ipyniivue

This is a Jupyter Widget for Niivue based on anywidget. It is designed to visualize neuroimaging volumes, meshes, streamlines and connectomes and their corresponding statistical maps. This repository is designed for advanced developers to build and extend ipyniivue. The minimal installation end users is pip install ipyniivue. Most users will likely find the example jupyter notebooks the ideal introduction to the capability.

ipyniivue is a Jupyter widget for the Niivue. It enables interactive visualization of neuroimaging data directly in Jupyter environments, including volumetric images, surface meshes, streamlines, connectomes, and their associated statistical maps.

This repository is intended solely for advanced developers who want to build and extend ipyniivue. For others, the example Jupyter notebooks provides a simple installation.

Installation

For most users, the best introduction to ipyniivue’s capabilities is the collection of example Jupyter notebooks. However, the minimal installation is simply:

pip install ipyniivue

Documentation

See the Documentation for usage.

Development

ipyniivue uses the recommended hatchling build system, which is convenient to use via the hatch CLI. We recommend installing hatch globally (e.g., via pipx) and running the various commands defined within pyproject.toml. hatch will take care of creating and synchronizing a virtual environment with all dependencies defined in pyproject.toml.

Getting Started

These steps set up a local development environment for ipyniivue with live JavaScript rebuilding via anywidget.

Prerequisites

• Python 3.10+
• Node.js (LTS recommended)
• hatch installed (e.g. via pipx install hatch)
• Ensure ipyniivue is not already installed in your Python environment

Step-by-step setup

0. Verify your Python environment

Make sure hatch is available and ipyniivue is not installed:

hatch --version
pip show ipyniivue || echo "ipyniivue not installed (good)"

1. Clone the repository

git clone https://github.com/niivue/ipyniivue.git

2. Enter the repository

cd ipyniivue

3. Install JavaScript dependencies

From the repository root:

npm install

4. Start a Hatch shell (Python environment)

hatch shell

This activates the project’s Python development environment.

5. Install Python dependencies (editable mode)

pip install -e ".[dev]"

This activates the project’s Python development environment.

6. Start the JavaScript dev server

In the same terminal:

cd js
npm run dev

This runs the JavaScript build in watch mode, rebuilding the widget automatically when files change.

Leave this terminal running.

7. Open a new terminal tab

Open a second terminal tab/window, again from the ipyniivue repository root.

8. Start another Hatch shell

hatch shell

9. Install ipyniivue in editable mode

pip install -e .

This ensures Python picks up the locally built widget assets.

10. Launch JupyterLab

jupyter lab

You can now open an example notebook and develop ipyniivue with live JavaScript updates.

Development tips

• For best results with live updates, set:

  export ANYWIDGET_HMR=1

  before launching JupyterLab.
• If the widget does not appear or fails to load, ensure:
  • npm run dev is still running
  • pip install -e . was run after starting the JS build
• If you switch branches or pull updates that change JS files, restart npm run dev.

Command Cheat Sheet

Run these commands from the root of the project:

Command	Description
hatch run format	Format the project with ruff format . and apply linting with ruff --fix .
hatch run lint	Lint the project with ruff check .
hatch run test	Run unit tests with pytest
hatch run docs	Build docs with Sphinx

Alternatively, you can manually create a virtual environment and manage installation and dependencies with pip:

python3 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"

Making Changes to the JavaScript Code

This is an anywidget project, meaning the codebase is a hybrid of Python and JavaScript. The JavaScript code resides under the js/ directory and uses esbuild for bundling. Whenever you make changes to the JavaScript code, you need to rebuild the files under src/ipyniivue/static.

You have two options:

1. Build Once: Build the JavaScript code one time:

   npm run build

2. Start Development Server: Start a development server that automatically rebuilds the code as you make changes:

   npm run dev

   We recommend this approach for a smoother development experience.

Working with Jupyter

Once the development server is running, you can start JupyterLab or Visual Studio Code to develop the widget. When you're finished, stop the development server with Ctrl+C.

Note: To have anywidget automatically apply changes as you work, set the environment variable ANYWIDGET_HMR=1. You can set this directly in a notebook cell:

%env ANYWIDGET_HMR=1

or in the shell:

export ANYWIDGET_HMR=1

Release Process

Releases are automated using GitHub Actions via the release.yml workflow.

Steps to Create a New Release

1. Commit Changes: Ensure all your changes are committed.

2. Create a Tag: Create a new tag matching the pattern v*:

   git tag -a vX.X.X -m "vX.X.X"
   git push --follow-tags

3. Workflow Actions: When triggered, the workflow will:

   • Publish the package to PyPI with the tag version.
   • Generate a changelog based on conventional commits.
   • Create a GitHub Release with the changelog.

Changelog Generation

• We generate a changelog for GitHub releases with antfu/changelogithub.
• Each changelog entry is grouped and rendered based on conventional commits.
• It's recommended to follow the Conventional Commits specification.

Generating the Notebook Gallery

ipyniivue includes tooling to generate a static HTML gallery from example notebooks.

Each notebook is:

1. Executed in isolation
2. Converted to static HTML
3. Snapshotted using the final rendered canvas
4. Added to a gallery page with thumbnails that link to the generated HTML

All generated artifacts live in a test-only output directory and are never committed.

Gallery output location

tests-out/
  html/          # generated static HTML per notebook
  gallery/
    thumbnails/  # canvas thumbnails
    index.html   # gallery page

The entire tests-out/ directory is gitignored and safe to delete at any time.

Generate the gallery

npm run gallery

This will:

• Execute notebooks found in examples/
• Generate static HTML into tests-out/html/
• Create thumbnails from the final rendered canvas
• Generate tests-out/gallery/index.html

You can open the gallery locally:

open tests-out/gallery/index.html

Clean generated gallery artifacts

To remove all generated HTML, thumbnails, and executed notebooks:

npm run clean:generated-html

To preview what would be removed:

npm run clean:generated-html:dry

End-to-End (E2E) Testing with Playwright

ipyniivue uses Playwright to run end-to-end tests against pre-executed static HTML, rather than interacting with live Jupyter notebooks.

This approach provides:

• Deterministic results
• No UI flakiness
• No dependency on JupyterLab state
• True visual regression testing for WebGL output

E2E testing workflow

The E2E pipeline consists of two steps:

1. Prepare static HTML
2. Run Playwright tests

npm run test:e2e

This runs:

scripts/prepare-e2e.cjs   # executes notebooks → static HTML in tests-out/
playwright test           # loads static HTML and runs assertions

What is tested?

For each notebook:

• The static HTML loads successfully
• A <canvas> element is rendered
• A WebGL context is available
• The rendered canvas matches a saved visual snapshot

Snapshots are stored under:

e2e-tests/__screenshots__/

These snapshot images should be committed, as they define the visual baseline.

Updating visual snapshots

When a visualization changes intentionally:

npx playwright test --update-snapshots

This regenerates the baseline screenshots.

Generated test artifacts

All generated HTML and intermediate files are written to:

tests-out/

This directory is:

• Automatically reused between runs
• Cleaned via npm run clean:generated-html
• Ignored by git

Why static HTML testing?

We intentionally avoid driving the JupyterLab UI in tests.

Static HTML testing provides:

• Stable WebGL output
• Faster CI
• Clear visual diffs
• No notebook state pollution
• No source-control churn

This is the recommended approach for testing interactive notebook-based visualization libraries.