Skip to content
Kitbay
Submit
Skill

data-visualization

MITUpdated on Nov 11, 202631,198GitHub stars

A charting skill from the NVIDIA example agent in LangChain's deepagents repository.

What the skill teaches

A charting skill from the NVIDIA example agent in LangChain's deepagents repository. It tells the agent to draw charts with matplotlib and seaborn in a headless sandbox, setting the non-interactive backend before plotting, applying print-ready defaults and a colour-blind-safe palette, and saving each figure as a 300 dpi PNG in the workspace folder before opening it so the user can see it. It gives ready patterns for bar, line, scatter, heatmap, histogram and box plots and for multi-panel summaries, usually drawn from results of cuDF analysis or cuML models. It sits in an example folder and assumes that example's GPU sandbox and file tools.

Capabilities

  • Draw bar, line, scatter, heatmap, histogram and box plots
  • Build multi-panel analysis figures
  • Save charts as high-resolution PNG files