer hand shows the latest ping value, while the inner hand tracks recent maximum pings, highlighting intermittent connectivity. Users can configure the device to ping a chosen host, such as a public server or a specific cloud instance, via its Wi-Fi access point. Powered by ESP32 microcontrollers, the clock features a custom PCB, e-paper display, and direct-drive BLDC motors for precise hand movement. It is energy-efficient, consuming approximately 2 W, and includes 3D-printed hubs for mounting. Ideal for troubleshooting network issues, the Ping Clock provides immediate feedback on connection health without requiring router resets.

TensorRT-YOLO is a highly efficient and user-friendly inference deployment tool for YOLO series models, optimized for NVIDIA devices. It integrates TensorRT plugins, CUDA kernels, and CUDA graphs to accelerate inference, supporting both C++ and Python for seamless deployment. The tool is compatible with a wide range of YOLO models (from YOLOv3 to YOLOv11) and supports multiple tasks including object detection, instance segmentation, image classification, pose estimation, and rotated object detection. Key features include CUDA acceleration, TensorRT integration, multi-context inference, and optimized memory management. It offers out-of-the-box functionality, CLI tools, Docker support, and compatibility across various platforms (Windows, Linux, ARM, x86). TensorRT-YOLO is designed for high performance, flexibility, and ease of use, making it ideal for diverse deployment scenarios.

Mopidy is an extensible music server written in Python, capable of playing music from local files, radio streams, and cloud services like Spotify, SoundCloud, and Google Play Music through extensions. It operates as a server on Linux or Mac systems, supporting HTTP and MPD protocols, and can be controlled via various MPD or web clients from any device. Mopidy is particularly popular on Raspberry Pi, often integrated with audio hardware like Pirate Audio or Pi Musicbox. Its hackable nature, supported by Python, JSON-RPC, and JavaScript APIs, allows for custom projects, such as embedding it in retro hardware. Mopidy is ideal for creating a shared, multi-client music experience with extensibility for new music sources and custom integrations.

Starlight is a modern, lightweight framework designed for building fast and scalable web applications. It offers a modular architecture, enabling developers to easily integrate and manage components. Key features include a robust API for seamless data handling, real-time updates through WebSocket support, and a built-in state management system for efficient application state control. Starlight emphasizes performance optimization with lazy loading and tree-shaking capabilities, ensuring minimal bundle sizes. It also provides a comprehensive set of developer tools, including hot module replacement and debugging utilities, to streamline the development process. With its focus on simplicity and flexibility, Starlight is ideal for both small projects and large-scale applications, offering a smooth development experience and high-performance outcomes.

Zeal is an offline documentation browser inspired by Dash, designed for developers to access technical documentation without an internet connection. It supports Windows and Linux, offering a user-friendly interface to download and manage docsets for various programming languages and frameworks. Users can search and filter docsets using specific syntax, and queries can also be initiated via the command line. Zeal allows customization by enabling users to create their own docsets following Dash’s guidelines. Built with CMake, Qt, libarchive, and SQLite, it is open-source under the GPLv3 license. Support and feedback are available through GitHub, IRC, Twitter, and email.

The iCloud Photos Downloader is a versatile command-line tool designed to download and manage iCloud photos across Linux, Windows, and macOS platforms. It supports multiple installation methods, including executables, Docker, PyPI, AUR, and npm. Key features include three operational modes (Copy, Sync, Move), support for Live Photos and RAW images, automatic de-duplication, and continuous monitoring for iCloud changes. The tool also offers optimizations for incremental runs, EXIF metadata updates, and experimental features for advanced users. It is maintained by volunteers and aims for weekly updates. Ideal for syncing iCloud photo collections locally, it provides flexible command-line options for customization and session management.

rot.js is a versatile RogueLike Toolkit in JavaScript, designed to simplify the development of roguelike games. It offers multiple integration methods, including direct download, npm installation, and prebuilt bundles, catering to various development environments. The library is written in TypeScript and supports modern ES2015 modules, bundling with tools like Babel and Rollup, and traditional ES5 scripts for older browsers. It is also compatible with Node.js for server-side usage, with specific support for terminal-based displays. rot.js is considered feature-complete but welcomes contributions for bug fixes, documentation, and new features. Its build process is streamlined with npm and make commands, ensuring easy setup and deployment. Ideal for both modern and legacy projects, rot.js is a robust toolkit for creating roguelike experiences.

Kanata is a cross-platform keyboard remapper for Linux, macOS, and Windows, designed to enhance keyboard comfort and functionality. It allows users to create multiple layers of key functionality, enabling advanced customization such as tap-hold actions, macros, and Unicode output. Kanata supports live configuration reloading, Vim-like leader sequences, and optional TCP server integration for program interaction. It features a human-readable configuration file and offers pre-built binaries or self-build options using Rust. Kanata aims to bring QMK-like customization to any keyboard, improving accessibility and user experience without requiring specialized hardware.

InStock is a comprehensive stock analysis and trading system designed for quantitative investment. It captures daily stock and ETF data, calculates technical indicators, identifies K-line patterns, and supports multi-criteria stock selection with over 200 customizable filters. The system includes built-in strategies, backtesting, and automatic trading capabilities. It efficiently runs on PCs, tablets, and mobile devices, offering a Docker image for easy installation. Key features include chip distribution analysis, strategy-based stock selection, and real-time data processing. InStock is optimized for performance, with tasks like data capture, indicator calculation, and strategy backtesting completed in approximately 4 minutes. It also provides a web-based interface for visualization and supports database storage for historical data analysis.

Amarok-Hider is a lightweight Android application designed to quickly and efficiently hide private files and apps with a single click. Unlike traditional encryption tools, Amarok disguises file names and headers to make them appear corrupted and unopenable, while also deactivating apps to render them invisible in the launcher. Key features include a panic button for urgent hiding, a quick settings tile for instant access, app lock with password or fingerprint, and a user-friendly Material3 interface. It supports root-free app hiding and is compatible with various modes like Shizuku and Dhizuku. Note: Amarok is not an encryption tool and is not recommended for protecting highly confidential data.