Quick Start

This guide gets you from zero to a running query. It uses SQLite in memory, so no database setup is required.

1. Install

pnpm add @mountsqli/core @mountsqli/driver-sqlite

2. Define a table

import { defineTable, int, text } from "@mountsqli/core";

export const users = defineTable("users", {
  id: int().pk(),
  email: text().notNull(),
  age: int(),
});

export type User = InferTable<typeof users>;

3. Open the database

import { mountsqli } from "@mountsqli/core";

export const db = mountsqli({
  driver: "sqlite",
  url: ":memory:",
  tables: [users],
});

4. Run a query

import { db, users } from "./db";

await db.from(users).insert({ email: "ada@example.com", age: 36 });

const adult = await db
  .from(users)
  .where("age", ">", 18)
  .orderBy("email", "asc")
  .findOne();

console.log(adult);
// → { id: 1, email: "ada@example.com", age: 36 }

That is it. The result is fully typed — adult is User | null.

What just happened?

defineTable built a TableDef with phantom-typed columns.
mountsqli created a Db with a driver and your tables.
The builder produced a QueryPlan, the dialect compiled it to SELECT ... WHERE "age" > ?, and the driver returned typed rows.

Real-world shape

In a real app you keep db.ts as the single source of truth and import db wherever you query:

import { db, users } from "../db";

export async function listAdults() {
  return db.from(users).where("age", ">=", 18).all();
}

Best practices

Put defineTable + mountsqli in one db.ts module.
Export InferTable<typeof x> types alongside each table.
Use :memory: for scripts and tests; use a file URL to persist.

Common mistakes

Awaiting mountsqli() — it is synchronous and returns a Db you can use now.
Using any for rows — let InferTable do the work.

Your First Project — a fuller scaffold with config.
Query Builder → Select — every read option.
Schema → Defining Tables — column builders in depth.