Welcome to ZudoJS

What ZudoJS is

ZudoJS is a backend framework for TypeScript on Node.js. It takes over the jobs you listed in What a framework does: lifecycle, dependency injection, routing, configuration, validation, errors, data access, security, testing and project structure.

It is not one big package. It is 38 small npm packages, each named @zudojs/something, plus a command-line tool published as zudojs. Each package owns one job. You install only the ones you use, and you can use most of them on their own, even in a project that is not a ZudoJS app.

That design follows the ideas from Backend architecture: separation of concerns, interfaces between parts, dependency injection, and dependencies that point in one direction only.

Packages on shelves

Think of the packages as sitting on shelves. A package may use packages on lower shelves, never higher ones. So there are no cycles, and each package can be built, tested and understood on its own.

  top      testing                   fakes and helpers for all the others
           http, cli                 talk to the outside world
           runtime, auth, cqrs, …    put the parts together into an application
           core, events, schema, …   framework machinery
           config, logger, …         single-purpose building blocks
  bottom   errors, types             depend on no other ZudoJS package

• At the bottom is @zudojs/errors. Almost every other package uses it, so an error from any package is a BaseError with a status code and a stable code. @zudojs/types sits beside it with shared type guards.
• In the middle are building blocks such as @zudojs/config, @zudojs/logger, @zudojs/container and @zudojs/schema. For example, @zudojs/schema uses only errors, constants and types.
• Higher up are the packages that combine others: @zudojs/runtime starts and stops an application, and @zudojs/http serves it on the network.

The Architecture and Dependency direction pages list every package's exact dependencies.

The packages, grouped by job

You do not need to remember this table. Come back to it when a lesson mentions a package. Each name links to its documentation.

A normal application starts with about a dozen of them. The rest arrive when you need them: queue when you have background jobs, tenancy when you serve several customers from one app, and so on. The package list has a search box.

How the packages fit together

Here is the path one request takes through a ZudoJS application, with the package that handles each step. Compare it with the BookStore, where you wrote every step yourself:

The packages that do not need Node.js also run in this page's browser terminal. Here are three of them working together: @zudojs/schema checks the input, @zudojs/errors reports failures, and @zudojs/container hands out the one shared service:

import { ContainerScope, createContainer, createToken } from "@zudojs/container";
import { BaseError, NotFoundError } from "@zudojs/errors";
import { schema } from "@zudojs/schema";
import type { Infer } from "@zudojs/schema";

const NewBookSchema = schema.object({
  title: schema.string().trim().min(1).max(200),
  priceCents: schema.number().int().min(0),
});
type Book = Infer<typeof NewBookSchema> & { readonly id: number };

class BookService {
  private readonly books: Book[] = [];

  add(input: unknown): Book {
    const book = { id: this.books.length + 1, ...NewBookSchema.parse(input) };
    this.books.push(book);
    return book;
  }

  get(id: number): Book {
    const book = this.books.find((b) => b.id === id);
    if (book === undefined) throw new NotFoundError(`Book ${id} not found`);
    return book;
  }
}

const BOOKS = createToken<BookService>("books");
const container = createContainer();
container.register(BOOKS, { useFactory: () => new BookService() }, { scope: ContainerScope.SINGLETON });

const books = container.resolve(BOOKS);
console.log(books.add({ title: "  Kindred ", priceCents: 1099 }));
console.log("same service everywhere:", container.resolve(BOOKS) === books);

for (const attempt of [() => books.add({ title: "", priceCents: -1 }), () => books.get(9)]) {
  try {
    attempt();
  } catch (error) {
    if (error instanceof BaseError) console.log(error.name, error.statusCode, error.code, error.message);
  }
}

{ id: 1, title: 'Kindred', priceCents: 1099 }
same service everywhere: true
SchemaError 400 ERR_SCHEMA_VALIDATION Validation failed
NotFoundError 404 ERR_RESOURCE_NOT_FOUND Book 9 not found

Press Run in browser and change things. What to notice:

• One description, two uses. NewBookSchema checks the data at runtime, and Infer turns it into the Book type. In the BookStore you wrote the interface and the parser separately.
• The container creates BookService the first time someone asks for the BOOKS token, and gives everyone that same instance (SINGLETON). No createRouter threading objects through by hand.
• One error family. The SchemaError thrown by @zudojs/schema and the NotFoundError you threw both extend BaseError from @zudojs/errors, so one instanceof check reads the status code of either.

The next lesson installs @zudojs/schema and @zudojs/errors on your computer and explains them in detail. The container gets its own lesson in the ZudoJS core part.

The command-line tool

Choosing a dozen packages and wiring them into a project by hand would bring back the problem you just left behind. So ZudoJS has a CLI (command-line interface), published on npm as zudojs. It gives you a command called zudojs, and a short name for it, zudo. Run on its own, it shows a menu of these commands:

You will install it and use every one of these in Create the Task API project.

Three shapes of application

zudojs create asks which architecture you want. There are three, and they differ in how many deployable programs you end up with:

A common and healthy path is monolith → modular monolith → microservices only where needed. The same ZudoJS packages work in all three, and the Architecture with ZudoJS part of the course walks that path.

Practice

import { NotFoundError } from "@zudojs/errors";

interface Book {
  readonly id: number;
  readonly title: string;
}

const books: Book[] = [{ id: 1, title: "Kindred" }];

function findByTitle(title: string): Book {
  const book = books.find((b) => b.title === title);
  if (book === undefined) throw new NotFoundError(`No book called "${title}"`);
  return book;
}

console.log(findByTitle("Kindred"));
try {
  findByTitle("Dune");
} catch (error) {
  if (error instanceof NotFoundError) console.log(error.statusCode, error.message);
}

{ id: 1, title: 'Kindred' }
404 No book called "Dune"

Recap

• ZudoJS is 38 small @zudojs/* packages plus the zudojs CLI. Each owns one job, and you install only what you use.
• Packages sit on shelves and only use packages below them. @zudojs/errors is at the bottom, so every error shares one base class.
• A request passes through http, security, auth, schema, container, database and errors, and runtime starts and stops it all.
• The zudojs CLI (short name zudo) creates, runs, builds, extends and checks projects.
• Applications can be a monolith, a modular monolith or microservices. Start with a monolith.

Next, you install your first two ZudoJS packages and use them in a Task API service.