This workspace contains utilities for extracting workout-related data from the MyFit production database and performing exploratory analysis in a Jupyter notebook. The workflow omits any user or session tables and focuses exclusively on training metrics.

• Python 3.12+
• Access to the production PostgreSQL database (Cockroach-compatible) with a valid connection string

1. Install and bootstrap uv

   # One-time install of uv (Linux/macOS)
   curl -LsSf https://astral.sh/uv/install.sh | sh
   
   # Inside the repo root
   cd /home/anonjr/Programming/myfit-analysis
   uv venv
   . .venv/bin/activate
   uv pip sync requirements.lock

   Tip: Add the environment to Jupyter via uv run python -m ipykernel install --user --name myfit-analysis if you plan to run the notebook outside VS Code. If you need an editable install of this repository (e.g., for packaging), run uv pip install -e . after syncing.

   Updating dependencies? Regenerate the lock with uv pip compile --extra dev pyproject.toml -o requirements.lock and re-run uv pip sync requirements.lock.

2. Configure environment variables

   Update .env with your PostgreSQL connection string:

   DATABASE_URL=postgresql://username:password@host:26257/myfit?sslmode=require

Run the extractor to pull the workout-related tables into local CSV files. The script reads DATABASE_URL from .env and skips any user or session data.

uv run python main.py --output-dir data

Use --dry-run to view the export plan without querying the database and --verbose for additional logging.

Open the notebook at notebooks/workout_analysis.ipynb in VS Code or launch Jupyter Lab with uv run jupyter lab and execute the cells. The notebook relies on the CSV exports located in ./data and produces seaborn and matplotlib visualisations covering:

• Weekly workout volume and consistency
• Distribution of total load per workout
• Average RIR trend over time

• Extend the extractor if new workout-adjacent tables are introduced.
• Layer in additional metrics (e.g., mesocycle adherence) or predictive models (e.g., fatigue scoring) on top of the exported data.