pyglobegl
anywidget wrapper for
globe.gl with integrations with
popular Python spatial packages.
📖 Full documentation: https://pyglobegl.pages.dev/
pyglobegl renders interactive WebGL globes in modern notebook environments
(Jupyter, JupyterLab, Colab, VS Code, marimo) and drives them from a friendly,
strongly typed Python API. It ships a prebuilt JupyterLab extension, so
pip install is all you need.
Installation
pip install pyglobegl
Optional spatial extras:
pip install "pyglobegl[geopandas]" # GeoPandas + Pandera helpers
pip install "pyglobegl[movingpandas]" # MovingPandas (includes GeoPandas)
See the installation guide
for uv and extras details.
Quick start
from IPython.display import display
from pyglobegl import (
GlobeConfig,
GlobeLayerConfig,
GlobeWidget,
PointDatum,
PointsLayerConfig,
)
points = [
PointDatum(lat=0, lng=0, altitude=0.25, color="#ff0000", label="Center"),
PointDatum(lat=15, lng=-45, altitude=0.12, color="#00ff00", label="West"),
]
config = GlobeConfig(
globe=GlobeLayerConfig(
globe_image_url="https://cdn.jsdelivr.net/npm/three-globe/example/img/earth-day.jpg"
),
points=PointsLayerConfig(points_data=points),
)
display(GlobeWidget(config=config))
Layer data are typed Pydantic models (no dynamic accessor remapping), and
defaults mirror globe.gl so omitted values still render predictably. See the
quick start for more.
Layers & features
Each globe.gl layer is exposed as typed data models:
Plus:
Documentation
Contributing
Build Assets (Release Checklist)
cd frontend && pnpm run builduv build
UI Test Artifacts
- Canvas captures are saved under
ui-artifacts as
{test-name}-pass-<timestamp>.png or {test-name}-fail-<timestamp>.png.
- Canvas comparisons use SSIM (structural similarity) with a fixed threshold
(currently
0.86).
