KerasHub is a production-grade Keras 3 library providing pretrained model architectures, layers, tokenizers, metrics, and sampling utilities for NLP, vision, multimodal, and audio tasks. The codebase is organized around a consistent Backbone → Task → Preprocessor pattern, where each model family (e.g., BERT, Gemma, ViT, Whisper) defines a backbone, a preprocessor (tokenizer + optional media converter), and task-specific heads (CausalLM, ImageClassifier, etc.). All models expose a unified `from_preset()` API for loading pretrained weights hosted on Kaggle Models, and `save_to_preset()` for export. The library includes extensive checkpoint conversion tooling from HuggingFace and TIMM, export to HuggingFace format and OpenVINO, and integration with vLLM for serving. While predominantly an application library, it has strong data/research character through its pretrained weight infrastructure, benchmark scripts, and research-oriented architectures (e.g., FNet, RWKV7, BLIP-2).
Nostalgist.js is a TypeScript library that provides a simplified API for running retro console emulators in web browsers using RetroArch compiled via Emscripten. It handles the complex setup of Emscripten modules, virtual file systems, and RetroArch configuration, allowing developers to launch games with minimal code. The library supports a wide range of consoles (NES, SNES, Game Boy, etc.) with sensible defaults for loading cores and ROMs from CDNs, while also allowing custom resolvers. It includes features like save/load states, SRAM management, screenshot capture, and programmatic input control. The codebase is well-structured with strong TypeScript types, comprehensive tests (integration and e2e), and a clean class hierarchy. Notable highlights include a flexible file resolution system that handles various input formats (URLs, Blobs, ArrayBuffers, etc.) and a clever patching mechanism for importing Emscripten modules as ES modules.
SyntaxUI is a comprehensive documentation and showcase application built with Next.js 15, designed to display a library of pre-built React UI components. The application serves as both a marketing site and a functional reference for developers, allowing them to browse components, view live previews, and copy source code directly. It leverages the Next.js App Router for routing and server-side rendering, with MDX files used extensively for documentation content. The architecture is monolithic, centered around a static site generation approach with client-side interactivity for features like search and component previews. The core functionality revolves around displaying UI components categorized into buttons, loaders, text effects, and layout blocks. It integrates Tailwind CSS for styling, Framer Motion for animations, and Lucide React for iconography. The site includes a custom client-side search implementation using FlexSearch, which indexes the MDX documentation at build time. It also features dynamic elements like a GitHub stars counter and a Pro banner, managed via Zustand stores. The codebase is structured to separate the showcase components (in `src/showcase`) from the documentation pages (in `src/app/(docs)`), facilitating easy addition of new components. From a code quality perspective, the repository demonstrates strong engineering practices. It uses TypeScript throughout, ensuring type safety for components and utilities. The project utilizes modern Next.js features, including server actions for file reading and custom webpack loaders for search indexing. The separation of concerns is clear, with reusable UI primitives (like `ComponentPreview`) abstracted away from the specific component showcases. The documentation is generated from MDX, allowing for rich content embedding alongside code examples. The repository is actively maintained, evidenced by recent commits and a low number of open issues. It keeps pace with the ecosystem, utilizing the latest stable versions of Next.js and Tailwind CSS. There is minimal technical debt, with no deprecated APIs or outdated dependencies visible in the manifest. The code is clean, well-organized, and follows a consistent pattern, making it a valuable resource for engineers looking to build documentation sites or component libraries.
NoPoSplat is a research application for novel view synthesis and camera pose estimation from unposed image pairs. It predicts 3D Gaussian parameters (means, covariances, harmonics, opacities) in canonical space using a vision transformer backbone (CroCo, ResNet, or DINO) with a DPT head, then renders images via differentiable Gaussian splatting. The system eliminates the need for known camera poses during training/inference by estimating poses implicitly through geometric consistency losses. Built on PyTorch Lightning with Hydra configuration, it supports multi-GPU training and evaluation on datasets like RealEstate10K, ACID, DTU, and ScanNet++. The codebase emphasizes modularity through component registries, composable loss functions, and chainable data preprocessing shims.
This library provides a polyfill/shim for the IndexedDB API, implementing IndexedDB functionality using WebSQL as a backend storage mechanism. It works in both browser environments (leveraging native WebSQL where available) and Node.js (via websql-configurable/SQLite3). The library offers both invasive (automatically setting global indexedDB) and non-invasive (requiring explicit setup) versions. It implements the full IndexedDB specification including databases, object stores, indexes, transactions, and cursors, with proper handling of JavaScript value serialization through a structured cloning algorithm. The project features comprehensive test suites validating W3C compliance and active maintenance with regular updates to address compatibility issues and specification changes.
ProjectQ is a comprehensive quantum computing framework that enables users to design, simulate, optimize, and execute quantum circuits across multiple hardware backends. The framework features a modular compiler engine pipeline that transforms high-level quantum operations into hardware-specific gate sequences through decomposition rules and optimization passes. It provides abstraction layers for various quantum hardware (IBM Quantum, AQT, AWS Braket, Azure Quantum, IonQ) and includes a high-performance simulator with C++ backends. Key capabilities include resource estimation for circuit optimization, quantum algorithm libraries (arithmetic, Fourier transform), circuit visualization via LaTeX/TikZ, and meta-programming constructs for reversible computing. The framework is actively maintained with recent updates to dependencies, CI/CD pipelines, and backend integrations.
This repository is a monorepo containing Blockly plugins and example applications that demonstrate how to extend and customize the Blockly visual programming library. It provides reusable UI components (custom fields, toolbox enhancements, workspace minimap), real-time collaboration features, and developer tools for creating custom blocks. The architecture follows a plugin pattern where each extension is self-contained with its own source, tests, and package.json. Examples showcase practical integrations like block factories, serialization, and multi-user editing. The codebase is actively maintained with regular updates to track Blockly releases and employs Lerna for workspace management.
This repository provides a React component for international phone number input. It allows users to select a country from a dropdown (with flag icons) and input a phone number, which is formatted and validated according to the selected country's numbering plan. The component supports various formats (national, international) and integrates with popular form libraries like React Hook Form. It is built with a modular architecture, separating concerns into distinct components for country selection, input handling, and utility functions for phone number parsing and formatting. The library emphasizes reusability and customization through props that allow overriding components (input, country select, flags), providing custom metadata, and adjusting behavior (e.g., making the country calling code non-editable). It includes comprehensive TypeScript definitions and supports multiple distribution formats (ESM, CJS, UMD).
OpenBudgeteer is a personal budgeting application built with .NET 9.0 featuring a Blazor WebAssembly frontend and ASP.NET Core Web API backend. It implements the bucket budgeting methodology, allowing users to manage financial accounts, transactions, budget categories (buckets), automated transaction rules, recurring transactions, and financial reporting. The application supports multiple database providers (MySQL, PostgreSQL) and uses Redis for session storage and application settings. The architecture follows a layered approach with distinct data, service, presentation, and API layers. The data layer uses Entity Framework Core with a repository pattern for data access. The service layer provides business logic through generic and EF Core-specific services. The presentation layer employs the MVVM pattern with Blazor components and ViewModels handling UI state and logic. The API layer exposes RESTful endpoints for the frontend to interact with the backend. Key reusable components include a generic repository pattern, a generic service pattern, a ViewModel base class for Blazor, a fluent CSV bank transaction mapper, and a MudBlazor theme persistence service. These components demonstrate craftsmanship in abstraction and separation of concerns, making them suitable for extraction and reuse in other projects. The application is actively maintained with recent updates to .NET 9.0, MudBlazor, and ApexCharts, indicating low technical drift.
The Dropbox SDK for JavaScript (dropbox-sdk-js) is a lightweight, promise-based client library that provides a programmatic interface to the Dropbox v2 API. It supports both Node.js and browser environments, with full OAuth 2.0/PKCE authentication, file upload and download with resumable and parallel streaming, exponential backoff retries, progress reporting, and content hash validation. The library is auto-generated from Dropbox's API specification (Stone) and exposes hundreds of endpoint methods covering files, sharing, team management, users, and more. It is designed for production use with robust error handling, rate-limit awareness, and environment-aware binary types. The codebase demonstrates strong architectural separation of concerns: authentication logic is encapsulated in `DropboxAuth`, file transfer logic in dedicated uploader/downloader classes, and API request routing in generated route definitions. The SDK is well-tested with unit, integration, and TypeScript type tests, and includes comprehensive examples for both browser and Node.js usage patterns.
no-neck-pain.nvim is a Neovim plugin that centers the focused buffer in the middle of the screen by creating empty side buffers. It provides a non-opinionated, highly configurable zen-like editing experience without the intrusive UI changes typical of other 'distraction-free' plugins. The plugin manages window layout across multiple tabs, supports splits/vsplits, and integrates with popular file trees (NvimTree, Neo-tree), dashboards (alpha, snacks), and other tools like undotree, neotest, nvim-dap-ui, outline, and aerial. The architecture is modular Lua with clear separation into configuration, state management, UI creation, color handling, and utility modules. The core logic revolves around dynamic layout scanning via Neovim's winlayout API, event-driven re-initialization, and per-tab state tracking.
Aruba is a Ruby gem designed for testing command-line applications. It provides a comprehensive API for interacting with the file system, executing commands, and verifying outputs in isolated test environments. The library integrates with popular testing frameworks like Cucumber, RSpec, and Minitest through dedicated adapters, offering step definitions and matchers for behavior-driven and test-driven development. Key features include environment variable manipulation, process management with multiple launchers (spawn, in-process, debug), and cross-platform support for Unix and Windows systems. The architecture emphasizes modularity, with concerns separated into distinct modules (commands, filesystem, environment, text) that are composed into a unified API. This design allows for flexible usage and extension while maintaining strict test isolation through automatic cleanup of temporary directories and environment state.
Zircle is a JavaScript library for building circular, zoomable user interfaces using native custom elements. It recreates the original Zircle component family (z-canvas, z-view, z-spot, z-list, z-dialog) with Orbit for radial layout and Zumly for zoom navigation, eliminating the previous Vue dependency. The library provides a declarative HTML API via templates and a programmatic controller via `createZircle()`, supporting multiple independent canvases, themes, responsive sizing, and accessibility. Internal components include radial knobs, sliders, scrollbars, and pagination, all built with pointer, keyboard, and reduced-motion support. The architecture separates engine concerns: Orbit handles polar geometry and SVG progress rings; Zumly manages view history, zoom transitions, and hash routing; Zircle provides the component vocabulary and canvas lifecycle. Key entry points are the `z-canvas` custom element for declarative usage and the `createZircle()` function for programmatic integration. External dependencies are limited to @zumer/orbit and zumly, and the build produces ESM, standalone ESM, IIFE, and CSS bundles. The codebase demonstrates strong software craftsmanship with thorough browser tests across Chromium, Firefox, and WebKit, TypeScript declarations, a reusable base class (ZSurface) for light-DOM content distribution, and clean separation of control logic in control-utils.js. Several components (z-knob, z-scroll, z-dialog, z-list) implement complex interactive behaviors in a framework-agnostic way, making them valuable references for custom element development.
Radix is a native macOS disk space analysis application built with SwiftUI, Swift concurrency, and deep integration with macOS filesystem APIs (APFS clones, FSEvents, getattrlistbulk). The core scanning engine uses actor-based parallelism with atomic directory summarization, incremental rescanning via FSEvent history, and cancellation-aware sort and layout operations. The app features interactive sunburst and treemap visualizations with viewport pan/zoom, a searchable file browser with sorted tables, discard pile for staged trash operations, cross-snapshot comparison showing file changes over time, and export/import of scan snapshots in a versioned compressed archive format (.radixscan). Architecture follows a clean separation of Models, Services, ViewModels, and Features, with heavy use of dependency injection closures for testability. The codebase includes extensive unit, integration, and benchmark tests (Swift Testing) covering correctness, performance, and concurrency safety. It demonstrates high craftsmanship in file tree storage (compact arena with logical scopes), cancellation patterns, APFS storage accounting, and interactive chart geometry.
HY-WU is a machine learning application for text-guided image editing that implements a Functional Neural Memory framework. The system generates instance-conditioned LoRA (Low-Rank Adaptation) weights on-the-fly from hybrid image-instruction representations and injects them into a frozen backbone model (HunyuanImage-3.0-Instruct) during forward pass, enabling test-time adaptation without optimization. The architecture centers around a WUPipeline orchestrator that coordinates a parameter generator, the base model with PG mixin integration, and image processing components. Key entry points include the infer.py CLI script and gradio/app.py web interface. The system processes user prompts and input images through VAE encoding, condition extraction, parameter generation, LoRA injection, and image decoding to produce edited outputs.
RohanKar Launcher is a desktop game launcher built with Electron that allows users to browse, download, install, and launch games from a curated Archive.org collection. The application features a polished UI with game library browsing, detailed game information panels, download management with progress tracking, game launching with executable detection, playtime tracking, favorites, collections organization, controller/gamepad support, Steam integration, and automatic updates. It uses SQLite for persistent storage and implements intelligent caching for thumbnails and efficient file system operations.
This repository implements an MCP (Model Context Protocol) server that serves as a bridge between MCP clients (like Claude Code) and OpenAI's Codex CLI. It exposes Codex CLI functionality as MCP tools, enabling AI-powered code analysis, generation, and review within editor environments. The server handles stdio communication, manages conversational sessions, spawns Codex CLI processes with appropriate arguments, and processes their output to provide structured responses to clients. The architecture follows a modular design with clear separation between server setup, tool definitions, handlers, session management, and command execution utilities.
rtlamr2mqtt is a Home Assistant add-on and standalone Docker container that reads utility meters (electric, gas, water) using an RTL-SDR USB receiver and publishes readings to MQTT. The application runs rtl_tcp to interface with the SDR hardware and rtlamr to decode meter transmissions, parsing the JSON output and publishing formatted readings, HA auto-discovery payloads, and status to an MQTT broker. It supports sleep/wake cycles to conserve resources on Raspberry Pi, automatic process restart with backoff, USB device reset on timeout, and listen mode for discovering unknown meter IDs. The architecture is fully async (asyncio TaskGroup) with three independent tasks: MeterReader (rtlamr output parsing), MQTTPublisher (aiomqtt-based MQTT communication), and ManagedProcess (generic async subprocess lifecycle manager). It can run as a Home Assistant add-on (auto-discovers MQTT via Supervisor API) or standalone with a YAML config file.
This repository is the @playcanvas/supersplat-viewer npm package, an embeddable 3D Gaussian splatting viewer built on the PlayCanvas engine. It bootstraps a full web application around a PlayCanvas AppBase: createViewer(options) mounts a canvas and UI, loads .gsplat/.ply content plus optional skybox and voxel/mesh collision assets, and returns a ViewerHandle for capture, seeking, camera control, XR, and teardown. The default rendering path is the engine's sorted gsplat renderer; an opt-in WebGPU stochastic renderer provides unsorted splat projection, temporal anti-aliasing, occlusion culling, and depth-based picking. The repo is an application, not a data/research pipeline, though it contains research-grade rendering experiments and a dedicated benchmark harness.
AlbumentationsX is a production-grade Python library for fast, flexible image augmentation in deep learning and computer vision pipelines. It provides a comprehensive catalog of 70+ transforms covering pixel-level (blur, noise, color, weather effects), geometric (flip, rotate, crop, resize, distortion, perspective), dropout (coarse, grid, mask, pixel), mixing (mosaic, copy-paste, overlay), and domain adaptation (FDA, histogram matching, stain normalization). The library supports 2D and 3D data, including images, masks, bounding boxes (HBB and OBB), keypoints, and volumes, with seamless integration into PyTorch via tensor transforms. Its architecture is built on a clean inheritance hierarchy (BasicTransform → DualTransform/ImageOnlyTransform/Transform3D) with Pydantic-based parameter validation, serialization to JSON/YAML, and deterministic replay via ReplayCompose. The codebase is backed by an extensive test suite (property-based, regression, contract, and golden vector tests), ASV benchmarks, and a sophisticated CI/CD system with path-aware planning, performance budgets, and legal integrity checks. Telemetry (opt-out) collects anonymous usage statistics to guide development. While primarily a software library, it incorporates research-level transforms (e.g., PlanckianJitter, HEStain, dithering) and is used in both academic and production settings.