Toggle Dark/Light/Auto modeToggle Dark/Light/Auto modeToggle Dark/Light/Auto modeBack to homepage

Model Sources

When drawing a diagram with Sprotty we need a place to define and update the schema of the diagram to draw. Sprotty uses model sources to do this. Sprotty currently offers two different model sources: The LocalModelSource for local models and the DiagramServer for remote ones.
flowchart TD;
ActionDispatcher <-.->|Action| ModelSource
ModelSource --- LocalModelSource
ModelSource --- DiagramServerProxy
DiagramServerProxy <-.->|Action| DiagramServer
Regardless of where your model-source is located, Sprotty handles them in a similar fashion. All communication between ActionDispatcher and model source is always through actions and is bi-directional. This is a powerful feature of Sprotty as it allows both flexibility regarding where and how the Diagram is generated, as well as changing or updating and reacting to interactions with the diagram simultaneously.

The following sections will explain how to use and work with the different types of model sources.

General usage

Regardless of the model source we are using, the first thing we have to do is to register our model source in the front-end DI-container like this:


After that, we can retrieve the model source with the following code to further configure and use it.

const modelSource = container.get<LocalModelSource>(TYPES.ModelSource);

Local Model Source

A LocalModelSource instance allows us to set and modify the model through function calls, and keeps the model schema saved locally. To see how to use this model source, let’s have a look at the following example:

import {SNode} from 'sprotty-protocol';
export default runExample() {
    const container = createContainer('sprotty-showcase');
    const modelSource = container.get<LocalModelSource>(TYPES.ModelSource);
        type: 'graph'
        children: [
                type: 'node',
                id: 'main_node',
                text: 'node1',
                position: {x: 0, y: 0}

    document.getElementById('addButton').addEventListener('click', () => {
            parentId: 'graph',
            element: <SNode>{
                type: 'node',
                id: 'new_node',
                text: 'new node',
                position: {x: 100, y: 100}

In this example, we have a hard-coded data structure containing all the initial elements which are set in the model. Be aware that by defining the model this way we are defining the model schema - a data structure describing the model - and not the actual model itself. The model in this context means instances of the classes containing logic Sprotty uses for rendering. The model schema is used by Sprotty to generate the model. Due to this, we should use the interfaces for our nodes, edges, etc. from sprotty-protocol and not the classes from the sprotty main package. For disambiguation, model classes are suffixed with Impl in Sprotty, while interfaces (contained in the package sprotty-protocol) are not. This means that SNode is an interface, while SNodeImpl is a class.

After defining the model schema we can then use methods like addElements() from our LocalModelSource to add new nodes at the click of a button. The LocalModelSource then handles updating the model and notifying the ActionDispatcher about the update, so that the view can receive an animated update.

Through methods like the ones outlined above, the LocalModelSource can also be used as a facade over the action-based API of Sprotty. It handles actions for bounds calculation and model updates.

Diagram Server

When the model needs to be generated from a remote source, like in a worker or from a server, we can use Sprotty’s DiagramServer model source. It communicates with the client through Action objects which can be serialized to plain JSON.

On the client-side, instead of registering an actual ModelSource we can use a DiagramServerProxy. The Proxy handles the communication and forwards actions to the ActionDispatcher. Out of the box, Sprotty offers the WebSocketDiagramServerProxy for communicating through WebSockets with the DiagramServer. Should a different form of communication be necessary we would have to create a custom proxy.

Using the WebSocketDiagramServerProxy is quite simple. We just need to call listen on the ModelSource and pass it the WebSocket we’re communicating with.

const modelSource = container.get<WebSocketDiagramServerProxy>(TYPES.ModelSource);

For creating the DiagramServer itself, let’s look at an example.

// Creating a new websocket server
const wss = new WebSocketServer.Server({ port: 8080 });

// create our DiagramServices
const elkFactory: ElkFactory = () => new SocketElkServer();
const services: DiagramServices = {
    DiagramGenerator: new RandomGraphGenerator(),
    ModelLayoutEngine: new ElkLayoutEngine(elkFactory)
// Creating connection using websocket
wss.on("connection", ws => {

    const diagramServer = new DiagramServer(action => {
    }, services)

    ws.on('message' data => {

In the example above, we assume we have a simple nodeJs WebSocket server and want to create a DiagramServer for it.

As we can see, there are two parts to creating the DiagramServer. First, we need a dispatch method to send actions from the server to the client. This can be as simple as calling ws.send() with the serialized action. Second, we need the DiagramServices. The DiagramServices type looks like this:

export interface DiagramServices {
    readonly DiagramGenerator: IDiagramGenerator
    readonly ModelLayoutEngine?: IModelLayoutEngine
    readonly ServerActionHandlerRegistry?: ServerActionHandlerRegistry

There are 3 components to the DiagramServices. One is mandatory, the other two are optional:

  • The DiagramGenerator which the server uses to create the schema of the Diagram
  • Optionally the ModelLayoutEngine, like the ElkLayoutEngine from sprotty-elk, if we want to do server-side layouting
  • Optionally the ServerActionHandlerRegistry for overwriting the default handling of incoming actions

Creating a Custom Model Source Proxy

In case communication between the DiagramServer and client does not work through WebSockets, for example when the DiagramServer is running in a worker or the sprotty client is in a vscode webview (see sprotty-vscode), we can easily implement our own proxy instead.

export class WebWorkerDiagramProxy extends DiagramServerProxy {
    constructor(private worker: Worker) {
        const proxy = this;
        worker.onmessage = function(event) {


    protected sendMessage(message: ActionMessage): void {

Following the example above, first we need to extend DiagramServerProxy. This already gives us most of our needed functionality and makes this proxy a ModelSource. Then we need to listen for incoming messages and pass them to the messageReceived() function, which deserializes and passes them to the ActionDispatcher. Lastly, we need to implement the sendMessage() method to allow actions coming from the ActionDispatcher to be transferred to the DiagramServer.

Now our custom model source proxy is able to propagate all actions between the ActionDispatcher and our DiagramServer running in the worker, which gives us access to all of sprotty’s functionality.