seaborn

name: seaborn description: | Creates statistical data visualizations and heatmaps. Use when: generating AIR heatmaps, violin plots for text features, topic distribution matrices, creating diverging color palettes, or building publication-quality statistical visualizations for HTML reports. allowed-tools: Read, Edit, Write, Glob, Grep, Bash

Seaborn Skill

Seaborn in this project is isolated to src/reporter.py for generating statistical visualizations embedded in HTML reports. All plots are rendered headlessly via matplotlib's Agg backend and converted to base64 PNG for embedding.

Quick Start

Diverging Heatmap with Threshold

import seaborn as sns
import matplotlib.pyplot as plt

fig, ax = plt.subplots(figsize=(12, 6))
cmap = sns.diverging_palette(10, 130, as_cmap=True)
sns.heatmap(
    pivot_df, annot=True, fmt=".2f", cmap=cmap, center=0.80,
    vmin=0.5, vmax=1.2, ax=ax, linewidths=0.5,
    cbar_kws={"label": "AIR (< 0.80 = material)"},
)

Violin Plot with Fallback

try:
    sns.violinplot(data=df, x="variant", y="flesch_reading_ease",
                   ax=ax, inner="box", palette="Set2", cut=0)
except Exception:
    sns.boxplot(data=df, x="variant", y="flesch_reading_ease",
                ax=ax, palette="Set2")

Color Palette Generation

# Categorical palette for discrete groups
colors = sns.color_palette("Set2", n_colors=6)

# Diverging palette for threshold-centered heatmaps
cmap = sns.diverging_palette(130, 10, as_cmap=True)  # Green to Red

Key Concepts

Concept	Usage	Example
Diverging palette	Threshold-centered heatmaps	`sns.diverging_palette(10, 130, as_cmap=True)`
Categorical palette	Discrete group colors	`sns.color_palette("Set2", n)`
Heatmap center	Statistical threshold	`center=0.80` for AIR materiality
Violin inner	Distribution + summary	`inner="box"` shows box inside violin
Cut parameter	Limit violin extent	`cut=0` constrains to data range

Common Patterns

Heatmap with Statistical Threshold

When: Visualizing AIR, effect sizes, or metrics with meaningful cutoffs.

cmap = sns.diverging_palette(10, 130, as_cmap=True)
sns.heatmap(
    pivot, annot=True, fmt=".2f", cmap=cmap, center=0.80,
    vmin=0.5, vmax=1.2, ax=ax, linewidths=0.5,
    cbar_kws={"label": "AIR (< 0.80 = material)"},
)

Count Heatmap

When: Showing frequency cross-tabulations.

sns.heatmap(ct, annot=True, fmt=".0f", cmap="YlOrRd", ax=ax,
            linewidths=0.5, cbar_kws={"label": "Count"})

Related Skills

matplotlib skill for figure creation and saving
pandas skill for pivot tables and crosstabs
numpy skill for numerical operations in visualizations