Py Math Viz

Use Python to produce clean, readable plots and export to PNG by default.

Assumptions (this workspace):

• Preferred interpreter: /root/.openclaw/workspace/.venv-math/bin/python
• Libraries: matplotlib, seaborn, numpy (already), opencv-python-headless (installed)

Defaults (quality rules)

• Output: PNG, dpi>=200, white background
• Always label axes; add title if it helps
• Use legend when multiple series exist (and place it so it doesn’t cover data)
• Avoid overlap:
  • prefer constrained_layout=True and/or plt.tight_layout()
  • rotate dense x-ticks, shorten tick labels
  • increase figure size instead of cramming
• Use consistent styling:
  • seaborn.set_theme(style="whitegrid")
  • colorblind-friendly palette if unsure

Quick start (script)

Render a plot from a JSON spec. To keep the workspace root clean, write outputs into out/ (or out/_scratch/ for throwaway experiments):

/root/.openclaw/workspace/.venv-math/bin/python \
  skills/py-math-viz/scripts/plot_from_spec.py \
  --spec spec.json --out out/plots/plot.png

Then send out/plots/plot.png (Telegram: use the message tool with filePath).

Plot spec

The spec is intentionally small. Read:

• references/spec.md — schema + examples
• references/quick-recipes.md — recipes index (points to JSON specs / seaborn snippets / image tiling)

OpenCV helpers

Tile multiple PNGs into one (for comparisons):

/root/.openclaw/workspace/.venv-math/bin/python \
  skills/py-math-viz/scripts/tile_images.py \
  --out tiled.png img1.png img2.png img3.png

When NOT to use

• Interactive dashboards (Plotly/Dash) unless explicitly asked
• 3D/mesh-heavy rendering (use specialized tools if needed)