TypeScript types

Without types, from("tasks") returns loosely typed rows. With types, the compiler knows each table’s columns, and a typo in eq("donee", false) is an error in your editor instead of a bug in production.

How it works

Each resource compiles to an OpenAPI description: a read model, a create body and an update body per table. generateTypes turns that into a TypeScript Database interface you pass to createClient. The create body tells which columns are required, which the generator uses to separate “always present” from “may be null”.

Generate

From a script
Per resource

import { generateTypes } from "@pawabase/client";
import { writeFileSync } from "node:fs";

// The OpenAPI description of your project's API
// (the same document the API Explorer is built from)
const spec = await (await fetch(`${process.env.PAWABASE_URL}/rest/v1/openapi.json`, {
  headers: { apikey: process.env.PAWABASE_SECRET_KEY! },
})).json();

writeFileSync("src/database.types.ts", generateTypes(spec, { name: "Database", source: "development" }));

Re-run it whenever you change columns: add it to your build or a npm run types script, and commit the file.

Use

import { createClient } from "@pawabase/client";
import type { Database } from "./database.types";

export const pawabase = createClient<Database>({ url, apiKey });

const { data } = await pawabase.from("tasks").eq("done", false).order("created_at", "desc");
data[0].title;        // string
data[0].assignee_id;  // string | null

await pawabase.from("tasks").create({ title: "Write docs" });          // `title` required
// await pawabase.from("tasks").create({ titel: "x" });                // compile error
// await pawabase.from("tasks").eq("donee", false);                    // compile error

The generated shape for each table is:

tasks: { Row: Task; Insert: TaskInsert; Update: TaskUpdate }

so reads, creates and updates each get the right type: Insert requires the required columns, Update makes everything optional, and Row has everything the server returns.

Without generation

You can write the shape by hand for a few tables:

type Task = { id: string; title: string; done: boolean; created_at: string };

const pawabase = createClient<{ resources: { tasks: { Row: Task; Insert: Partial<Task>; Update: Partial<Task> } } }>({ url, apiKey });

Python

Python has no generated types, but the client is typed and rows are plain dictionaries. Use a TypedDict or a dataclass of your own at the edge:

from typing import TypedDict

class Task(TypedDict):
    id: str
    title: str
    done: bool

tasks: list[Task] = pb.from_("tasks").eq("done", False).execute().data

Keep types honest

Types are generated from the definition, so they reflect your columns, not the data’s mess. A column marked nullable is typed | null.
Generate types from the same environment your app reads from, so they match what it gets back.
Treat the generated file as read-only; change columns in Studio or code and regenerate.

Columns and types

What the types come from.

API Explorer

See the compiled description.

Getting started

Define your data

Change safely

Read and write

Protect and shape

Studio tools

Build well

How it works

Generate

Use

Without generation

Python

Keep types honest

Columns and types

API Explorer

​How it works

​Generate

​Use

​Without generation

​Python

​Keep types honest

Columns and types

API Explorer

How it works

Generate

Use

Without generation

Python

Keep types honest