Contributing to JupyterLab

If you’re reading this section, you’re probably interested in contributing to JupyterLab. Welcome and thanks for your interest in contributing!

Please take a look at the Contributor documentation, familiarize yourself with using JupyterLab, and introduce yourself to the community (on the mailing list or discourse) and share what area of the project you are interested in working on. Please also see the Jupyter Community Guides.

We have labeled some issues as good first issue or help wanted that we believe are good examples of small, self-contained changes. We encourage those that are new to the code base to implement and/or ask questions about these issues. You are not required to ask for a permission to work on such issue, but if you do and do not get a reply within 48 hours please assume that no one else is working on it (even if someone previously volunteered) and open a pull request with proposed implementation. If you are not certain about the implementation, using draft pull requests is encouraged.

If you believe you’ve found a security vulnerability in JupyterLab or any Jupyter project, please report it to If you prefer to encrypt your security reports, you can use this PGP public key.

General Guidelines for Contributing

For general documentation about contributing to Jupyter projects, see the Project Jupyter Contributor Documentation and Code of Conduct.

We maintain the two most recently released major versions of JupyterLab, JupyterLab v2 and JupyterLab v3. After JupyterLab v4 is released, we will no longer maintain v2. All JupyterLab v2 users are strongly advised to upgrade as soon as possible.

All source code is written in TypeScript. See the Style Guide.

All source code is formatted using prettier. When code is modified and committed, all staged files will be automatically formatted using pre-commit git hooks (with help from the lint-staged and husky libraries). The benefit of using a code formatter like prettier is that it removes the topic of code style from the conversation when reviewing pull requests, thereby speeding up the review process.

You may also use the prettier npm script (e.g. npm run prettier or yarn prettier or jlpm prettier) to format the entire code base. We recommend installing a prettier extension for your code editor and configuring it to format your code with a keyboard shortcut or automatically on save.

Submitting a Pull Request Contribution

Generally, an issue should be opened describing a piece of proposed work and the issues it solves before a pull request is opened. A triager will ensure that your issue meets our definition of ready before we can merge any pull requests that relate to it.

Issue Management

Opening an issue lets community members participate in the design discussion, makes others aware of work being done, and sets the stage for a fruitful community interaction. When you open a new bug or enhancement request, please provide all the requested information in the issue template so that a responder will be able to triage your bug without delay.

A pull request should reference the issue it is addressing. Once the pull request is merged, the issue related to it will also be closed. If there is additional discussion around implementation the issue may be re-opened. Once 30 days have passed with no additional discussion, the lock bot will lock the issue. If additional discussion is desired, or if the pull request doesn’t fully address the locked issue, please open a new issue referencing the locked issue.

New issues are subject to triage. A developer with triage permissions (a triager) will do the following:

  1. Read the issue

  2. Search the existing issues and mark it as a duplicate if necessary

  3. If additional information is required, add a comment requesting it

  4. If the issue is ready to be worked on, assign it to a milestone

  5. Apply appropriate labels to the issue (see examples below)

A developer may start to work on an issue as soon as it is filed. Please work with a triager if they have any questions about your issue so that your changes can be merged in without delay.

Definition of Ready

One of the main goals of triage is to get issues into a state where they are ready for someone to work on. Once a triager is satisfied that an issue meets the definition below, they will remove the status:Needs Triage label from it. We will not merge a pull request for an issue that still needs triage.

Triagers should also ensure that the issue has appropriate labels that describe it, such as labels with the pkg: prefix for issues that affect one or more packages.

All requested information, where applicable, is provided. From the templates in JupyterLab’s issues:

For a bug:

  • Description, preferably including screen shots

  • Steps to reproduce

  • Expected behavior

  • Context, such as OS, browser, JupyterLab version, and output or log excerpts

For a feature request:

  • Description of the problem

  • Description of the proposed solution

  • Additional context

The issue should represent real, relevant, feasible work. In short, if a knowledgeable person were to be assigned this issue, they would be able to complete it with a reasonable amount of effort and assistance, and it furthers the goals of the Jupyter project.

  • Issues should be unique; triage is the best time to identify duplicates.

  • Bugs represent valid expectations for use of Jupyter products and services.

  • Expectations for security, performance, accessibility, and localization match generally-accepted norms in the community that uses Jupyter products.

  • The issue represents work that one developer can commit to owning, even if they collaborate with other developers for feedback. Excessively large issues should be split into multiple issues, each triaged individually, or into team-compass issues to discuss more substantive changes.

Labels Used by Triagers

All new bugs and enhancement requests have the status:Needs Triage label.

On a regular basis, Jupyter contributors (triage reviewers or triagers) review JupyterLab issues tagged with status:Needs Triage, starting with the oldest, and determine whether they meet the definition of ready.

Once triaged, if the issue is ready, the reviewer removes the status:Needs Triage label; no additional label is required. If there is not enough information in the issue as filed, the triage reviewer applies the status:Needs Info label and leaves status:Needs Triage in place. If an issue has remained in status:Needs Info for more than 14 days without any follow-up communication, the reviewer should apply status:Blocked. A blocked issue should be closed after another 14 days pass without a reply that unblocks it.

Our expectation is that every new issue should be examined within a week of its creation.

Tagging Issues with Labels

Users without the commit rights to the JupyterLab repository can tag issues with labels using the @meeseeksdev bot. For example: To apply the label foo and bar baz to an issue, comment @meeseeksdev tag foo "bar baz" on the issue.

Contributing from within the browser

Using the web interface - documented here - you can create and propose a change purely within your browser.

Using Binder, you can test the current master branch and your changes within the browser as well. We recommend you have at least 8 GB of RAM for this. To build and launch an instance of the latest JupyterLab master, open this link in a new tab. The build takes about 7 minutes to complete.

To test your own branch hosted on GitHub, enter it on If everything goes right, filling out the form takes about 2 minutes, and the build should take about 7 minutes again.

Setting up a local development environment

This section explains how to set up a local development environment. We assume you use GNU/Linux, Mac OS X, or Windows Subsystem for Linux.

Installing Node.js and jlpm

Building JupyterLab from its GitHub source code requires Node.js. The development version requires Node.js version 12+, as defined in the engines specification in dev_mode/package.json.

If you use conda, you can get it with:

conda install -c conda-forge nodejs

The canvas node package is not properly packaged for Mac OS X with ARM architectures (M1 and M2). To build JupyterLab on such platforms, you need a few additional packages, and to specify the pkg-config path:

conda install -c conda-forge pkg-config pango libpng cairo jpeg giflib librsvg glib
export PKG_CONFIG_PATH=$CONDA_PREFIX/lib/pkgconfig

If you use Homebrew on Mac OS X:

brew install node

You can also use the installer from the Node.js website.

To check which version of Node.js is installed:

node -v

Installing JupyterLab

Fork the JupyterLab repository.

Then use the following steps:

git clone<your-github-username>/jupyterlab.git
cd jupyterlab
pip install -e .
jlpm install
jlpm run build  # Build the dev mode assets (optional)

Additionally, you might want to execute the following optional commands:

# Build the core mode assets (optional)
jlpm run build:core

# Build the app dir assets (optional)
jupyter lab build


  • A few of the scripts will run “python”. If your target python is called something else (such as “python3”) then parts of the build will fail. You may wish to build in a conda environment, or make an alias.

  • Some of the packages used in the development environment require Python 3.0 or higher. If you encounter an ImportError during the installation, make sure Python 3.0+ is installed. Also, try using the Python 3.0+ version of pip or pip3 install -e . command to install JupyterLab from the forked repository.

  • If you see an error that says Call to 'pkg-config pixman-1 --libs' returned exit status 127 while in binding.gyp while running the pip install command above, you may be missing packages required by canvas. On macOS with Homebrew, you can add these packages by running brew install pkg-config cairo pango libpng jpeg giflib librsvg. If you are using mamba or conda, you can install the necessary packages with conda install -c conda-forge pkg-config glib pango pixman.

  • The jlpm command is a JupyterLab-provided, locked version of the yarn package manager. If you have yarn installed already, you can use the yarn command when developing, and it will use the local version of yarn in jupyterlab/yarn.js when run in the repository or a built application directory.

  • If you decide to use the jlpm command and encounter the jlpm: command not found error, try adding the user-level bin directory to your PATH environment variable. You already installed jlpm along with JupyterLab in the previous command, but jlpm might not be accessible due to PATH environment variable related issues. If you are using a Unix derivative (FreeBSD, GNU / Linux, OS X), you can achieve this by using export PATH="$HOME/.local/bin:$PATH" command.

  • At times, it may be necessary to clean your local repo with the command npm run clean:slate. This will clean the repository, and re-install and rebuild.

  • If pip gives a VersionConflict error, it usually means that the installed version of jupyterlab_server is out of date. Run pip install --upgrade jupyterlab_server to get the latest version.

  • To install JupyterLab in isolation for a single conda/virtual environment, you can add the --sys-prefix flag to the extension activation above; this will tie the installation to the sys.prefix location of your environment, without writing anything in your user-wide settings area (which are visible to all your envs):

  • You can run jlpm run build:dev:prod to build more accurate sourcemaps that show the original Typescript code when debugging. However, it takes a bit longer to build the sources, so is used only to build for production by default.

If you are using a version of Jupyter Notebook earlier than 5.3, then you must also run the following command to enable the JupyterLab server extension:

jupyter serverextension enable --py --sys-prefix jupyterlab

For installation instructions to write documentation, please see Writing Documentation

Run JupyterLab

Start JupyterLab in development mode:

jupyter lab --dev-mode

Development mode ensures that you are running the JavaScript assets that are built in the dev-installed Python package. Note that when running in dev mode, extensions will not be activated by default - refer documentation on extension development to know more.

When running in dev mode, a red stripe will appear at the top of the page; this is to indicate running an unreleased version.

If you want to change the TypeScript code and rebuild on the fly (needs page refresh after each rebuild):

jupyter lab --dev-mode --watch

Build and Run the Tests

jlpm run build:testutils
jlpm test

You can run tests for an individual package by changing to the appropriate package folder:

cd packages/notebook
jlpm run build:test
jlpm test

We use jest for all tests, so standard jest workflows apply. Tests can be debugged in either VSCode or Chrome. It can help to add an it.only to a specific test when debugging. All of the test* scripts in each package accept jest cli options.

VSCode Debugging

To debug in VSCode, open a package folder in VSCode. We provide a launch configuration in each package folder. In a terminal, run jlpm test:debug:watch. In VSCode, select “Attach to Jest” from the “Run” sidebar to begin debugging. See VSCode docs on debugging for more details.

Chrome Debugging

To debug in Chrome, run jlpm test:debug:watch in the terminal. Open Chrome and go to chrome://inspect/. Select the remote device and begin debugging.

Testing Utilities

There are some helper functions in testutils (which is a public npm package called @jupyterlab/testutils) that are used by many of the tests.

For tests that rely on @jupyterlab/services (starting kernels, interacting with files, etc.), there are two options. If a simple interaction is needed, the Mock namespace exposed by testutils has a number of mock implementations (see testutils/src/mock.ts). If a full server interaction is required, use the JupyterServer class.

We have a helper function called testEmission to help with writing tests that use Lumino signals, as well as a framePromise function to get a Promise for a requestAnimationFrame. We sometimes have to set a sentinel value inside a Promise and then check that the sentinel was set if we need a promise to run without blocking.


Translatable strings update

The translatable strings update cannot occur on patch release. They must be delayed on minor or major versions.

Performance Testing

If you are making a change that might affect how long it takes to load JupyterLab in the browser, we recommend doing some performance testing using Lighthouse. It let’s you easily compute a number of metrics, like page load time, for the site.

To use it, first build JupyterLab in dev mode:

jlpm run build:dev

Then, start JupyterLab using the dev build:

jupyter lab --dev-mode --NotebookApp.token=''  --no-browser

Now run Lighthouse against this local server and show the results:

jlpm run lighthouse --view

Using throttling

Lighthouse recommends using the system level comcast tool to throttle your network connection and emulate different scenarios. To use it, first install that tool using go:

go get

Then, before you run Lighthouse, enable the throttling (this requires sudo):

run lighthouse:throttling:start

This enables the “WIFI (good)” preset of comcast, which should emulate loading JupyterLab over a local network.

Then run the lighthouse tests:

jlpm run lighthouse [...]

Then disable the throttling after you are done:

jlpm run lighthouse:throttling:stop

Comparing results

Performance results are usually only useful in comparison to other results. For that reason, we have included a comparison script that can take two lighthouse results and show the changes between them.

Let’s say we want to compare the results of the production build of JupyterLab with the normal build. The production build minifies all the JavaScript, so should load a bit faster.

First, we build JupyterLab normally, start it up, profile it and save the results:

jlpm build:dev
jupyter lab --dev --NotebookApp.token='' --no-browser

# in new window
jlpm run lighthouse --output json --output-path normal.json

Then rebuild with the production build and retest:

jlpm run build:dev:prod
jupyter lab --dev --NotebookApp.token='' --no-browser

# in new window
jlpm run lighthouse --output json --output-path prod.json

Now we can use compare the two outputs:

jlpm run lighthouse:compare normal.json prod.json

This gives us a report of the relative differences between the audits in the two reports:

Resulting Output

normal.json -> prod.json

First Contentful Paint
- -62% Δ
- 1.9 s -> 0.7 s
- First Contentful Paint marks the time at which the first text or image is painted. Learn more.
First Meaningful Paint
- -50% Δ
- 2.5 s -> 1.3 s
- First Meaningful Paint measures when the primary content of a page is visible. Learn more.
Speed Index
- -48% Δ
- 2.6 s -> 1.3 s
- Speed Index shows how quickly the contents of a page are visibly populated. Learn more.
Estimated Input Latency
- 0% Δ
- 20 ms -> 20 ms
- Estimated Input Latency is an estimate of how long your app takes to respond to user input, in milliseconds, during the busiest 5s window of page load. If your latency is higher than 50 ms, users may perceive your app as laggy. Learn more.
Max Potential First Input Delay
- 9% Δ
- 200 ms -> 210 ms
- The maximum potential First Input Delay that your users could experience is the duration, in milliseconds, of the longest task. Learn more.
First CPU Idle
- -50% Δ
- 2.5 s -> 1.3 s
- First CPU Idle marks the first time at which the page’s main thread is quiet enough to handle input. Learn more.
Time to Interactive
- -52% Δ
- 2.5 s -> 1.2 s
- Time to interactive is the amount of time it takes for the page to become fully interactive. Learn more.
Avoid multiple page redirects
- -2% Δ
- Potential savings of 10 ms -> Potential savings of 10 ms
- Redirects introduce additional delays before the page can be loaded. Learn more.
Minimize main-thread work
- -54% Δ
- 2.1 s -> 1.0 s
- Consider reducing the time spent parsing, compiling and executing JS. You may find delivering smaller JS payloads helps with this.
JavaScript execution time
- -49% Δ
- 1.1 s -> 0.6 s
- Consider reducing the time spent parsing, compiling, and executing JS. You may find delivering smaller JS payloads helps with this. Learn more.
Preload key requests
- -100% Δ
- Potential savings of 240 ms ->
- Consider using <link rel=preload> to prioritize fetching resources that are currently requested later in page load. Learn more.
Uses efficient cache policy on static assets
- 0% Δ
- 1 resource found -> 1 resource found
- A long cache lifetime can speed up repeat visits to your page. Learn more.
Avoid enormous network payloads
- -86% Δ
- Total size was 30,131 KB -> Total size was 4,294 KB
- Large network payloads cost users real money and are highly correlated with long load times. Learn more.
Minify JavaScript
- -100% Δ
- Potential savings of 23,041 KB ->
- Minifying JavaScript files can reduce payload sizes and script parse time. Learn more.
Enable text compression
- -86% Δ
- Potential savings of 23,088 KB -> Potential savings of 3,112 KB
- Text-based resources should be served with compression (gzip, deflate or brotli) to minimize total network bytes. Learn more.
Avoid an excessive DOM size
- 0% Δ
- 1,268 elements -> 1,268 elements
- Browser engineers recommend pages contain fewer than ~1,500 DOM elements. The sweet spot is a tree depth < 32 elements and fewer than 60 children/parent element. A large DOM can increase memory usage, cause longer style calculations, and produce costly layout reflows. Learn more.

Visual Regression and UI Tests

As part of JupyterLab CI workflows, UI tests are run with visual regression checks. Galata is used for UI testing. Galata provides a high level API to control and inspect JupyterLab UI programmatically, testing tools and CLI to manage tests and other tasks.

UI tests are run for each commit into JupyterLab project in PRs or direct commits. Code changes can sometimes cause UI tests to fail for various reasons. After each test run, Galata generates a user friendly test result report which can be used to inspect failing UI tests. Result report shows the failure reason, call-stack up to the failure and detailed information on visual regression issues. For visual regression errors, reference image and test capture image, along with diff image generated during comparison are provided in the report. You can use these information to debug failing tests. Galata test report can be downloaded from GitHub Actions page for a UI test run. Test artifact is named ui-test-output and once you extract it, you can access the report by opening test/report/index.html in a browser window.

Main reasons for UI test failures are:

  1. A visual regression caused by code changes:

    Sometimes unintentional UI changes are introduced by modifications to project source code. Goal of visual regression testing is to detect this kind of UI changes. If your PR / commit is causing visual regression, then debug and fix the regression caused. You can locally run and debug the UI tests to fix the visual regression. Follow the instructions in steps 5-7 of Adding a new UI test suite guide in UI Testing documentation to locally debug and fix UI tests. Once you have a fix, you can push the change to your GitHub branch and test with GitHub actions.

  2. An intended update to user interface:

    If your code change is introducing an update to UI which causes existing UI Tests to fail, then you will need to update reference image(s) for the failing tests. In order to do that, you can post a comment on your PR with the following content:

    • please update galata snapshots: A bot will push a new commit to your PR updating galata test snaphsots.

    • please update snapshots: Combine the two previous comments effects.

For more information on UI Testing, please read the UI Testing developer documentation and Galata documentation.

Good Practices for Integration tests

Here are some good practices to follow when writing integration tests:

  • Don’t compare multiple screenshots in the same test; if the first comparison breaks, it will require running multiple times the CI workflow to fix all tests.

Contributing to the debugger front-end

To make changes to the debugger extension, a kernel with support for debugging is required.

Check out the user documentation to learn how to install such kernel: Debugger.

Then refresh the page and the debugger sidebar should appear in the right area.

The Debugger Adapter Protocol

The following diagram illustrates the types of messages sent between the JupyterLab extension and the kernel.

UML sequence diagram illustrating the interaction between a user, JupyterLab, and the kernel. From top to bottom, the timeline starts with opening the notebook and includes annotations where the debugger is started and stopped. Specific interactions and message types are discussed in the subsequent text.

Inspecting Debug Messages in VS Code

Inspecting the debug messages in VS Code can be useful to understand when debug requests are made (for example triggered by a UI action), and to compare the behavior of the JupyterLab debugger with the Python debugger in VS Code.

The first step is to create a test file and a debug configuration (launch.json):

An editor showing the menu for creating a debug configuration.
   "version": "0.2.0",
   "configurations": [
         "name": "Python: Current File",
         "type": "python",
         "request": "launch",
         "program": "${file}",
         "console": "integratedTerminal",
         "env": { "DEBUGPY_LOG_DIR": "/path/to/logs/folder" }

Then start the debugger:

A started debugging session in the editor. There are additional buttons in the upper right for navigating the session.

The content of the log file looks like this:


D00000.032: IDE --> {
               "command": "initialize",
               "arguments": {
                  "clientID": "vscode",
                  "clientName": "Visual Studio Code",
                  "adapterID": "python",
                  "pathFormat": "path",
                  "linesStartAt1": true,
                  "columnsStartAt1": true,
                  "supportsVariableType": true,
                  "supportsVariablePaging": true,
                  "supportsRunInTerminalRequest": true,
                  "locale": "en-us"
               "type": "request",
               "seq": 1



  • IDE = VS Code

  • PYD = pydev debugger

  • Messages follow the DAP


Build and run the stand-alone examples

To install and build the examples in the examples directory:

jlpm run build:examples

To run a specific example, change to the examples directory (i.e. examples/filebrowser) and enter:


Debugging in the Browser

All methods of building JupyterLab produce source maps. The source maps should be available in the source files view of your browser’s development tools under the webpack:// header.

When running JupyterLab normally, expand the ~ header to see the source maps for individual packages.

When running in --dev-mode, the core packages are available under packages/, while the third party libraries are available under ~. Note: it is recommended to use jupyter lab --watch --dev-mode while debugging.

When running a test, the packages will be available at the top level (e.g. application/src), and the current set of test files available under /src. Note: it is recommended to use jlpm run watch in the test folder while debugging test options. See above for more info.

High level Architecture

The JupyterLab application is made up of two major parts:

  • an npm package

  • a Jupyter server extension (Python package)

Each part is named jupyterlab. The developer tutorial documentation provides additional architecture information.

The NPM Packages

The repository consists of many npm packages that are managed using the lerna build tool. The npm package source files are in the packages/ subdirectory.

Build the NPM Packages from Source

git clone
cd jupyterlab
pip install -e .
jlpm run build:packages


jlpm run clean
jlpm run build:packages

Writing Documentation

Documentation is written in Markdown and reStructuredText. In particular, the documentation on our Read the Docs page is written in reStructuredText. To ensure that the Read the Docs page builds, you’ll need to install the documentation dependencies with conda:

conda env create -f docs/environment.yml
conda activate jupyterlab_documentation

To test the docs run:

python -m pytest --check-links -k .md . || python -m pytest --check-links -k .md --lf .

The Read the Docs pages can be built using make:

cd docs
make html

Or with jlpm:

jlpm run docs

Writing Style

  • The documentation should be written in the second person, referring to the reader as “you” and not using the first person plural “we.” The author of the documentation is not sitting next to the user, so using “we” can lead to frustration when things don’t work as expected.

  • Avoid words that trivialize using JupyterLab such as “simply” or “just.” Tasks that developers find simple or easy may not be for users.

  • Write in the active tense, so “drag the notebook cells…” rather than “notebook cells can be dragged…”

  • The beginning of each section should begin with a short (1-2 sentence) high-level description of the topic, feature or component.

  • Use “enable” rather than “allow” to indicate what JupyterLab makes possible for users. Using “allow” connotes that we are giving them permission, whereas “enable” connotes empowerment.

User Interface Naming Conventions

Documents, Files, and Activities

Files are referred to as either files or documents, depending on the context.

Documents are more human centered. If human viewing, interpretation, interaction is an important part of the experience, it is a document in that context. For example, notebooks and markdown files will often be referring to as documents unless referring to the file-ness aspect of it (e.g., the notebook filename).

Files are used in a less human-focused context. For example, we refer to files in relation to a file system or file name.

Activities can be either a document or another UI panel that is not file backed, such as terminals, consoles or the inspector. An open document or file is an activity in that it is represented by a panel that you can interact with.

Element Names

  • The generic content area of a tabbed UI is a panel, but prefer to refer to the more specific name, such as “File browser.” Tab bars have tabs which toggle panels.

  • The menu bar contains menu items, which have their own submenus.

  • The main work area can be referred to as the work area when the name is unambiguous.

  • When describing elements in the UI, colloquial names are preferred (e.g., “File browser” instead of “Files panel”).

The majority of names are written in lower case. These names include:

  • tab

  • panel

  • menu bar

  • sidebar

  • file

  • document

  • activity

  • tab bar

  • main work area

  • file browser

  • command palette

  • cell inspector

  • code console

The following sections of the user interface should be in title case, directly quoting a word in the UI:

  • File menu

  • Files tab

  • Running panel

  • Tabs panel

  • Simple Interface mode

The capitalized words match the label of the UI element the user is clicking on because there does not exist a good colloquial name for the tool, such as “file browser” or “command palette”.

See The JupyterLab Interface for descriptions of elements in the UI.

The Jupyter Server Extension

The Jupyter server extension source files are in the jupyterlab/ subdirectory. To use this extension, make sure the Jupyter Notebook server version 4.3 or later is installed.

Build the JupyterLab server extension

When you make a change to JupyterLab npm package source files, run:

jlpm run build

to build the changes, and then refresh your browser to see the changes.

To have the system build after each source file change, run:

jupyter lab --dev-mode --watch

Build Utilities

There is a range of build utilities for maintaining the repository. To get a suggested version for a library use jlpm run get:dependency foo. To update the version of a library across the repo use jlpm run update:dependency foo ^latest. To remove an unwanted dependency use jlpm run remove:dependency foo.

The key utility is jlpm run integrity, which ensures the integrity of the packages in the repo. It will:

  • Ensure the core package version dependencies match everywhere.

  • Ensure imported packages match dependencies.

  • Ensure a consistent version of all packages.

  • Manage the meta package.

The packages/metapackage package is used to build all of the TypeScript in the repository at once, instead of 50+ individual builds.

The integrity script also allows you to automatically add a dependency for a package by importing from it in the TypeScript file, and then running: jlpm run integrity from the repo root.

We also have scripts for creating and removing packages in packages/, jlpm run create:package and jlpm run remove:package. When creating a package, if it is meant to be included in the core bundle, add the jupyterlab: { coreDependency: true } metadata to the package.json. Packages with extension or mimeExtension metadata are considered to be a core dependency unless they are explicitly marked otherwise.

Testing Changes to External Packages

Linking/Unlinking Packages to JupyterLab

If you want to make changes to one of JupyterLab’s external packages (for example, Lumino) and test them out against your copy of JupyterLab, you can easily do so using the link command:

  1. Make your changes and then build the external package

  2. Register a link to the modified external package

    • navigate to the external package dir and run jlpm link

  3. Link JupyterLab to modded package

    • navigate to top level of your JupyterLab repo, then run jlpm link "<package-of-interest>"

You can then (re)build JupyterLab (eg jlpm run build) and your changes should be picked up by the build.

To restore JupyterLab to its original state, you use the unlink command:

  1. Unlink JupyterLab and modded package

    • navigate to top level of your JupyterLab repo, then run jlpm unlink "<package-of-interest>"

  2. Reinstall original version of the external package in JupyterLab

    • run jlpm install --check-files

You can then (re)build JupyterLab and everything should be back to default.

Possible Linking Pitfalls

If you’re working on an external project with more than one package, you’ll probably have to link in your copies of every package in the project, including those you made no changes to. Failing to do so may cause issues relating to duplication of shared state.

Specifically, when working with Lumino, you’ll probably have to link your copy of the "@lumino/messaging" package (in addition to whatever packages you actually made changes to). This is due to potential duplication of objects contained in the MessageLoop namespace provided by the messaging package.

Keyboard Shortcuts

Typeset keyboard shortcuts as follows:

  • Monospace typeface, with spaces between individual keys: Shift Enter.

  • For modifiers, use the platform independent word describing key: Shift.

  • For the Accel key use the phrase: Command/Ctrl.

  • Don’t use platform specific icons for modifier keys, as they are difficult to display in a platform specific way on Sphinx/RTD.

Screenshots and Animations

Our documentation should contain screenshots and animations that illustrate and demonstrate the software. Here are some guidelines for preparing them:

  • Make sure the screenshot does not contain copyrighted material (preferable), or the license is allowed in our documentation and clearly stated.

  • If taking a png screenshot, use the Firefox or Chrome developer tools to do the following:

    • set the browser viewport to 1280x720 pixels

    • set the device pixel ratio to 1:1 (i.e., non-hidpi, non-retina)

    • screenshot the entire viewport using the browser developer tools. Screenshots should not include any browser elements such as the browser address bar, browser title bar, etc., and should not contain any desktop background.

  • If creating a movie, adjust the settings as above (1280x720 viewport resolution, non-hidpi) and use a screen capture utility of your choice to capture just the browser viewport.

  • For PNGs, reduce their size using pngquant --speed 1 <filename>. The resulting filename will have -fs8 appended, so make sure to rename it and use the resulting file. Commit the optimized png file to the main repository. Each png file should be no more than a few hundred kilobytes.

  • For movies, upload them to the IPython/Jupyter YouTube channel and add them to the jupyterlab-media repository. To embed a movie in the documentation, use the website, which can be found by clicking on the ‘privacy-enhanced’ embedding option in the Share dialog on YouTube. Add the following parameters the end of the URL ?rel=0&amp;showinfo=0. This disables the video title and related video suggestions.

  • Screenshots or animations should be preceded by a sentence describing the content, such as “To open a file, double-click on its name in the File Browser:”.

  • We have custom CSS that will add box shadows, and proper sizing of screenshots and embedded YouTube videos. See examples in the documentation for how to embed these assets.

To help us organize screenshots and animations, please name the files with a prefix that matches the names of the source file in which they are used:


This will help us to keep track of the images as documentation content evolves.


  • By default, the application will load from the JupyterLab staging directory (default is <sys-prefix>/share/jupyter/lab/build. If you wish to run the core application in <git root>/jupyterlab/build, run jupyter lab --core-mode. This is the core application that will be shipped.

  • If working with extensions, see the extension documentation.

  • The npm modules are fully compatible with Node/Babel/ES6/ES5. Simply omit the type declarations when using a language other than TypeScript.