Skip to content
MountSQLiMountSQLI

Introduction

MountSQLI is a type-safe ORM and backend platform. It gives you a typed query builder, schema definitions, migrations, auth, an API layer, storage, realtime, caching, and an AI assistant — all from one compiler.

What is it?

Section titled “What is it?”

Most ORMs are a class hierarchy over your database. You define models as classes, and the library generates SQL at runtime. MountSQLI takes a different bet:

The ORM is a compiler plus an intermediate representation (QueryPlan IR), not a class hierarchy.

You describe what you want in TypeScript. MountSQLI compiles that into a QueryPlan (a plain data structure), then a Dialect turns the plan into SQL with bound parameters. The database never sees string-concatenated values.

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


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


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


const row = await db.from(users).where("email", "=", "a@b.c").findOne();
//      ^? { id: number; email: string } | null

Why does it exist?

Section titled “Why does it exist?”

Other ORMs force a trade-off:

  • Prisma / TypeORM are safe but heavy. They allocate object graphs per row and can’t tree-shake.
  • Drizzle is lighter and SQL-first, but still builds queries as chained method state.

MountSQLI keeps Drizzle’s “native SQL first” feel while making the query a compiled, immutable data structure. That unlocks edge deployments, near-zero allocation, and a single validator that both you and an AI use.

When should you use it?

Section titled “When should you use it?”

Use MountSQLI when you want:

  • End-to-end type safety from schema to result row.
  • One tool for schema, queries, migrations, auth, API, and storage.
  • SQL you can read, with injection safety by construction.
  • AI-generated queries validated by the same compiler as yours.

How it works (the 30-second version)

Section titled “How it works (the 30-second version)”
graph TD
  A[TypeScript schema + builder] --> B[QueryPlan IR]
  B --> C[Dialect compiles to SQL + params]
  C --> D[Driver runs query]
  D --> E[Typed rows]

The plan is the contract. Everything else — dialects, drivers, RLS, the AI validator — reads or writes the same plan.

Best practices

Section titled “Best practices”
  • Define tables once with defineTable and infer types with InferTable.
  • Let the compiler catch bad column names and operators at build time.
  • Prefer the builder over raw SQL; escape hatches exist but are guarded.

Common mistakes

Section titled “Common mistakes”
  • Treating MountSQLI like an ActiveRecord ORM (it has no model instances).
  • Hand-writing SQL strings instead of using the builder’s raw-SQL helpers.
Section titled “Related”