% Copyright (c) Jupyter Development Team.

% Distributed under the terms of the Modified BSD License.

(developer-extension-points)=

Common Extension Points

Most of the component parts of JupyterLab are designed to be extensible, and they provide services that can be requested in extensions via tokens. A list of common core tokens that extension authors can request is given in {ref}core-tokens.

Following the list of core tokens is a guide for using some of JupyterLab's most commonly-used extension points. However, it is not an exhaustive account of how to extend the application components, and more detailed descriptions of their public APIs may be found in the JupyterLab API and Lumino API documentation.

:depth: 1
:local: true

(core-tokens)=

Core Plugins

The core packages of JupyterLab provide the following plugins. They can be enabled or disabled using the command jupyter labextension enable <plugin-id> or jupyter labextension disable <plugin-id>.

Core Tokens

The core packages of JupyterLab provide many services for plugins. The tokens for these services are listed here, along with short descriptions of when you might want to use the services in your extensions.

Commands

Add a Command to the Command Registry

Perhaps the most common way to add functionality to JupyterLab is via commands. These are lightweight objects that include a function to execute combined with additional metadata, including how they are labeled and when they are to be enabled. The application has a single command registry, keyed by string command IDs, to which you can add your custom commands.

The commands added to the command registry can then be used to populate several of the JupyterLab user interface elements, including menus and the launcher.

Here is a sample block of code that adds a command to the application (given by app):

const commandID = 'my-command';
let toggled = false;

app.commands.addCommand(commandID, {
  label: 'My Cool Command',
  isEnabled: () => true,
  isVisible: () => true,
  isToggled: () => toggled,
  iconClass: 'some-css-icon-class',
  describedBy: {
    args: {
      type: 'object',
      properties: {
        text: {
          type: 'string',
          description: 'Optional text to log',
          default: ''
        },
        count: {
          type: 'number',
          description: 'Optional number of times to log the text',
          default: 1
        }
      }
    }
  },
  execute: args => {
    const text = args?.text || '';
    const count = args?.count || 1;
    for (let i = 0; i < count; i++) {
      console.log(`Executed ${commandID} with text: ${text}`);
    }
  }
});

This example adds a new command, which, when triggered, calls the execute function. isEnabled indicates whether the command is enabled, and determines whether renderings of it are greyed out. isToggled indicates whether to render a check mark next to the command. isVisible indicates whether to render the command at all. iconClass specifies a CSS class which can be used to display an icon next to renderings of the command. describedBy is an optional but recommended property that provides a JSON schema describing the command's arguments, which is useful for documentation, tooling, and ensuring consistency in how the command is invoked.

Each of isEnabled, isToggled, and isVisible can be either a boolean value or a function that returns a boolean value, in case you want to do some logic in order to determine those conditions.

Likewise, each of label and iconClass can be either a string value or a function that returns a string value.

There are several more options which can be passed into the command registry when adding new commands. These are documented here.

After a command has been added to the application command registry you can add them to various places in the application user interface, where they will be rendered using the metadata you provided.

For example, you can add a button to the Notebook toolbar to run the command with the CommandToolbarButtonComponent.

Add a Command to the Command Palette

In order to add an existing, registered command to the command palette, you need to request the {ts:interface}apputils.ICommandPalette token in your extension. Here is an example showing how to add a command to the command palette (given by palette):

palette.addItem({
  command: commandID,
  category: 'my-category',
  args: {}
});

The command ID is the same ID that you used when registering the command. You must also provide a category, which determines the subheading of the command palette in which to render the command. It can be a preexisting category (e.g., 'notebook'), or a new one of your own choosing.

The args are a JSON object that will be passed into your command's functions at render/execute time. You can use these to customize the behavior of your command depending on how it is invoked. For instance, you can pass in args: { isPalette: true }. Your command label function can then check the args it is provided for isPalette, and return a different label in that case. This can be useful to make a single command flexible enough to work in multiple contexts.

:::{note} Try this section in a browser playground:

1. Click Load Interactive Example.
2. In the playground editor toolbar (or command palette), run Load Current File As Extension.
3. A notification appears; click Open Command Palette, then run Extension Points Demo Command from category Common Extension Points. :::

<script type="text/plain" id="jp-plugin-playground-source-extension-points">
import { JupyterFrontEnd, JupyterFrontEndPlugin } from '@jupyterlab/application';

import { ICommandPalette } from '@jupyterlab/apputils';

const plugin: JupyterFrontEndPlugin<void> = {
  id: 'extension-points-command-demo:plugin',
  autoStart: true,
  requires: [ICommandPalette],
  activate: (app: JupyterFrontEnd, palette: ICommandPalette) => {
    const commandID = 'extension-points-command-demo:toggle';
    let toggled = false;

    app.commands.addCommand(commandID, {
      label: 'Extension Points Demo Command',
      caption: 'Added from the interactive docs example',
      isToggled: () => toggled,
      execute: () => {
        toggled = !toggled;
      }
    });

    palette.addItem({
      command: commandID,
      category: 'Common Extension Points'
    });

    void app.commands.execute('apputils:notify', {
      message: 'The example is ready. To see the command open Command Palette.',
      type: 'success',
      options: {
        autoClose: false,
        actions: [
          {
            label: 'Open Command Palette',
            commandId: 'apputils:activate-command-palette',
            displayType: 'accent'
          }
        ]
      }
    });
  }
};

export default plugin;
</script>
<div
  class="jp-plugin-playground-embed"
  data-playground-hide="all"
  data-playground-source-id="jp-plugin-playground-source-extension-points"
  data-playground-file-name="index.ts"
  data-playground-title="Common extension points interactive example"
  data-playground-description="Interactive command registry + command palette example."
></div>

<div
  class="jp-plugin-playground-embed"
  data-playground-hide="all"
  data-playground-query="fromURL=https://raw.githubusercontent.com/jupyterlab/extension-examples/0ff7dc53f876a2ad9388eb71c188156a572014fc/command-palette/src/index.ts"
  data-playground-title="Extension examples command palette interactive example"
  data-playground-description="Extension examples: command-palette."
></div>

(context-menu)=

Context Menu

JupyterLab has an application-wide context menu available as app.contextMenu. The application context menu is shown when the user right-clicks, and is populated with menu items that are most relevant to the thing that the user clicked.

The context menu system determines which items to show based on CSS selectors. It propagates up the DOM tree and tests whether a given HTML element matches the CSS selector provided by a given command.

Items can be added in the context menu in two ways:

1. Using the settings - this is the preferred way as they are configurable by the user.
2. Using the API - this is for advanced cases like dynamic menu or semantic items.

Here is an example showing how to add a command to the application context menu using the settings.

{
  "jupyter.lab.menus": {
    "context": [
      {
        "command": "my-command",
        "selector": ".jp-Notebook",
        "rank": 500
      }
    ]
  }
}

In this example, the command with id my-command is shown whenever the user right-clicks on a DOM element matching .jp-Notebook (that is to say, a notebook). The selector can be any valid CSS selector, and may target your own UI elements, or existing ones. A list of CSS selectors currently used by context menu commands is given in {ref}css-selectors.

Item must follow this definition:

:language: json
:lines: 14-34

where menuItem definition is:

:language: json

The same example using the API is shown below. See the Lumino docs for the item creation options.

app.contextMenu.addItem({
  command: commandID,
  selector: '.jp-Notebook'
});

<div
  class="jp-plugin-playground-embed"
  data-playground-hide="menu,statusbar"
  data-playground-query="path=extension-examples/context-menu/src/index.ts"
  data-playground-title="Extension examples context menu interactive example"
  data-playground-description="Extension examples: context-menu."
></div>

If you don't want JupyterLab's custom context menu to appear for your element, because you have your own right click behavior that you want to trigger, you can add the data-jp-suppress-context-menu data attribute to any node to have it and its children not trigger it.

For example, if you are building a custom React element, it would look like this:

function MyElement(props: {}) {
  return (
    <div data-jp-suppress-context-menu>
      <p>Hi</p>
      <p onContextMenu={() => {console.log("right clicked")}}>There</p>
    </div>
  )
}

Alternatively, you can use a 'contextmenu' event listener and call event.stopPropagation to prevent the application context menu handler from being called (it is listening in the bubble phase on the document). At this point you could show your own Lumino contextMenu, or simply stop propagation and let the system context menu be shown. This would look something like the following in a Widget subclass:

// In `onAfterAttach()`
this.node.addEventListener('contextmenu', this);

// In `handleEvent()`
case 'contextmenu':
  event.stopPropagation();

(copy-shareable-link)=

Copy Shareable Link

The file browser provides a context menu item "Copy Shareable Link". The desired behavior will vary by deployment and the users it serves. The file browser supports overriding the behavior of this item.

import { IFileBrowserFactory } from '@jupyterlab/filebrowser';

import {
  JupyterFrontEnd,
  JupyterFrontEndPlugin
} from '@jupyterlab/application';

const shareFile: JupyterFrontEndPlugin<void> = {
  activate: activateShareFile,
  id: commandID,
  requires: [IFileBrowserFactory],
  autoStart: true
};

function activateShareFile(
  app: JupyterFrontEnd,
  factory: IFileBrowserFactory
): void {
  const { commands } = app;
  const { tracker } = factory;

  commands.addCommand('filebrowser:share-main', {
    execute: () => {
      const widget = tracker.currentWidget;
      if (!widget) {
        return;
      }
      const path = encodeURI(widget.selectedItems().next().path);
      // Do something with path.
    },
    isVisible: () =>
      tracker.currentWidget &&
      Array.from(tracker.currentWidget.selectedItems()).length === 1,
    iconClass: 'jp-MaterialIcon jp-LinkIcon',
    label: 'Copy Shareable Link'
  });
}

Note that an extension providing a replacement plugin like this must either {ref}automatically disable <disabledExtensions> the replaced core plugin or the user must disable the core plugin manually:

jupyter labextension disable @jupyterlab/filebrowser-extension:share-file

Icons

See {ref}ui-components

Keyboard Shortcuts

There are two ways of adding keyboard shortcuts in JupyterLab. If you don't want the shortcuts to be user-configurable, you can add them directly to the application command registry:

app.commands.addKeyBinding({
  command: commandID,
  args: {},
  keys: ['Accel T'],
  selector: '.jp-Notebook'
});

In this example my-command command is mapped to Accel T, where Accel corresponds to Cmd on a Mac and Ctrl on Windows and Linux computers.

The behavior for keyboard shortcuts is very similar to that of the context menu: the shortcut handler propagates up the DOM tree from the focused element and tests each element against the registered selectors. If a match is found, then that command is executed with the provided args. Full documentation for the options for addKeyBinding can be found here.

JupyterLab also provides integration with its settings system for keyboard shortcuts. Your extension can provide a settings schema with a jupyter.lab.shortcuts key, declaring default keyboard shortcuts for a command:

{
  "jupyter.lab.shortcuts": [
    {
      "command": "my-command",
      "keys": ["Accel T"],
      "selector": ".jp-mod-searchable"
    }
  ]
}

Shortcuts added to the settings system will be editable by users.

From JupyterLab version 3.1 onwards, it is possible to execute multiple commands with a single shortcut. This requires you to define a keyboard shortcut for apputils:run-all-enabled command:

{
  "command": "apputils:run-all-enabled",
  "keys": ["Accel T"],
  "args": {
    "commands": ["my-command-1", "my-command-2"],
    "args": [{}, {}]
  },
  "selector": "body"
}

In this example my-command-1 and my-command-2 are passed in args of apputils:run-all-enabled command as commands list. You can optionally pass the command arguments of my-command-1 and my-command-2 in args of apputils:run-all-enabled command as args list.

Launcher

As with menus, keyboard shortcuts, and the command palette, new items can be added to the application launcher via commands. You can do this by requesting the {ts:interface}launcher.ILauncher token in your extension:

launcher.add({
  command: commandID,
  category: 'Other',
  rank: 0
});

In addition to providing a command ID, you also provide a category in which to put your item, (e.g. 'Notebook', or 'Other'), as well as a rank to determine its position among other items.

<div
  class="jp-plugin-playground-embed"
  data-playground-hide="all"
  data-playground-query="path=extension-examples/launcher/src/index.ts"
  data-playground-title="Extension examples launcher interactive example"
  data-playground-description="Extension examples: launcher."
></div>

(shell)=

Jupyter Front-End Shell

The Jupyter front-end shell ({ts:interface}application.JupyterFrontEnd.IShell) is used to add and interact with content in the application. The {ts:interface}application.JupyterFrontEnd.IShell interface provides an add() method for adding widgets to the application. In JupyterLab, the application shell consists of:

• A top area for things like top-level toolbars and information.
• A menu area for top-level menus, which is collapsed into the top area in multiple-document mode and put below it in single-document mode.
• left and right sidebar areas for collapsible content.
• A main work area for user activity.
• A down area for information content; like log console, contextual help.
• A bottom area for things like status bars.
• A header area for custom elements.

Top Area

The top area is intended to host most persistent user interface elements that span the whole session of a user. A toolbar named TopBar is available on the right of the main menu bar. For example, JupyterLab adds a user dropdown to that toolbar when started in collaborative mode.

See {ref}generic toolbars <generic-toolbar> to see how to add a toolbar or a custom widget to a toolbar.

You can use a numeric rank to control the ordering of top bar items in the settings; see {ref}Toolbar definitions <toolbar-settings-definition>.

JupyterLab adds a spacer widget to the top bar at rank 50 by default. You can then use the following guidelines to place your items:

• rank <= 50 to place items to the left side in the top bar
• rank > 50 to place items to the right side in the top bar

Left/Right Areas

The left and right sidebar areas of JupyterLab are intended to host more persistent user interface elements than the main area. That being said, extension authors are free to add whatever components they like to these areas. The outermost-level of the object that you add is expected to be a Lumino Widget, but that can host any content you like (such as React components).

As an example, the following code executes an application command to a terminal widget and then adds the terminal to the right area:

app.commands
  .execute('terminal:create-new')
  .then((terminal: WidgetModuleType.Terminal) => {
    app.shell.add(terminal, 'right');
  });

You can use a numeric rank to control the ordering of the left and right tabs:

app.shell.add(terminal, 'left', { rank: 600 });

The recommended ranges for this rank are:

• 0-500: reserved for first-party JupyterLab extensions.
• 501-899: reserved for third-party extensions.
• 900: The default rank if none is specified.
• 1000: The JupyterLab extension manager.

(mainmenu)=

Main Menu

There are two ways to extend JupyterLab's main menu.

1. Using the settings - this is the preferred way as they are configurable by the user.
2. Using the API - this is for advanced cases like dynamic menu or semantic items.

Settings-defined menu

JupyterLab provides integration with its settings system for menu definitions. Your extension can provide a settings schema with a jupyter.lab.menus key, declaring default menus. You don't need to set anything in the TypeScript code (except the command definitions).

To add a new menu with your extension command:

{
  "jupyter.lab.menus": {
    "main": [
      {
        "id": "jp-mainmenu-myextension",
        "label": "My Menu",
        "items": [
          {
            "command": "my-command",
            "rank": 500
          }
        ],
        "rank": 100
      }
    ]
  }
}

The menu item label will be set with the command label. For menus (and submenus), the label needs to be set explicitly with the label property.

Menu and item have a rank that will determine the elements order.

To add a new entry in an existing menu:

{
  "jupyter.lab.menus": {
    "main": [
      {
        "id": "jp-mainmenu-file",
        "items": [
          {
            "command": "my-command",
            "rank": 500
          }
        ]
      }
    ]
  }
}

Here is the list of default menu ids:

• File menu: jp-mainmenu-file

  • New file submenu: jp-mainmenu-file-new

• Edit menu: jp-mainmenu-edit

• View menu: jp-mainmenu-view

  • Appearance submenu: jp-mainmenu-view-appearance

• Run menu: jp-mainmenu-run

• Kernel menu: jp-mainmenu-kernel

• Tabs menu: jp-mainmenu-tabs

• Settings menu: jp-mainmenu-settings

• Help menu: jp-mainmenu-help

The default main menu is defined in the mainmenu-extension package settings.

A menu must respect the following schema:

:language: json
:lines: 5-13

And an item must follow:

:language: json

Menus added to the settings system will be editable by users using the mainmenu-extension settings. In particular, they can be disabled at the item or the menu level by setting the property disabled to true.

API-defined menu

To use the API, you should request the {ts:interface}mainmenu.IMainMenu token for your extension.

There are three main ways to extend:

1. You can add your own menu to the menu bar.
2. You can add new commands to the existing menus.
3. You can register your extension with one of the existing semantic menu items.

Adding a New Menu

To add a new menu to the menu bar, you need to create a new Lumino menu.

You can then add commands to the menu in a similar way to the command palette, and add that menu to the main menu bar:

const menu = new Menu({ commands: app.commands });
menu.addItem({
  command: commandID,
  args: {}
});

mainMenu.addMenu(menu, { rank: 40 });

As with the command palette, you can optionally pass in args to customize the rendering and execution behavior of the command in the menu context.

Adding a New Command to an Existing Menu

In many cases you will want to add your commands to the existing JupyterLab menus rather than creating a separate menu for your extension. Because the top-level JupyterLab menus are shared among many extensions, the API for adding items is slightly different. In this case, you provide a list of commands and a rank, and these commands will be displayed together in a separate group within an existing menu.

For instance, to add a command group with firstCommandID and secondCommandID to the File menu, you would do the following:

mainMenu.fileMenu.addGroup(
  [
    {
      command: firstCommandID
    },
    {
      command: secondCommandID
    }
  ],
  40 /* rank */
);

Registering a Semantic Menu Item

There are some commands in the JupyterLab menu system that are considered common and important enough that they are treated differently.

For instance, we anticipate that many activities may want to provide a command to close themselves and perform some cleanup operation (like closing a console and shutting down its kernel). Rather than having a proliferation of similar menu items for this common operation of "closing-and-cleanup", we provide a single command that can adapt itself to this use case, which we term a "semantic menu item". For this example, it is the File Menu closeAndCleaners set.

Here is an example of using the closeAndCleaners semantic menu item:

mainMenu.fileMenu.closeAndCleaners.add({
  id: 'notebook:close-and-shutdown',
  isEnabled: (w: Widget) => tracker.currentWidget !== null && tracker.has(w)
});

In this example, tracker is a {ref}widget-tracker, which allows the menu item to determine whether to delegate the menu command to your activity, and id is a the command identifier.

More examples for how to register semantic menu items are found throughout the JupyterLab code base. The available semantic menu items are:

• {ts:interface}mainmenu.IEditMenu.IUndoer: an activity that knows how to undo and redo.
• {ts:interface}mainmenu.IEditMenu.IClearer: an activity that knows how to clear its content.
• IEditMenu.IGoToLiner: an activity that knows how to jump to a given line.
• IFileMenu.ICloseAndCleaner: an activity that knows how to close and clean up after itself.
• IFileMenu.IConsoleCreator: an activity that knows how to create an attached code console for itself.
• {ts:interface}mainmenu.IKernelMenu.IKernelUser: an activity that can perform various kernel-related operations.
• {ts:interface}mainmenu.IRunMenu.ICodeRunner: an activity that can run code from its content.
• {ts:interface}mainmenu.IViewMenu.IEditorViewer: an activity that knows how to set various view-related options on a text editor that it owns.

Status Bar

JupyterLab's status bar is intended to show small pieces of contextual information. Like the left and right areas, it only expects a Lumino Widget, which might contain any kind of content. Since the status bar has limited space, you should endeavor to only add small widgets to it.

The following example shows how to place a status item that displays the current "busy" status for the application. This information is available from the {ts:interface}application.ILabStatus token, which we reference by a variable named labStatus. We place the statusWidget in the middle of the status bar. When the labStatus busy state changes, we update the text content of the statusWidget to reflect that.

const statusWidget = new Widget();
labStatus.busySignal.connect(() => {
  statusWidget.node.textContent = labStatus.isBusy ? 'Busy' : 'Idle';
});
statusBar.registerStatusItem('lab-status', {
  align: 'middle',
  item: statusWidget
});

(toolbar-registry)=

Toolbar Registry

JupyterLab provides a {ts:interface}apputils.IToolbarWidgetRegistry to define and customize toolbar widgets from the settings, which is similar to that defining the context menu and the main menu bar.

Document Widgets

A typical example is the notebook toolbar as in the snippet below:

function activatePlugin(
  app: JupyterFrontEnd,
  // ...
  toolbarRegistry: IToolbarWidgetRegistry | null,
  settingRegistry: ISettingRegistry | null
): NotebookWidgetFactory.IFactory {
  const { commands } = app;
  let toolbarFactory:
    | ((widget: NotebookPanel) => DocumentRegistry.IToolbarItem[])
    | undefined;

  // Register notebook toolbar specific widgets
  if (toolbarRegistry) {
    toolbarRegistry.addFactory<NotebookPanel>(FACTORY, 'cellType', panel =>
      ToolbarItems.createCellTypeItem(panel, translator)
    );

    toolbarRegistry.addFactory<NotebookPanel>(
      FACTORY,
      'kernelStatus',
      panel => Toolbar.createKernelStatusItem(panel.sessionContext, translator)
    );
    // etc...

    if (settingRegistry) {
      // Create the factory
      toolbarFactory = createToolbarFactory(
        toolbarRegistry,
        settingRegistry,
        // Factory name
        FACTORY,
        // Setting id in which the toolbar items are defined
        '@jupyterlab/notebook-extension:panel',
        translator
      );
    }
  }

  const factory = new NotebookWidgetFactory({
    name: FACTORY,
    fileTypes: ['notebook'],
    modelName: 'notebook',
    defaultFor: ['notebook'],
    // ...
    toolbarFactory,
    translator: translator
  });
  app.docRegistry.addWidgetFactory(factory);

The registry addFactory method allows an extension to provide special widget for a unique pair (factory name, toolbar item name). Then the helper createToolbarFactory can be used to extract the toolbar definition from the settings and build the factory to pass to the widget factory.

The default toolbar items can be defined across multiple extensions by providing an entry in the "jupyter.lab.toolbars" mapping. For example for the notebook panel:

(toolbar-settings-definition)=

"jupyter.lab.toolbars": {
  "Notebook": [ // Factory name
    // Item with non-default widget - it must be registered within an extension
    {
      "name": "save", // Unique toolbar item name
      "rank": 10 // Item rank
    },
    // Item with default button widget triggering a command
    { "name": "insert", "command": "notebook:insert-cell-below", "rank": 20 },
    { "name": "cut", "command": "notebook:cut-cell", "rank": 21 },
    { "name": "copy", "command": "notebook:copy-cell", "rank": 22 },
    { "name": "paste", "command": "notebook:paste-cell-below", "rank": 23 },
    { "name": "run", "command": "runmenu:run", "rank": 30 },
    { "name": "interrupt", "command": "kernelmenu:interrupt", "rank": 31 },
    { "name": "restart", "command": "kernelmenu:restart", "rank": 32 },
    {
      "name": "restart-and-run",
      "command": "notebook:restart-run-all",
      "rank": 33 // The default rank is 50
    },
    { "name": "cellType", "rank": 40 },
    // Horizontal spacer widget
    { "name": "spacer", "type": "spacer", "rank": 100 },
    { "name": "kernelName", "rank": 1000 },
    { "name": "kernelStatus", "rank": 1001 }
  ]
},
"jupyter.lab.transform": true,
"properties": {
  "toolbar": {
    "title": "Notebook panel toolbar items",
    "items": {
      "$ref": "#/definitions/toolbarItem"
    },
    "type": "array",
    "default": []
  }
}

The settings registry will merge those definitions from settings schema with any user-provided overrides (customizations) transparently and save them under the toolbar property in the final settings object. The toolbar list will be used to create the toolbar. Both the source settings schema and the final settings object are identified by the plugin ID passed to createToolbarFactory. The user can customize the toolbar by adding new items or overriding existing ones (like providing a different rank or adding "disabled": true to remove the item).

:::{note} You need to set jupyter.lab.transform to true in the plugin id that will gather all items. :::

What are transforms? The jupyter.lab.transform flag tells JupyterLab to wait for a transform function before loading the plugin. This allows dynamic modification of settings schemas, commonly used to merge toolbar/menu definitions from multiple extensions.

Loading order pitfall: Extensions providing transforms must register them early in activation, before dependent plugins load, otherwise those plugins will timeout waiting for the transform.

The current widget factories supporting the toolbar customization are:

• Notebook: Notebook panel toolbar
• Cell: Cell toolbar
• Editor: Text editor toolbar
• HTML Viewer: HTML Viewer toolbar
• CSVTable: CSV (Comma Separated Value) Viewer toolbar
• TSVTable: TSV (Tabulation Separated Value) Viewer toolbar

(toolbar-item)=

And the toolbar item must follow this definition:

:language: json

(generic-toolbar)=

Generic Widget with Toolbar

The logic detailed in the previous section can be used to customize any widgets with a toolbar.

The additional keys used in jupyter.lab.toolbars settings attributes are:

• Cell: Cell toolbar
• FileBrowser: Default file browser panel toolbar items
• TopBar: Top area toolbar (right of the main menu bar)

Here is an example for enabling a toolbar on a widget:

function activatePlugin(
  app: JupyterFrontEnd,
  // ...
  toolbarRegistry: IToolbarWidgetRegistry,
  settingRegistry: ISettingRegistry
): void {

  const browser = new FileBrowser();

  // Toolbar
  // - Define a custom toolbar item
  toolbarRegistry.addFactory(
    'FileBrowser', // Factory name
    'uploader',
    (browser: FileBrowser) =>
      new Uploader({ model: browser.model, translator })
  );

  // - Link the widget toolbar and its definition from the settings
  setToolbar(
    browser, // This widget is the one passed to the toolbar item factory
    createToolbarFactory(
      toolbarRegistry,
      settings,
      'FileBrowser', // Factory name
      plugin.id,
      translator
    ),
    // You can explicitly pass the toolbar widget if it is not accessible as `toolbar` attribute
    // toolbar,
  );

See {ref}Toolbar definitions <toolbar-settings-definition> example on how to define the toolbar items in the settings.

(widget-tracker)=

Widget Tracker

Often extensions will want to interact with documents and activities created by other extensions. For instance, an extension may want to inject some text into a notebook cell, or set a custom keymap, or close all documents of a certain type. Actions like these are typically done by widget trackers. Extensions keep track of instances of their activities in instances of {ts:class}apputils.WidgetTracker class, which are then provided as tokens so that other extensions may request them.

For instance, if you want to interact with notebooks, you should request the {ts:interface}notebook.INotebookTracker token. You can then use this tracker to iterate over, filter, and search all open notebooks. You can also use it to be notified via signals when notebooks are added and removed from the tracker.

Widget tracker tokens are provided for many activities in JupyterLab, including notebooks, consoles, text files, mime documents, and terminals. If you are adding your own activities to JupyterLab, you might consider providing a WidgetTracker token of your own, so that other extensions can make use of it.

Completion Providers

Both code completer and inline completer can be extended by registering an (inline) completion provider on the completion manager provided by the {ts:interface}completer.ICompletionProviderManager token.

Code Completer

A minimal code completion provider needs to implement the fetch and isApplicable methods, and define a unique identifier property, but the {ts:interface}completer.ICompletionProvider interface allows for much more extensive customization of the completer.

import {
  CompletionHandler,
  ICompletionProviderManager,
  ICompletionContext,
  ICompletionProvider
} from '@jupyterlab/completer';

class MyProvider implements ICompletionProvider {
  readonly identifier = 'my-provider';

  async isApplicable(context: ICompletionContext) {
    return true;
  }

  async fetch(
    request: CompletionHandler.IRequest,
    context: ICompletionContext
  ) {
    return {
      start: request.offset,
      end: request.offset,
      items: [{ label: 'option 1' }, { label: 'option 2' }]
    };
  }
}

const plugin: JupyterFrontEndPlugin<void> = {
  id: 'my-completer-extension:provider',
  autoStart: true,
  requires: [ICompletionProviderManager],
  activate: (
    app: JupyterFrontEnd,
    manager: ICompletionProviderManager
  ): void => {
    const provider = new MyProvider();
    manager.registerProvider(provider);
  }
};

A more detailed example is provided in the extension-examples repository.

For an example of an extensively customised completion provider, see the jupyterlab-lsp extension.

Inline Completer

% versionadded::4.1 % Experimental Inline Completion API was added in JupyterLab 4.1. % We welcome feedback on making it better for extension authors.

A minimal inline completion provider extension would only implement the required method fetch and define identifier and name properties, but a number of additional fields can be used for enhanced functionality, such as streaming, see the {ts:interface}completer.IInlineCompletionProvider documentation.

import {
  CompletionHandler,
  ICompletionProviderManager,
  IInlineCompletionContext,
  IInlineCompletionProvider
} from '@jupyterlab/completer';

class MyInlineProvider implements IInlineCompletionProvider {
  readonly identifier = 'my-provider';
  readonly name = 'My provider';

  async fetch(
    request: CompletionHandler.IRequest,
    context: IInlineCompletionContext
  ) {
    return {
      items: [{ insertText: 'suggestion 1' }, { insertText: 'suggestion 2' }]
    };
  }
}

const plugin: JupyterFrontEndPlugin<void> = {
  id: 'my-completer-extension:inline-provider',
  autoStart: true,
  requires: [ICompletionProviderManager],
  activate: (
    app: JupyterFrontEnd,
    manager: ICompletionProviderManager
  ): void => {
    const provider = new MyInlineProvider();
    manager.registerInlineProvider(provider);
  }
};

For an example of an inline completion provider with streaming support, see jupyterlab-transformers-completer.

State Database

The state database can be accessed by importing {ts:interface}statedb.IStateDB from @jupyterlab/statedb and adding it to the list of requires for a plugin:

const id = 'foo-extension:IFoo';

const IFoo = new Token<IFoo>(id);

interface IFoo {}

class Foo implements IFoo {}

const plugin: JupyterFrontEndPlugin<IFoo> = {
  id,
  autoStart: true,
  requires: [IStateDB],
  provides: IFoo,
  activate: (app: JupyterFrontEnd, state: IStateDB): IFoo => {
    const foo = new Foo();
    const key = `${id}:some-attribute`;

    // Load the saved plugin state and apply it once the app
    // has finished restoring its former layout.
    Promise.all([state.fetch(key), app.restored]).then(([saved]) => {
      /* Update `foo` with `saved`. */
    });

    // Fulfill the plugin contract by returning an `IFoo`.
    return foo;
  }
};

Kernel Subshells

Kernel subshells enable concurrent code execution within kernels that support them. Subshells are separate threads of execution that allow interaction with a kernel while it's busy executing long-running code, enabling non-blocking communication and parallel execution.

Kernel Support

Subshells are supported by:

• ipykernel 7.0.0+ (Python kernels) - Kernels advertise support via supported_features: ['kernel subshells'] in kernel info replies
• Other kernels implementing JEP 91

User Interface

For user interface details, see {ref}subshell-console.

Extension Development

Extension developers can use subshell functionality through the kernel service API:

import { INotebookTracker } from '@jupyterlab/notebook';

// Get the current kernel from a notebook
const current = tracker.currentWidget;
if (!current) return;

const kernel = current.sessionContext.session?.kernel;
if (!kernel) return;

// Check if kernel supports subshells
if (kernel.supportsSubshells) {
  // Create a new subshell
  const reply = await kernel.requestCreateSubshell({}).done;
  const subshellId = reply.content.subshell_id;
  console.log(`Created subshell: ${subshellId}`);

  // List existing subshells
  const listReply = await kernel.requestListSubshell({}).done;
  console.log(`Active subshells: ${listReply.content.subshell_id}`);

  // Execute code in a specific subshell
  const future = kernel.requestExecute(
    { code: 'print("Hello from subshell!")' },
    false, // disposeOnDone
    { subshell_id: subshellId } // metadata
  );
  await future.done;

  // Delete a subshell when done
  await kernel.requestDeleteSubshell({ subshell_id: subshellId }).done;
  console.log(`Deleted subshell: ${subshellId}`);
}

For detailed specifications, see JEP 91.

LSP Features

JupyterLab provides an infrastructure to communicate with the language servers. If the LSP services are activated and users have language servers installed, JupyterLab will start the language servers for the language of the opened notebook or file.

Extension authors can access the virtual documents and the associated LSP connection of opened document by requiring the {ts:interface}lsp.ILSPDocumentConnectionManager token from @jupyterlab/lsp.

Here is an example for making requests to the language server.

const plugin: JupyterFrontEndPlugin<void> = {
  id,
  autoStart: true,
  requires: [ILSPDocumentConnectionManager],
  activate: async (app: JupyterFrontEnd, manager: ILSPDocumentConnectionManager): Promise<void> => {

    // Get the path to the opened notebook of file
    const path = ...

    // Get the widget adapter of opened document
    const adapter = manager.adapters.get(path);
    if (!adapter) {
      return
    }
    // Get the associated virtual document of the opened document
    const virtualDocument = adapter.virtualDocument;

    // Get the LSP connection of the virtual document.
    const connection = manager.connections.get(virtualDocument.uri);
    ...
    // Send completion request to the language server
    const response = await connection.clientRequests['textDocument/completion'].request(params);
    ...
  }
};

Occasionally, LSP extensions include a CodeMirror extension to modify the code editor. In those cases, you can follow this example:

const renamePlugin: JupyterFrontEndPlugin<void> = {
  id,
  autoStart: true,
  requires: [
    ILSPDocumentConnectionManager,
    ILSPFeatureManager,
    IWidgetLSPAdapterTracker
  ],
  activate: (
    app: JupyterFrontEnd,
    connectionManager: ILSPDocumentConnectionManager,
    featureManager: ILSPFeatureManager,
    tracker: IWidgetLSPAdapterTracker
  ) => {
    const FEATURE_ID = 'rename_symbol';
    const extensionFactory: EditorAdapter.ILSPEditorExtensionFactory = {
      name: FEATURE_ID,
      factory: options => {
        const { editor, widgetAdapter } = options;

        // Get the editor
        const ceEditor: CodeEditor.IEditor | null = editor.getEditor();
        if (!ceEditor) {
          return null;
        }

        // Get the associated virtual document of the opened document
        if (!widgetAdapter.virtualDocument) {
          return null;
        }

        // Get the LSP connection of the virtual document.
        const connection = connectionManager.connections.get(
          widgetAdapter.virtualDocument.uri
        );
        if (!connection || !connection.provides('renameProvider')) {
          return null;
        }

        // Create a CodeMirror extension that listens for double click, gets the
        // selected code and makes a LSP request to rename it and prints the results.
        const ext = EditorView.domEventHandlers({
          dblclick: (e, view) => {
            const range = ceEditor.getSelection();
            const res = connection.clientRequests[
              'textDocument/rename'
            ].request({
              newName: 'test',
              position: {
                line: range.start.line,
                character: range.start.column
              },
              textDocument: { uri: widgetAdapter.virtualDocument!.uri }
            });

            res
              .then(value => {
                console.debug(value);
              })
              .catch(e => console.error);
          }
        });

        // Wrap the CodeMirror extension in the extension registry object.
        return EditorExtensionRegistry.createImmutableExtension(ext);
      }
    };

    // Register the extension with the LSP feature
    featureManager.register({
      id: FEATURE_ID,
      extensionFactory
    });
  }
};

Content Provisioning

The file system interactions can be customized by adding:

• a content provider, selectively replacing the way in which content is fetched and synchronized
• a drive, adding a new source of content, analogous to a physical hard drive

While both the content provider and drive are meant to provide custom implementations of the Contents API methods such as get() and save(), and optionally a custom sharedModelFactory, the intended use cases, and the way these are exposed in the user interface are different:

• Drive:

  • Use case: provision of additional content, not available on the default drive.
  • UI: paths of files and directories from the drive are prefixed with the drive name and colon.

• Content Provider:

  • Use case: modification of the protocol used for data retrieval (e.g., streaming of the content, real-time collaboration), by extending the Contents API methods for files which already exist on one of the drives.
  • UI: users will choose a widget factory with an associated content provider when selecting how to open a file using the "Open with" dropdown.

To register a custom drive, use the contents manager's addDrive method. The drive needs to follow the {ts:interface}services.Contents.IDrive interface. For drives that use a jupyter-server compliant REST API you may wish to extend or re-use the built-in {ts:class}services.Drive class, as demonstrated below:

import { Drive, ServerConnection } from '@jupyterlab/services';

const customDrivePlugin: JupyterFrontEndPlugin<void> = {
  id: 'my-extension:custom-drive',
  autoStart: true,
  activate: (app: JupyterFrontEnd) => {
    const myDrive = new Drive({
      apiEndpoint: 'api/contents',
      name: 'MyNetworkDrive',
      serverSettings: {
        baseUrl: 'https://your-jupyter-server.com'
        // ...
      } as ServerConnection.ISettings
    });
    app.serviceManager.contents.addDrive(myDrive);
  }
};

To use a content provider, first register it on a drive (or multiple drives):

import {
  Contents,
  ContentsManager,
  RestContentProvider
} from '@jupyterlab/services';

interface IMyContentChunk {
  /** URL allowing to fetch the content chunk */
  url: string;
}

interface CustomContentsModel extends Contents.IModel {
  /**
   * Specializes the content (which in `Contents.IModel` is just `any`).
   */
  content: IMyContentChunk[];
}

class CustomContentProvider extends RestContentProvider {
  async get(
    localPath: string,
    options?: Contents.IFetchOptions
  ): Promise<CustomContentsModel> {
    // Customize the behaviour of the `get` action to fetch a list of
    // content chunks from a custom API endpoint instead of the `get`

    try {
      return getChunks(); // this method needs to be implemented
    } catch {
      // fall back to the REST API on errors:
      const model = await super.get(localPath, options);
      return {
        ...model,
        content: []
      };
    }
  }

  // ...
}

const customContentProviderPlugin: JupyterFrontEndPlugin<void> = {
  id: 'my-extension:custom-content-provider',
  autoStart: true,
  activate: (app: JupyterFrontEnd) => {
    const drive = (app.serviceManager.contents as ContentsManager).defaultDrive;
    const registry = drive?.contentProviderRegistry;
    if (!registry) {
      // If content provider is a non-essential feature and support for JupyterLab <4.4 is desired:
      console.error(
        'Cannot initialize content provider: no content provider registry.'
      );
      return;
    }
    const customContentProvider = new CustomContentProvider({
      // These options are only required if extending the `RestContentProvider`.
      apiEndpoint: '/api/contents',
      serverSettings: app.serviceManager.serverSettings
    });
    registry.register('my-custom-provider', customContentProvider);
  }
};

and then create and register a widget factory which will understand how to make use of your custom content provider:

class ExampleWidgetFactory extends ABCWidgetFactory<
  ExampleDocWidget,
  ExampleDocModel
> {
  protected createNewWidget(
    context: DocumentRegistry.IContext<ExampleDocModel>
  ): ExampleDocWidget {
    return new ExampleDocWidget({
      context,
      content: new ExamplePanel(context)
    });
  }
}

const widgetFactoryPlugin: JupyterFrontEndPlugin<void> = {
  id: 'my-extension:custom-widget-factory',
  autoStart: true,
  activate: (app: JupyterFrontEnd) => {
    const widgetFactory = new ExampleWidgetFactory({
      name: FACTORY,
      modelName: 'example-model',
      fileTypes: ['example'],
      defaultFor: ['example'],
      // Instructs the document registry to use the custom provider
      // for context of widgets created with `ExampleWidgetFactory`.
      contentProviderId: 'my-custom-provider'
    });
    app.docRegistry.addWidgetFactory(widgetFactory);
  }
};

Where ExampleDocModel can now expect the CustomContentsModel rather than Contents.IModel:

class ExampleDocModel implements DocumentRegistry.IModel {
  // ...

  fromJSON(chunks: IMyContentChunk[]): void {
    this.sharedModel.transact(() => {
      let i = 0;
      for (const chunk of chunks) {
        const chunk = fetch(chunk.url);
        this.sharedModel.set(`chunk-${i}`, chunk);
        i += 1;
      }
    });
  }

  fromString(data: string): void {
    const chunks = JSON.parse(data) as IMyContentChunk[];
    return this.fromJSON(chunks);
  }
}

For a complete example of a widget factory (although not using a content provider), see the documents extension example.