krumjahn/applehealth

What it solves

This project provides a terminal-based AI assistant that allows users to analyze their Apple Health export.xml data. It solves the problem of making vast amounts of personal health metrics—often spanning years—accessible and actionable through a conversational interface, while allowing users to maintain full control over their data privacy.

How it works

The tool converts Apple Health XML exports into processed CSV or JSON datasets. It then uses a Chat REPL (Read-Eval-Print Loop) to allow users to ask questions in plain English. The system supports a wide range of AI models, including cloud-based LLMs (GPT-4o, Claude, Gemini) and local models via Ollama, Jan, LM Studio, or llama.cpp, ensuring that data can be processed entirely on-device if desired.

Who it’s for

It is designed for developers, biohackers, and researchers who want to perform deep analysis on their health data, spot trends, and generate charts without relying on generic health apps.

Highlights

• Conversational Analysis: Chat with your health data using plain English to identify trends and correlations.
• Local AI Support: Full integration with Ollama and other local servers for 100% private, on-device analysis.
• Multi-Model Flexibility: Supports over 50 AI models via LiteLLM.
• Data Reconciliation: Automatically handles overlapping measurements from multiple sources (Apple Watch, iPhone, Oura, Whoop) and normalizes units.
• Automated Visualizations: Generates charts for heart rate, sleep patterns, and workout intensity.
• High-Fidelity Export: Converts XML data into usable CSV/JSON formats while preserving metadata.