napari-colab 0.1.1

Interactive napari viewer inside Google Colab via noVNC

napari-colab

                         _  ____      _       _
 _ __   __ _ _ __   __ _| |/ ___|___ | | __ _| |__
| '_ \ / _` | '_ \ / _` | | |   / _ \| |/ _` | '_ \
| | | | (_| | |_) | (_| | | |__| (_) | | (_| | |_) |
|_| |_|\__,_| .__/ \__,_|_|\____\___/|_|\__,_|_.__/
             |_|

napari :heart: Colab — because microscopy images deserve a real viewer, even in the cloud.

Interactive napari viewer inside Google Colab via noVNC.

Warning: This package is designed exclusively for Google Colab. It relies on root access via apt-get, a virtual framebuffer (Xvfb), and the Colab proxy to tunnel ports to your browser. It will not work on local machines, JupyterHub, or other hosted notebook services. If you can run napari locally, you should — it will be faster and more responsive.

Installation

pip install napari-colab

Note: the system packages (xvfb, x11vnc, scrot, xcb libraries) are installed automatically by setup() on first run via apt-get.

Quick start

# Cell 1 — once per session
from napari_colab import setup, open_viewer, screenshot, shutdown

# Cell 2 — launch viewer (URL printed here)
viewer = open_viewer(width=1800, height=1000)

# Cell 3 — load your data
from pathlib import Path
images = sorted(Path('my_images/').glob('*.tiff'))
labels = sorted(Path('my_labels/').glob('*.tiff'))

viewer.open(images, stack=True, name='raw')
viewer.open(labels, stack=True, layer_type='labels', name='segmentation')

# Cell 4 — optional extras
viewer.dims.ndisplay = 3
viewer.camera.angles = (30, 45, 0)
screenshot()   # inline preview

API

Function	Description
setup()	Install deps, start Xvfb + x11vnc + noVNC
open_viewer(width, height)	Launch napari subprocess, return ViewerProxy
screenshot()	Capture virtual display inline in notebook
shutdown()	Kill napari and display helpers

ViewerProxy

viewer.open(paths, stack=True, name='layer', layer_type='image'|'labels')
viewer.add_image(np_array, name='...', colormap='...', blending='...')
viewer.add_labels(np_array, name='...')
viewer.dims.ndisplay = 2 | 3
viewer.camera.angles = (rx, ry, rz)
viewer.camera.zoom   = 1.5
viewer.reset_view()
viewer.window.resize(w, h)

How it works

Xvfb (:99)   <- napari + Qt render here (software OpenGL / Mesa)
    |
x11vnc       <- captures framebuffer -> VNC (port 5900)
    |
websockify   <- VNC -> WebSocket bridge (port 6080)
    |
noVNC        <- web client at /vnc.html
    |
Colab proxy  <- tunnels port 6080 to your browser tab

napari runs in its own subprocess so the Colab kernel never crashes. Commands (open, add_image, camera, etc.) are forwarded via a JSON-over-TCP socket. A QTimer drains the command queue every 200 ms on the Qt main thread — the only thread allowed to touch Qt objects.

Troubleshooting

OpenCV Qt conflict (Cellpose and similar packages)

Packages like Cellpose pull in opencv-python, which bundles its own Qt plugins. These conflict with the Qt used by napari, causing a fatal xcb plugin error:

qt.qpa.plugin: Could not load the Qt platform plugin "xcb"

napari-colab depends on opencv-python-headless (which is functionally identical but Qt-free) to avoid this. However, if you install a package that depends on opencv-python after napari-colab, pip may re-install the full OpenCV and reintroduce the conflict.

Fix:

pip uninstall -y opencv-python && pip install opencv-python-headless

Prevention: install OpenCV-dependent packages (cellpose, etc.) before napari-colab so that the headless variant wins:

pip install cellpose
pip install napari-colab

License

MIT