📘️ openapi-typescript

🚀 Convert static OpenAPI schemas to TypeScript types quickly using pure Node.js. Fast, lightweight, (almost) dependency-free, and no Java/node-gyp/running OpenAPI servers necessary.

Features

• ✅ Supports YAML and JSON formats
• ✅ Supports advanced OpenAPI 3.1 features like discriminators
• ✅ Supports loading via remote URL (simple authentication supported with the --auth flag)
• ✅ Supports remote references: $ref: "external.yaml#components/schemas/User"
• ✅ Fetches remote schemas quickly using undici

Examples

👀 See examples

Usage

Note:️ openapiTS requires VALID OpenAPI 3.x schemas to work, and this library does not handle validation. There are several quality tools that handle this like @apidevtools/swagger-parser. Make sure to validate your schemas first!

🖥️ CLI

🗄️ Reading a local schema

npx openapi-typescript schema.yaml --output schema.ts

# 🔭 Loading spec from schema.yaml…
# 🚀 schema.yaml -> schema.ts [250ms]

🦠 Globbing local schemas

npx openapi-typescript "specs/**/*.yaml" --output schemas/

# 🔭 Loading spec from specs/one.yaml…
# 🔭 Loading spec from specs/two.yaml…
# 🔭 Loading spec from specs/three.yaml…
# 🚀 specs/one.yaml -> schemas/specs/one.ts [250ms]
# 🚀 specs/two.yaml -> schemas/specs/two.ts [250ms]
# 🚀 specs/three.yaml -> schemas/specs/three.ts [250ms]

Thanks, @sharmarajdaksh!

☁️ Reading remote schemas

npx openapi-typescript https://petstore3.swagger.io/api/v3/openapi.yaml --output petstore.d.ts

# 🔭 Loading spec from https://petstore3.swagger.io/api/v3/openapi.yaml…
# 🚀 https://petstore3.swagger.io/api/v3/openapi.yaml -> petstore.d.ts [650ms]

Thanks, @psmyrdek!

🟦 Using in TypeScript

Import any top-level item from the generated spec to use it. It works best if you also alias types to save on typing:

import { components } from "./generated-schema.ts";

type APIResponse = components["schemas"]["APIResponse"];

Because OpenAPI schemas may have invalid TypeScript characters as names, the square brackets are a safe way to access every property.

🏗️ Operations

Operations can be imported directly by their operationId:

import { operations } from "./generated-schema.ts";

type getUsersById = operations["getUsersById"];

Thanks, @gr2m!

⚾ Fetching data

Fetching data can be done simply and safely using an automatically-typed fetch wrapper:

• openapi-fetch (recommended)
• openapi-typescript-fetch by @ajaishankar

Example (openapi-fetch)

import createClient from "openapi-fetch";
import { paths } from "./v1"; // (generated from openapi-typescript)

const { get, post, put, patch, del } = createClient<paths>({
  baseUrl: "https://myserver.com/api/v1/",
  headers: {
    Authorization: `Bearer ${import.meta.env.VITE_AUTH_TOKEN}`,
  },
});

See each project’s respective pages for usage & install instructions.

✨ Tip: Automatically-typed fetch wrappers are better to use than manually-assigned generics. The latter is not only more work, but it can be error-prone (which makes your OpenAPI typing worthless if it can’t catch all your errors!).

📖 Options

The following flags can be appended to the CLI command.

Option	Alias	Default	Description
--help			Display inline help message and exit
--version			Display this library’s version and exit
--output [location]	-o	(stdout)	Where should the output file be saved?
--auth [token]			Provide an auth token to be passed along in the request (only if accessing a private schema)
--header	-x		Provide an array of or singular headers as an alternative to a JSON object. Each header must follow the key: value pattern
--headers-object="{…}"	-h		Provide a JSON object as string of HTTP headers for remote schema request. This will take priority over --header
--http-method	-m	GET	Provide the HTTP Verb/Method for fetching a schema from a remote URL
--immutable-types		false	Generates immutable types (readonly properties and readonly array)
--additional-properties		false	Allow arbitrary properties for all schema objects without additionalProperties: false
--empty-objects-unknown		false	Allow arbitrary properties for schema objects with no specified properties, and no specified additionalProperties
--default-non-nullable		false	Treat schema objects with default values as non-nullable
--export-type	-t	false	Export type instead of interface
--path-params-as-types		false	Allow dynamic string lookups on the paths object
--support-array-length		false	Generate tuples using array minItems / maxItems
--alphabetize		false	Sort types alphabetically

🚩 --path-params-as-types

By default, your URLs are preserved exactly as-written in your schema:

export interface paths {
  "/user/{user_id}": components["schemas"]["User"];
}

Which means your type lookups also have to match the exact URL:

import { paths } from "./my-schema";

const url = `/user/${id}`;
type UserResponses = paths["/user/{user_id}"]["responses"];

But when --path-params-as-types is enabled, you can take advantage of dynamic lookups like so:

import { paths } from "./my-schema";

const url = `/user/${id}`;
type UserResponses = paths[url]["responses"]; // automatically matches `paths['/user/{user_id}']`

Though this is a contrived example, you could use this feature to automatically infer typing based on the URL in a fetch client or in some other useful place in your application.

Thanks, @Powell-v2!

🚩 --support-array-length

This option is useful for generating tuples if an array type specifies minItems or maxItems.

For example, given the following schema:

components:
  schemas:
    TupleType
      type: array
      items:
        type: string
      minItems: 1
      maxItems: 2

Enabling --support-array-length would change the typing like so:

  export interface components {
    schemas: {
-     TupleType: string[];
+     TupleType: [string] | [string, string];
    };
  }

This results in more explicit typechecking of array lengths.

Note: this has a reasonable limit, so for example maxItems: 100 would simply flatten back down to string[];

Thanks, @kgtkr!

🐢 Node

npm i --save-dev openapi-typescript

import fs from "node:fs";
import openapiTS from "openapi-typescript";

// example 1: load [object] as schema (JSON only)
const schema = await fs.promises.readFile("spec.json", "utf8"); // must be OpenAPI JSON
const output = await openapiTS(JSON.parse(schema));

// example 2: load [string] as local file (YAML or JSON; released in v4.0)
const localPath = new URL("./spec.yaml", import.meta.url); // may be YAML or JSON format
const output = await openapiTS(localPath);

// example 3: load [string] as remote URL (YAML or JSON; released in v4.0)
const output = await openapiTS("https://myurl.com/v1/openapi.yaml");

The Node API may be useful if dealing with dynamically-created schemas, or you’re using within context of a larger application. Pass in either a JSON-friendly object to load a schema from memory, or a string to load a schema from a local file or remote URL (it will load the file quickly using built-in Node methods). Note that a YAML string isn’t supported in the Node.js API; either use the CLI or convert to JSON using js-yaml first.

📖 Node options

The Node API supports all the CLI flags above in camelCase format, plus the following additional options:

Name	Type	Default	Description
commentHeader	string		Override the default “This file was auto-generated …” file heading
inject	string		Inject arbitrary TypeScript types into the start of the file
transform	Function		Override the default Schema Object ➝ TypeScript transformer in certain scenarios
postTransform	Function		Same as transform but runs after the TypeScript transformation

🤖 transform / postTransform

If using the Node.js API, you can override the normal Schema Object transformer with your own. This is useful for providing enhanced functionality for specific parts of your schema.

For example, say your schema has the following property:

properties:
  updated_at:
    type: string
    format: date-time

By default, openapiTS will generate updated_at?: string; because it’s not sure which format you want by "date-time" (formats are nonstandard and can be whatever you’d like). But we can enhance this by providing our own custom formatter, like so:

const types = openapiTS(mySchema, {
  transform(schemaObject, metadata): string {
    if ("format" in schemaObject && schemaObject.format === "date-time") {
      return schemaObject.nullable ? "Date | null" : "Date";
    }
  },
});

That would result in the following change:

-  updated_at?: string;
+  updated_at?: Date;

Any Schema Object present in your schema will be run through this formatter (even remote ones!). Also be sure to check the metadata parameter for additional context that may be helpful.

There are many other uses for this besides checking format. Because this must return a string you can produce any arbitrary TypeScript code you’d like (even your own custom types).

✨ Don’t forget about postTransform() as well! It works the same way, but runs after the TypeScript transformation so you can extend/modify types as-needed.

🏅 Project Goals

1. Support converting any valid OpenAPI schema to TypeScript types, no matter how complicated.
2. This library does NOT validate your schema, there are other libraries for that.
3. The generated TypeScript types must match your schema as closely as possible (e.g. no renaming to PascalCase)
4. This library should never require Java, node-gyp, or some other complex environment to work. This should require Node.js and nothing else.
5. This library will never require a running OpenAPI server to work.

🤝 Contributing

PRs are welcome! Please see our CONTRIBUTING.md guide.

Contributors ✨

Thanks goes to these wonderful people (emoji key):

Drew Powers 💻 📖 🚇 ⚠️	Przemek Smyrdek 💻 📖 🤔 ⚠️	Dan Enman 🐛 💻	Atle Frenvik Sveen 💻 📖 🤔 ⚠️	Tim de Wolf 💻 🤔	Tom Barton 💻 📖 🤔 ⚠️	Sven Nicolai Viig 🐛 💻 ⚠️
Sorin Davidoi 🐛 💻 ⚠️	Nathan Schneirov 💻 📖 🤔 ⚠️	Lucien Bénié 💻 📖 🤔 ⚠️	Boris K 📖	Anton 🐛 💻 🤔 ⚠️	Tim Shelburne 💻 ⚠️	Michał Miszczyszyn 💻
Sam K Hall 💻 ⚠️	Matt Jeanes 💻	Kristofer Giltvedt Selbekk 💻	Elliana May 💻 ⚠️	Henrik Hall 💻 📖 ⚠️	Gregor Martynus 💻 ⚠️ 🐛	Sam Mesterton-Gibbons 💻 🐛 ⚠️
Rendall 💻 🐛 ⚠️	Robert Massaioli 💻 🐛	Jan Kuča 💻 ⚠️	Thomas Valadez 📖	Asitha de Silva 💻 🐛	Mikhail Yermolayev 🐛	Alex Batalov 💻 ⚠️
Federico Bevione 🐛 💻 ⚠️	Daisuke Yamamoto 💻 🐛 ⚠️	dnalborczyk 📖 💻 ⚠️	FabioWanner 🐛 💻 ⚠️	Ash Smith 💻 🐛 ⚠️	Micah Halter 💻 ⚠️ 🐛	Yuto Yoshihara 💻 🐛 ⚠️
Dakshraj Sharma 💻	Shaosu Liu 💻	Vytenis 💻	Eric Zorn 💻 ⚠️ 📖	Max Belsky 💻 🐛	Peter Bech 💻 🐛	Rusty Conover 💻
Dave Carlson 💻	ottomated 💻 🐛	Artem Shuvaev 💻 🐛	ajaishankar 📖	Dominik Dosoudil 💻 ⚠️	tkr 💻 📖	berzi 💻 📖 ⚠️
Philip Trauner 💻 📖 ⚠️	Pavel Yermolin 💻 📖 ⚠️	Duncan Beevers 💻 🐛 ⚠️ 📖	Timofey Kukushkin 💻 ⚠️ 🐛	Dmitry Semigradsky 🐛 ⚠️ 💻	Jeremy Liberman 💻 ⚠️	Axel Hernández Ferrera 💻 🐛 ⚠️
Loïc Fürhoff 💻 ⚠️ 🐛	Bartosz Szczeciński 💻 🐛 ⚠️	Marco Salomone 💻 ⚠️ 🐛	Yacine Hmito 💻 ⚠️ 🐛	Sajad Torkamani 📖	Marius van den Beek 💻 🐛 ⚠️	Steven Grimm 💻 🐛 ⚠️
Erik Hughes 💻 🐛 ⚠️	Matthieu Monsch 💻 🐛 ⚠️	Mitchell Merry 💻 ⚠️ 🐛	François Risoud 💻 🐛 ⚠️	shoffmeister 📖

This project follows the all-contributors specification. Contributions of any kind welcome!