Using Prisma with Turborepo

Prisma (opens in a new tab) is an extremely popular ORM with automated migrations, type safety and integrated tooling. Using it with Turborepo can cut time you spend generating code, and easily make sure your generated Prisma code is always up-to-date.

Guide

This guide shows you how to:

1. Set up Prisma in a monorepo
2. Handle migration and code generation scripts
3. Cache those scripts with Turborepo
4. Ensure that they're always run whenever dev or build is run

If you've already got Prisma set up in your database, you can skip to step 4.

1. Create your monorepo

If you don't have an existing project, use our quickstart to create a new monorepo.

2. Add a new database package

Create a new folder called database inside packages with a package.json inside:

{
  "name": "database",
  "version": "0.0.0",
  "dependencies": {
    "@prisma/client": "latest"
  },
  "devDependencies": {
    // Replace "latest" with the latest version
    "prisma": "latest"
  }
}

If you're using pnpm, you should add a file at the root called .npmrc:

public-hoist-pattern[]=*prisma*

Run your package manager's install step to install the new dependencies.

3. Run prisma init

cd into packages/database:

cd packages/database

Run npx prisma init.

This should create several files inside packages/database:

prisma/schema.prisma
.gitignore
.env

• schema.prisma is where your Prisma schema (opens in a new tab) lives. Here, you'll be able to modify the shape of your database.
• .gitignore adds some ignored files to git
• .env lets you manually specify your DATABASE_URL for prisma.

At this point, you should refer to the Prisma docs for connecting your database to Prisma (opens in a new tab).

Once you've got a database connected, you can move on.

4. Setting up the scripts

Let's add some scripts to the package.json inside packages/database:

{
  "scripts": {
    "db:generate": "prisma generate",
    "db:push": "prisma db push --skip-generate"
  }
}

Let's also add these scripts to turbo.json in the root:

{
  "pipeline": {
    "db:generate": {
      "cache": false
    },
    "db:push": {
      "cache": false
    }
  }
}

We can now run turbo db:push db:generate from the root of our repository to automatically migrate our database and generate our type safe Prisma client.

5. Exporting your client

Next, we need to export the @prisma/client so we can use it in our applications. Let's add a new file to packages/database:

export * from '@prisma/client';

Following the internal packages pattern, we'll also need to add index.ts to main and types inside packages/database/package.json.

{
  "main": "./index.ts",
  "types": "./index.ts"
}

Importing database

Let's now import our database package into one of our apps to test it out. Let's say you have an app at apps/web. Add the dependency to apps/web/package.json:

{
  "dependencies": {
    "database": "*"
  }
}

Run your package manager's install command.

You can now import PrismaClient from database anywhere in your app:

import { PrismaClient } from 'database'
 
const client = new PrismaClient();

6. Figuring out the scripts

We're now in a pretty good position. We have a reusable database module that we can import into any of our applications. We've got a turbo db:push script we can use to push our changes to the database.

However, our db:generate scripts aren't optimized yet. They provide crucial code to our dev and build tasks. If a new developer runs dev on an application without running db:generate first, they'll get errors.

So, let's make sure that db:generate is always run before the user runs dev:

{
  "pipeline": {
    "dev": {
      "dependsOn": ["^db:generate"],
      "cache": false
    },
    "build": {
      "dependsOn": ["^db:generate"],
      "outputs": ["your-outputs-here"]
    },
    "db:generate": {
      "cache": false
    }
  }
}

Check out the section on running tasks to learn more about the ^db:generate syntax.

7. Caching the results of prisma generate

prisma generate outputs files to the filesystem, usually inside node_modules. In theory, it should be possible to cache the output of prisma generate with Turborepo to save a few seconds.

However, Prisma behaves differently with different package managers. This can lead to unpredictable results, which might lead to broken deployments in some situations. Instead of documenting the intricacies of each approach, we recommend not caching the results of prisma generate. Since prisma generate usually only takes 5-6 seconds, and tends not to take longer with larger schema files, this seems like a fine trade-off.

You may wish to experiment with this yourself. If you find a solution that you feel works, feel free to add an issue (opens in a new tab) and we can update this section.