nolegend

Remove the legend to become one.

Tufte-style templates and helpers for Plotly. One import, beautiful charts.

Based on Edward Tufte's The Visual Display of Quantitative Information: maximize data-ink ratio, remove chartjunk, direct-label instead of legends, use range frames, and apply colorblind-safe palettes.

Why this exists

There is no Tufte package for Plotly on PyPI. matplotlib has matplotx, plotnine has theme_tufte(), ggplot2 has ggthemes — Plotly has nothing. The closest built-in template (simple_white) removes gridlines but stops there. The default Plotly output is noisy: legends, gridlines, aggressive colors, titles that describe axes instead of insights.

nolegend fills that gap with opinionated defaults and easy overrides.

What this is NOT

• Not a general visualization framework. Plotly only.
• Not a dashboard framework. Use marimo for layout.
• Not a Tufte orthodoxy enforcer. We take the useful parts (data-ink ratio, range frames, direct labeling, small multiples) and skip the dogma.
• Not a replacement for domain judgment. The library guides chart selection and color use, but the human picks the story.

Install

uv add nolegend

Claude Code skill

nolegend ships a visualization skill that teaches Claude Code the px → go workflow, Tufte principles, color rules, and marimo patterns.

npx skills add mojzis/nolegend

Once installed, Claude Code will automatically apply nolegend conventions whenever you create Plotly charts.

Quick start

import plotly.express as px
import nolegend

nolegend.activate()  # set as default template for all figures

df = px.data.gapminder().query("continent == 'Europe' and year == 2007")
fig = px.bar(df.nlargest(10, "gdpPercap"), x="country", y="gdpPercap")
fig.show()

The px → go workflow

Start fast with Plotly Express, then refine:

import plotly.express as px
import nolegend

# 1. Quick draft with px
fig = px.line(
    df, x="year", y="revenue", color="product",
    template="tufte",
)

# 2. Refine: axes span only the data
nolegend.range_frame(fig)

# 3. Refine: labels on the lines, not in a legend box
nolegend.direct_label(fig)

# 4. Refine: call out the key moment
nolegend.annotate_point(fig, x=2023, y=peak, text="Record quarter")

fig.show()

Color palettes

Six curated, colorblind-safe palettes:

# Use a specific palette
fig = px.line(df, ..., template=nolegend.with_palette("earth"))

# Spotlight: highlight one series, gray out the rest
colors = nolegend.QUALITATIVE.spotlight(index=2)

Hard ceiling of 5 colors per chart. Gray is a deliberate design tool, not a leftover.

Sparklines

Word-sized graphics, inline with text:

spark = nolegend.sparkline([1, 4, 2, 5, 3, 7], width=150, height=30)
spark.show()

Dark mode

nolegend.activate(dark=True)
# or per-figure:
fig = px.scatter(df, ..., template="tufte_dark")

In marimo notebooks

import marimo as mo
import nolegend

# Auto-detect marimo theme
nolegend.activate(dark=mo.app_meta().theme == "dark")

# Reactive plotly charts
fig = px.scatter(df, x="x", y="y", template="tufte")
chart = mo.ui.plotly(fig)

API

License

MIT