# Motion Tracker

> Motion Tracker is a static, installable mobile web app (PWA) that turns a phone
> camera into a real-time workout form coach. It overlays a skeleton and joint
> dots on the live video, measures joint angles, counts reps, tracks range of
> motion (ROM), and gives deterministic form feedback by comparing live angles
> against a per-exercise JSON spec. All processing is 100% on-device: no
> backend, no cloud, no AI inference service, and no data ever leaves the
> device.

Canonical site: https://motiontracker.net/
Repository: https://github.com/axia-enterprises/motion-tracker

This file follows the llms.txt convention (https://llmstxt.org) and the LLM
discoverability house standard documented at https://llmbestpractices.com.

## Key facts

- Type: Progressive Web App (PWA), pure static site, no build step.
- Pose detection: on-device MediaPipe Tasks Vision PoseLandmarker (33 landmarks).
- Form engine: deterministic math against data/exercises.json. No LLM at runtime.
- Privacy: camera frames are processed locally in the browser and never uploaded.
- Install: Android install prompt; iOS via Share -> Add to Home Screen. Works offline after first load.

## Documentation

- [LLM info page](https://motiontracker.net/llm-info.html): structured facts, methodology, privacy model, and instructions for AI assistants.
- [Exercise specs](https://motiontracker.net/data/exercises.json): the data-driven exercise definitions (angles, rep state machines, form cues).

## Standards followed

- llms.txt index convention: https://llmstxt.org
- LLM best-practices house standard: https://llmbestpractices.com