APIs with PlanetScale, Prisma and Nitric

What we'll be doing

In this guide we will create an API using serverless services, then use PlanetScale and Prisma for data persistence. We will then deploy the application to the cloud of your choice, AWS, GCP, or Azure.

To keep things light, the API we'll create generates memes by allowing uploads of image templates, then generating new images with overlaid text. Feel free to adapt the steps for any other API you'd like to build.

For image editing, we used a library from NPM called jimp, but you could use anything else you like.

When you're done, you'll have an API which accepts a template image and can generate multiple memes from it.

Prerequisites

To complete this guide, here are things you'll need setup ahead of time:

An account with PlanetScale it's free
Node.js
The Nitric CLI
(optional) Your choice of an AWS, GCP or Azure account

Getting started

Let's start with a new Nitric project and select the TypeScript starter template:

# create the project
nitric new api-guide ts-starter

# navigate to the new project directory
cd api-guide

# install dependencies
npm install

Once you have the project, you can run it locally to check that everything is working correctly:

nitric start

The example app contains a hello world style example service. Once it's running you can test it with an HTTP request:

curl http://localhost:4001/hello/John
Hello John

Since we won't use the example function you can delete the services/hello.ts file.

Since we'll also use Jimp later for image editing, let's install that now too:

npm install jimp

Database and schema setup

Next, you'll need a PlanetScale database. If you have an account already, skip to the next step. Otherwise, you can sign up for a free account.

Create the database

You can create a new database with the PlanetScale CLI or through the web dashboard.

Here is an example using the CLI:

pscale database create planetnitric --region us-east

You can pick a different region for you database if you prefer, see: available regions

Prisma setup

Now you're ready to setup Prisma, let's start by adding it to our project with npm:

npm install --save-dev prisma @prisma/client

Then, we can initialize Prisma and generate our first schema file:

npx prisma init

This gives you a new prisma schema in a folder called prisma and a new .env file containing config we'll use to connect to our PlanetScale database.

Building the schema

Overwrite the contents of prisma.schema with the schema below. We'll use this to initialize our database.

// This is your Prisma schema file,
// learn more about it in the docs: https://pris.ly/d/prisma-schema

generator client {
  provider        = "prisma-client-js"
  binaryTargets   = ["linux-musl"]
  output   = "./client"
}

datasource db {
  provider             = "mysql"
  url                  = env("DATABASE_URL")
  relationMode         = "prisma"
}

model MemeTemplate {
  name          String         @id @unique
  createdAt     DateTime       @default(now())
  textPositions TextPosition[]
  Meme          Meme[]
}

model TextPosition {
  id       String       @id @default(cuid())
  name     String
  memeId   String
  posX     Int
  posY     Int
  width    Int
  height   Int
  template MemeTemplate @relation(fields: [memeId], references: [name], onDelete: Cascade)

  @@index([memeId])
}

model Meme {
  id         String       @id @default(cuid())
  createdAt  DateTime     @default(now())
  templateId String
  template   MemeTemplate @relation(fields: [templateId], references: [name], onDelete: Cascade)

  @@index([templateId])
}

Next, let's generate the prisma client from the schema:

npx prisma generate

Finally, let's make it easy to import an instance of the prisma client by creating the file prisma/index.ts and adding this code:

import { PrismaClient } from './client'

export * from './client'

let prisma: PrismaClient

if (process.env.NITRIC_ENVIRONMENT !== 'build') {
  if (process.env.NODE_ENV === 'production') {
    prisma = new PrismaClient({
      errorFormat: 'minimal',
    })
  } else {
    globalThis['prisma'] =
      globalThis['prisma'] ||
      new PrismaClient({
        errorFormat: 'pretty',
      })
    prisma = globalThis['prisma']
  }
}

export default prisma

Connecting to PlanetScale

Now our schema is ready, let's get Prisma connected to PlanetScale. The easiest method is to use the connect button in the PlanetScale dashboard and select Prisma from the dropdown. This will give you values you can copy into the .env for your project.

create function hook screenshot - step 2

When you're done, the .env file will look something like this:

DATABASE_URL='<Your URL from the above screenshot>'

With the schema available and the connection details configured, you can push the Prisma schema to PlanetScale:

npx prisma db push

Add cloud resources to our application

Apps built with Nitric define their resources in code, you can write this in the root of any .js or .ts file. To keep things organized, we recommend grouping resources together. So, let's start by defining the resources we'll need to support our API in a new resources directory.

First, let's declare an API gateway. Create a new file called apis.ts in a new folder called resources and this code:

import { api } from '@nitric/sdk'

export const memeApi = api('meme')

This creates a new api resource with the name "meme" and exports it as a resource that can be referenced elsewhere in the project.

Next, let's also create some buckets to store our meme image files. Create a new file called buckets.ts under resources and populate it with the following:

import { bucket } from '@nitric/sdk'

export const templates = bucket('templates')
export const memes = bucket('memes')

Again, we're declaring new resources, buckets in this case, and giving them unique names within the app. We export those resources so they can be referenced again without being declared repeatedly.

Create the meme template service

Now that the resources are declared, let's create the first service. This service allows API consumers to register new meme templates by providing a base image for memes and a configurable set of text locations.

In the /services directory create a new file called templates.ts and populate it with the following code:

import Jimp from 'jimp'
import prisma, { MemeTemplate, TextPosition } from '../prisma'
import { memeApi } from '../resources/apis'
import { templates } from '../resources/buckets'

export interface CreateTemplateRequest
  extends Omit<MemeTemplate, 'filepath' | 'createdAt'> {
  source: string
  textPositions: Omit<TextPosition, 'id' | 'memeId'>[]
}

const templateImgs = templates.allow('write')

export const normalizeName = (name: string) => {
  return name.replace(' ', '-').replace(/[^\w-]*/g, '')
}

// POST: /templates - Create new meme templates
memeApi.post('/templates', async ({ req, res }) => {
  const {
    textPositions,
    source,
    name: rawName,
  } = req.json() as CreateTemplateRequest
  const name = normalizeName(rawName)
  const img = await Jimp.read(source)

  try {
    const template = await prisma.memeTemplate.create({
      data: {
        name,
        textPositions: {
          create: textPositions,
        },
      },
    })

    // Limit width to 512px max to save space
    const resizeFactor = 512 / img.getWidth()
    img.resize(img.getWidth() * resizeFactor, img.getHeight() * resizeFactor)

    // store the image in the bucket
    const buf = await img.getBufferAsync(img.getMIME())
    await templateImgs.file(name).write(buf)

    res.json(template)
  } catch (e) {
    res.status = 409
    res.body = `Name already taken: ${name}: ${e.message}`
  }
})

// GET: /templates - List all meme templates
memeApi.get('/templates', async ({ res }) => {
  const memeTemplates = await prisma.memeTemplate.findMany({
    include: {
      textPositions: true,
    },
  })

  res.json(memeTemplates)
})

In this example we're importing the API gateway memeApi we created in our resources directory, and registering route and method handlers using methods like get and post, much like you would in frameworks such as Express.

Additionally, we're importing the bucket used to store template images templateImages from the resources directory. We also declare our intended use of the bucket with the for method, which lets Nitric know what permissions your code needs and applies them during deployments. In this instance we're only giving our template service write access to the templates bucket.

The incoming context object (which has been destructured into req and res) contains request and response details like path params, query params, headers, body, status, etc.

Create the meme service

Similar to the templates example, we'll create another new file services/memes.ts, with the code below:

import Jimp from 'jimp'
import prisma, { Meme } from '../prisma'
import { memes, templates } from '../resources/buckets'
import { memeApi } from '../resources/apis'

interface MemeCreationRequest extends Omit<Meme, 'id' | 'templateId'> {
  templateName: string
  texts: {
    name: string
    value: string
  }[]
}

const templateImgs = templates.allow('read')
const memesImgs = memes.allow('read', 'write')

// POST: /memes - Create new meme images
memeApi.post('/memes', async ({ req, res }) => {
  const meme = req.json() as MemeCreationRequest

  const template = await prisma.memeTemplate.findFirst({
    include: {
      textPositions: true,
    },
    where: {
      name: {
        equals: meme.templateName,
      },
    },
  })

  const imgBytes = await templateImgs.file(template.name).read()

  // Load the image and font
  const [img, font] = await Promise.all([
    Jimp.read(Buffer.from(imgBytes)),
    Jimp.loadFont(Jimp.FONT_SANS_32_WHITE),
  ])

  // Apply text to the template image to create the meme
  meme.texts.forEach((text) => {
    // get the text template
    const matchingText = template.textPositions.find(
      (tp) => tp.name === text.name,
    )

    if (!matchingText) return

    // ignore if anchor tags don't match
    img.print(
      font,
      img.getWidth() * (matchingText.posX / 100),
      img.getHeight() * (matchingText.posY / 100),
      {
        text: text.value,
        alignmentX: Jimp.HORIZONTAL_ALIGN_CENTER,
        alignmentY: Jimp.VERTICAL_ALIGN_MIDDLE,
      },
      img.getWidth() * (matchingText.width / 100),
      img.getHeight() * (matchingText.height / 100),
    )
  })

  const [newMeme, buf] = await Promise.all([
    prisma.meme.create({
      data: {
        templateId: meme.templateName,
      },
    }),
    img.getBufferAsync(Jimp.MIME_PNG),
  ])

  await memesImgs.file(newMeme.id).write(buf)

  res.json(newMeme)
})

// GET: /memes - List all created memes
memeApi.get('/memes', async ({ res }) => {
  const memes = await prisma.meme.findMany()
  return res.json(memes)
})

// GET: /memes/:id - Get a meme image by it's ID
memeApi.get('/memes/:id', async ({ req, res }) => {
  const { id } = req.params
  const signedUrl = await memesImgs.file(id).getDownloadUrl()
  res.status = 303
  res.headers['Location'] = [signedUrl]
})

Again, this file re-uses the memeApi resource, but defines new routes under the /memes path.

We also request read access to the templateImages bucket and read-write access to the memeImages bucket.

Testing it locally

Now we've got the API established, let's test it out locally.

npm run dev

When running APIs locally, Nitric will sub-route them by their name. So in this example to create a new meme template you'll need to make your POST request to https://localhost:4001/templates.

Here are some example requests you can use to test the API:

Create a meme template

curl -X POST http://localhost:4001/templates \
   -H 'Content-Type: application/json' \
   -d '{"name":"my-meme","source":"https://www.meme-arsenal.com/memes/89f28a7e83e28f15b1d8e560c788b4fc.jpg","textPositions":[{"name":"topText","posX":50,"posY":0,"width":50,"height":50},{"name":"bottomText","posX":50,"posY":50,"width":50,"height":50}]}'

Full Request Body

{
  "name": "my-meme",
  "source": "https://www.meme-arsenal.com/memes/89f28a7e83e28f15b1d8e560c788b4fc.jpg",
  "textPositions": [
    {
      "name": "topText",
      "posX": 50,
      "posY": 0,
      "width": 50,
      "height": 50
    },
    {
      "name": "bottomText",
      "posX": 50,
      "posY": 50,
      "width": 50,
      "height": 50
    }
  ]
}

For source provide a URL hosting a meme template image in a common format like .png or .jpg

Create a new meme using the template

curl -X POST http://localhost:4001/memes \
   -H 'Content-Type: application/json' \
   -d '{"templateName":"my-meme","texts":[{"name":"topText","value":"top text content"},{"name":"bottomText","value":"bottom text content"}]}'

Full Request Body

{
  "templateName": "my-meme",
  "texts": [
    {
      "name": "topText",
      "value": "top text content"
    },
    {
      "name": "bottomText",
      "value": "bottom text content"
    }
  ]
}

Retrieve the image

Using the Meme ID returned from the previous request, open a browser and navigate to http://localhost:4001/memes/[id].

Deploy to the cloud

If you're ready, you can deploy this project to AWS, Azure or Google Cloud.

Start by defining a stack. Stacks are essentially named deployment targets, which represent instances of your application running in the cloud.

You can create a new stack by running nitric stack new and following the prompts.

nitric stack new

Finally, run the up command to deploy the stack and push your code to the cloud:

nitric up

You can use the URL returned from the up command to make requests to your newly deployed API. Then, when you're done, you can destroy the stack with the down command:

nitric down