jherr/tanmaxx-17

TanMaxx is a single-user workout tracker built to show off every library in the TanStack ecosystem at once. There are 17 TanStack packages in use here, from routing and data fetching to forms, tables, hotkeys, and AI integration. The project was created specifically as a demo app for a video tour of what TanStack can do, so the goal is breadth over practicality. The core idea is simple: log your lifting sets, track your workout history, and let an AI generate training programs for you. You can type a set entry in plain English and the app parses it automatically. The program generator runs as a four-step background workflow, and there is a streaming chat interface where you can ask questions about your sessions or request a new program. The app's most distinctive feature is something called The Maxx, a two-handle slider that goes from Deload all the way up through Volume, Strength, Peaking, and eventually into modes labeled GIGAMAXX and INJURY ZONE. The slider does two things at once: it tells the AI how intensely to program your training, and it changes the entire visual look of the app as you crank it higher. The style gets progressively more dramatic at higher settings. On the practical side, the tracker works offline and saves data locally using SQLite. There is a catalog of 5,238 exercises you can browse, a history view, and keyboard shortcuts for logging sets without touching the mouse. The hotkey system covers everything from adjusting weight and reps to jumping between screens with short key chords. The project is structured as a monorepo with a web app and a shared package that exposes the tracker as a discoverable skill for AI coding agents. That package lets tools like Cursor or Claude pick up the available server functions and call them directly. To run it yourself you need Node 24 and pnpm, plus an Anthropic API key for the AI features.