• Serverless Workflow Specification - TypeScript SDK
  • Status
  • SDK Structure
    • Types and Interfaces
    • Classes
    • Fluent Builders
    • Validation Function
    • Other tools
  • Getting Started
    • Installation
    • Usage
      • Create a Workflow Definition from YAML or JSON
      • Create a Workflow Definition by Casting an Object
      • Create a Workflow Definition Using a Class Constructor
      • Create a Workflow Definition Using the Builder API
      • Serialize a Workflow Definition to YAML or JSON
      • Validate Workflow Definitions
      • Generate a directed graph
      • Generate a MermaidJS flowchart
    • Building Locally

Serverless Workflow Specification - TypeScript SDK

This SDK provides a TypeScript API for working with the Serverless Workflow Specification.

With this SDK, you can:

• Parse workflow definitions in JSON and YAML formats
• Programmatically build workflow definitions
• Validate workflow definitions
• Consume the SDK from Node.js runtimes and modern browsers through CommonJS, native ESM, and a standalone UMD bundle

Status

The npm @serverlessworkflow/sdk package versioning aligns with the versioning of the specification:

Latest Releases	Conformance to Spec Version
v1.0.*	v1.0.*
v0.8.4	v0.8

The repository is currently developed and validated on Node.js 22+.

Warning

Previous versions of the SDK were published with a typo in the scope: @severlessworkflow/sdk-typescript instead of @serverlessworkflow/sdk-typescript

Latest Releases	Conformance to Spec Version
v1.0.0	v0.6
v2.0.0	v0.7
v3.0.0	v0.8

SDK Structure

This SDK includes the following key components:

Types and Interfaces

The SDK provides various TypeScript types and interfaces that ensure type safety and enhance the development experience by catching type errors during compile time.

To avoid confusion with classes, these types and interfaces are exported under the Specification object, e.g., Specification.Workflow.

Classes

The SDK includes classes that correspond to the aforementioned types and interfaces. These classes offer:

• Instance Checking: Easily verify if an object is an instance of a specific class.
• Self-Validation: Validate the internal state of an object to ensure it adheres to the expected structure.
• Normalization: Methods to normalize object data, ensuring consistent formatting and values.

To avoid confusion with type definitions, these classes are exported under the Classes object, e.g., Classes.Workflow.

Fluent Builders

The fluent builders wrap the core classes and provide a fluent API for constructing objects. This API allows you to chain method calls and configure objects in a more readable and convenient manner.

The fluent builders are directly exported as *<desired-type>*Builder, e.g., workflowBuilder.

By default, built objects are self-validated and self-normalized. BuildOptions can be passed to the build() method to disable validation or normalization.

Validation Function

The SDK includes a validation function to check if objects conform to the expected schema. This function ensures that your workflow objects are correctly structured and meet the required specifications.

The validate function is directly exported and can be used as validate('Workflow', workflowObject).

Other Tools

The SDK also ships tools to build directed graph and MermaidJS flowcharts from a workflow.

Getting Started

Installation

npm install @serverlessworkflow/sdk

Usage

Create a Workflow Definition from YAML or JSON

You can deserialize a YAML or JSON representation of the workflow using the static method Classes.Workflow.deserialize:

import { Classes } from '@serverlessworkflow/sdk';

// const text = await readFile('/some/path/my-workflow-definition.yaml', { encoding: 'utf8' });
// const text = await fetch('https://myserver.com/my-workflow-definition.json');
const text = `
document:
  dsl: 1.0.3
  name: test
  version: 1.0.0
  namespace: default
do:
- step1:
    set:
      variable: 'my first workflow'
`;
const workflow = Classes.Workflow.deserialize(text);

Create a Workflow Definition by Casting an Object

You can type-cast an object to match the structure of a workflow definition:

import { Classes, Specification, validate } from '@serverlessworkflow/sdk';

// Simply cast an object:
const workflow = {
  document: {
    dsl: '1.0.3',
    name: 'test',
    version: '1.0.0',
    namespace: 'default',
  },
  do: [
    {
      step1: {
        set: {
          variable: 'my first workflow',
        },
      },
    },
  ],
} as Specification.Workflow;

// Validate it
try {
  validate('Workflow', workflow);
  // Serialize it
  const definitionTxt = Classes.Workflow.serialize(workflow);
} catch (ex) {
  // Invalid workflow definition
}

Create a Workflow Definition Using a Class Constructor

You can create a workflow definition by calling a constructor:

import { Classes, validate } from '@serverlessworkflow/sdk';

// Simply use the constructor
const workflow = new Classes.Workflow({
  document: {
    dsl: '1.0.3',
    name: 'test',
    version: '1.0.0',
    namespace: 'default',
  },
  do: [
    /*
    {
      step1: {
        set: {
          variable: 'my first workflow',
        },
      },
    },
  */
  ],
});
workflow.do.push({
  step1: new Classes.SetTask({
    set: {
      variable: 'my first workflow',
    },
  }),
});

// Validate it
try {
  workflow.validate();
  // Serialize it
  const definitionTxt = workflow.serialize();
} catch (ex) {
  // Invalid workflow definition
}

Create a Workflow Definition Using the Builder API

You can use the fluent API to build a validated and normalized workflow definition:

import { documentBuilder, setTaskBuilder, taskListBuilder, workflowBuilder } from '@serverlessworkflow/sdk';

const workflow = workflowBuilder(/*workflowObject*/)
  .document(documentBuilder().dsl('1.0.3').name('test').version('1.0.0').namespace('default').build())
  .do(
    taskListBuilder()
      .push({
        step1: setTaskBuilder()
          .set({
            variable: 'my first workflow',
          })
          .build(),
      })
      .build(),
  )
  .build(/*{
    validate: false,
    normalize: false
  }*/);

Serialize a Workflow Definition to YAML or JSON

You can serialize a workflow definition either by using its serialize method if it's an instance or the static method with the same name:

import { Classes } from '@serverlessworkflow/sdk';

// const workflow = <Your preferred method>;
if (workflow instanceof Classes.Workflow) {
  const yaml = workflow.serialize(/*'yaml' | 'json' */);
} else {
  const json = Classes.Workflow.serialize(workflow, 'json');
}

Note

The default serialization format is YAML.

Validate Workflow Definitions

Validation can be achieved in two ways: via the validate function or the instance validate method:

import { Classes, validate } from '@serverlessworkflow/sdk';

const workflow = /* <Your preferred method> */;
try {
  if (workflow instanceof Classes.Workflow) {
    workflow.validate();
  }
  else {
    validate('Workflow', workflow);
  }
}
catch (ex) {
  // Workflow definition is invalid
}

Generate a directed graph

A directed graph of a workflow can be generated using the buildGraph function, or alternatives:

• Workflow instance .toGraph();
• Static Classes.Workflow.toGraph(workflow)

import { buildGraph } from '@serverlessworkflow/sdk';

const workflow = {
  document: {
    dsl: '1.0.3',
    name: 'using-plain-object',
    version: '1.0.0',
    namespace: 'default',
  },
  do: [
    {
      step1: {
        set: {
          variable: 'my first workflow',
        },
      },
    },
  ],
};
const graph = buildGraph(workflow);
// const workflow = new Classes.Workflow({...}); const graph = workflow.toGraph();
// const graph = Classes.Workflow.toGraph(workflow);
/*{
  id: 'root',
  type: 'root',
  label: undefined,
  parent: null,
  nodes: [...], // length 3 - root entry node, step1 node, root exit node
  edges: [...], // length 2 - entry to step1, step1 to exit
  entryNode: {...}, // root entry node
  exitNode: {...} // root exit node
}*/

If you need a flattened graph, with all nested subgraph nodes and edges hoisted into a single collection, use buildFlatGraph instead. Pass true as the second argument to also strip the entry/exit port nodes:

import { buildFlatGraph } from '@serverlessworkflow/sdk';

const flatGraph = buildFlatGraph(workflow);
// const flatGraphWithoutPorts = buildFlatGraph(workflow, true);

Generate a MermaidJS flowchart

Generating a MermaidJS flowchart can be achieved in two ways: using the convertToMermaidCode, the legacy MermaidDiagram class, or alternatives:

• Workflow instance .toMermaidCode();
• Static Classes.Workflow.toMermaidCode(workflow)

import { convertToMermaidCode, MermaidDiagram } from '@serverlessworkflow/sdk';

const workflow = {
  document: {
    dsl: '1.0.3',
    name: 'using-plain-object',
    version: '1.0.0',
    namespace: 'default',
  },
  do: [
    {
      step1: {
        set: {
          variable: 'my first workflow',
        },
      },
    },
  ],
};
const mermaidCode = convertToMermaidCode(workflow); /* or  */
// const mermaidCode = new MermaidDiagram(workflow).sourceCode();
// const workflow = new Classes.Workflow({...}); const mermaidCode = workflow.toMermaidCode();
// const mermaidCode = Classes.Workflow.toMermaidCode(workflow);
/*
flowchart TD
    root-entry-node(( ))
    root-exit-node((( )))
    /do/0/step1["step1"]
    /do/0/step1 --> root-exit-node
    root-entry-node --> /do/0/step1


classDef hidden width: 1px, height: 0px, opacity: 0;
*/

flowchart TD
    root-entry-node(( ))
    root-exit-node((( )))
    /do/0/step1["step1"]
    /do/0/step1 --> root-exit-node
    root-entry-node --> /do/0/step1


classDef hidden width: 1px, height: 0px, opacity: 0;

Loading

You can refer to the Mermaid browser examples for a live demo. The browser samples under examples/browser include native ESM examples and legacy UMD-global examples under examples/browser/umd.

Building Locally

To build the project and run the full validation pipeline locally, use the following commands:

git clone https://github.com/serverlessworkflow/sdk-typescript.git
cd sdk-typescript
npm install
npm run lint
npm run test
npm run validate:package

You can also build the package and run the browser examples by executing:

npm run serve

This serves the browser examples in examples/browser, including both native ESM samples and UMD-global samples in examples/browser/umd.

Generated files live under src/lib/generated/ and should be refreshed with:

npm run codegen

If you're interested in contributing, the contributing guide and tooling architecture are the best places to start.