Defining Schemas with Zod & Inferring Types

With the project running, model the actual data the API will accept and return, starting from Zod schemas instead of hand-written interfaces.

import { z } from 'zod';

const TaskSchema = z.object({
  id: z.string().uuid(),
  title: z.string().min(1).max(200),
  done: z.boolean().default(false),
  priority: z.enum(['low', 'medium', 'high']),
  tags: z.array(z.string()).optional(),
});

type Task = z.infer<typeof TaskSchema>;

Each field reads like a description of the constraint, z.string().min(1).max(200) isn't just "a string," it's "a string between 1 and 200 characters," checked for real when .parse() runs, not just documented in a comment.

Deriving request schemas from the base schema

const CreateTaskSchema = TaskSchema.omit({ id: true, done: true });
// clients shouldn't set id or done themselves when creating a task

const UpdateTaskSchema = TaskSchema.partial().omit({ id: true });
// every field optional (a client only sends what's changing), id still can't be changed

type CreateTaskInput = z.infer<typeof CreateTaskSchema>;
type UpdateTaskInput = z.infer<typeof UpdateTaskSchema>;

.omit() and .partial() on a Zod schema work exactly like Omit<T, K> and Partial<T> from the intermediate TypeScript course, except they reshape the runtime validator, not just the compile-time type, and z.infer immediately gives you the matching TypeScript type for free.

Custom validation with refine

const CreateTaskSchemaStrict = CreateTaskSchema.refine(
  (data) => data.title.trim().length > 0,
  { message: 'Title cannot be only whitespace', path: ['title'] }
);

.refine() adds a check .min()/.max()/.enum() can't express on their own, an arbitrary function returning true/false, with a custom error message attached to a specific field.