This repository is a full-stack web application that simulates the WeatherStar 4000, the iconic 1990s cable television weather display system. It is a monolith composed of a Node.js/Express backend (`index.mjs`) and a single-page application frontend built with client-side JavaScript modules (`server/scripts/modules/`). The backend serves an EJS template and acts as a caching proxy for external weather APIs, while the frontend renders a retro-styled, TV-like interface with scanlines, custom fonts, and modular weather displays. The primary data flow involves fetching real-time weather data from the National Weather Service (NWS/NOAA) API, Iowa State Mesonet, and the HRRR model archive. The application supports two deployment modes: a static mode where the frontend calls APIs directly, and a server mode where the Express backend proxies and caches requests to reduce load and improve performance. Static data for cities and stations is generated by a suite of CLI tools in the `datagenerators/` directory. The codebase demonstrates high craftsmanship in several areas. The `proxy/cache.mjs` module implements a sophisticated HTTP cache with stale-while-revalidate and request deduplication, a genuinely reusable component. The `future-radar/zarr.mjs` module includes a custom Zarr v2 reader with a Blosc/LZ4 decoder, showcasing non-obvious data engineering skill. The `utils/setting.mjs` module provides a clean, observable pattern for UI settings. The data generators, particularly `stations-postprocessor.mjs`, employ a priority-based conflict resolution system for cleaning airport data. The project is actively maintained, with a well-defined build process (Gulp), CI/CD pipelines (GitHub Actions), and pinned dependencies. Its primary risk is tight coupling to the NOAA API ecosystem, making it US-only. The test suite is a smoke test rather than comprehensive unit tests. Overall, it is a strong, well-architected codebase with several genuinely reusable components, scoring an 8.0 on the quality rubric.