Open Source Ecosystems

Take a PyVista plotter object and show it on Streamlit as an interactive-ish component (as in it can be zoomed in/out, moved and rotated, but the visualization state is not returned). It can use either Panel or Trame to export the PyVista plotter to HTML; stpyvista places it within an iframe in Streamlit.

📦 Installation

pip install stpyvista

📚 Demo and documentation

✨ Minimal example

import streamlit as st
import pyvista as pv
from stpyvista import stpyvista

## Initialize a plotter object
plotter = pv.Plotter(window_size=[400,400])

## Create a mesh with a cube 
mesh = pv.Cube()

## Add some scalar field associated to the mesh
mesh['my_scalar'] = mesh.points[:, 2] * mesh.points[:, 0]

## Add mesh to the plotter
plotter.add_mesh(mesh, scalars='my_scalar', cmap='bwr')

## Final touches
plotter.view_isometric()
plotter.add_scalar_bar()
plotter.background_color = 'white'

## Pass a key to avoid re-rendering at each page change
stpyvista(plotter, key="pv_cube")

🎈 Deploying to Streamlit Community Cloud

By default, Community Cloud will run Python 3.12 🎈. Check this on New App → Advanced settings... → Python version.
Add stpyvista to the requirements.txt file.
Install procps, libgl1-mesa-glx and xvfb by adding them to the packages.txt file.
The Community Cloud is a Debian headless machine and Pyvista requires a virtual framebuffer to work. stpyvista.utils.start_xvfb checks if Xvfb is running and starts it if it was not.
```
from stpyvista.utils import start_xvfb

if "IS_XVFB_RUNNING" not in st.session_state:
  start_xvfb()
  st.session_state.IS_XVFB_RUNNING = True 
```

🚩 Known issues

cadquery conflicts: cadquery bundles an outdated version of vtk, which ends up conflicting with the vtk version that pyvista requires. Current solution is to install stpyvista before cadquery in a virtual environment.
NSInternalInconsistencyException thrown when running on macOS. Current solution is to deploy using a VM.

🍹 Solved issues

RuntimeError thrown when running Python 3.9. Upgrading to panel>=1.4.0 or using Python 3.10 fixes this.
Buggy colormaps when using panel==1.4.0. This issue was fixed with panel==1.4.1.
cmocean==4.0.1 will raise an exception while trying to register its colormaps to matplotlib, making PyVista to crash. This issue was fixed in cmocean==4.0.3.