Today on Github, several interesting new projects emerged, spanning areas like PDF generation, AI orchestration and underwater scene understanding.

• apps.apple.com: This repository archives the complete Svelte/TypeScript frontend source code of the Apple App Store website, offering educational insights into its architecture and implementation.
• ModpackDebuggerKit: This Python application streamlines modpack debugging with features like snapshotting, new mod detection, dependency management, and automated binary search.
• xpet: xpet is a lightweight, customizable X11 desktop pet application built in C using XPM animation frames.
• NAUTILUS: NAUTILUS introduces a large multimodal model and a corresponding large-scale underwater dataset for advanced underwater scene understanding, incorporating a novel feature enhancement module inspired by underwater physical models.
• claude-code-resources: This repository provides a comprehensive guide and automation tools to help developers master Claude Code, offering documentation, production-ready agents, templates, and real-world examples for practical application.
• flawless-pdf-generator: This project offers free, high-quality HTML to PDF conversion with excellent font rendering and styling, overcoming limitations of online converters, especially for languages like Persian.
• blake-ai-orchestrator: This open-source AI orchestrator showcases modularity and clean API design via a decoupled system suitable for both learning and production environments.
• test-passkey: This library simplifies the creation of WebAuthn Passkeys for testing application authentication flows, particularly with tools like Playwright.
• YoloMarkFlow: YoloMarkFlow is a user-friendly image annotation tool designed for YOLO training, featuring a streamlined interface, plugin support for model training, and a unified image pool architecture for efficient data management.
• constraint-cache: This repository offers a production-ready solution for reducing LLM costs by normalizing queries and caching the results in Redis, achieving high cache hit rates without requiring embeddings or ML training.