# ML / Data Scientist — Homebrew packages.
# ── base: every kitout workstation ──────────────────────────────────
brew "git"
brew "gh"
brew "ripgrep"
brew "fd"
brew "bat"
brew "eza"
brew "fzf"
brew "jq"
brew "starship"
brew "uv"                 # ubiquitous Python / tool runner
brew "node"               # coding agents + MCP servers
cask "ghostty"
cask "visual-studio-code"
cask "claude-code"

# ── ml / data scientist ──
# ── ML / DS (most of the stack is uv/pip, not brew) ──
brew "python@3.12"
brew "mlx"                      # Apple-silicon array framework (+ mlx-lm via pip)
brew "jupyterlab"
brew "xgboost"
brew "lightgbm"
brew "duckdb"
# cask "miniforge"             # if you prefer conda over uv
# torch, scikit-learn, pandas, polars, mlflow → `uv pip install` per project (script step)
