Create a serverless API with TypeScript, GraphQL and MongoDB

Updated on 4th December, 2020 to update ZEIT’s change to Vercel.

Updated on 30th July, 2020 to show that this can be used in Next.js API Routes.

Updated on 3rd May, 2020 to reflect the changes on how Vercel handles Environment Variables.

If you would like to use Prisma instead of MongoDB, check out my post here.

In this post we’ll create a GraphQL API to create, view and delete notes that we can deploy to Vercel using Apollo Server, TypeScript and MongoDB.

Initial Setup

First Vercel will need be downloaded, installed and logged into.

shell

npm i -g vercel
vercel login

The nice thing about using Vercel for development is that we don’t have to worry about manually configuring TypeScript.

Then create a folder, initialize a project in it and install typescript as a dev dependency.

shell

mkdir serverless-graphql-api-example
cd serverless-graphql-api-example
npm init --yes
npm install typescript -D

Setup the Database

To start setting up the database first go to MongoDB Atlas, create a database and get the connection URI.

If you want a more detailed explanation check out my previous guide here. Note that you can’t follow it exactly here because you can no longer create new accounts in MLab (use MongoDB Atlas instead) and that guide is for an always running server, not a serverless one.

Then install the mongoose dependency. Since we’ll be working with TypeScript we’ll need the types for mongoose as well.

shell

npm install mongoose
npm install @types/mongoose -D

After let’s create the model for the notes. Create a folder called src. After that create a folder called database in src. Then in that create another folder called models and in that create the file note.ts.

Start by importing mongoose.

src/database/models/note.ts

typescript

import mongoose from "mongoose";

Next create an interface for the model which extends from mongoose.Document. We’ll need to export this interface for use later as well.

src/database/models/note.ts

typescript

import mongoose from "mongoose";
export interface INote extends mongoose.Document {
  title: string;
  content: string;
  date: Date;
}

After that we’ll create the schema definition.

src/database/models/note.ts

typescript

import mongoose from "mongoose";
export interface INote extends mongoose.Document {
  title: string;
  content: string;
  date: Date;
}
const schema: mongoose.SchemaDefinition = {
  title: { type: mongoose.SchemaTypes.String, required: true },
  content: { type: mongoose.SchemaTypes.String, required: true },
  date: { type: mongoose.SchemaTypes.Date, required: true },
};

Then define the name of the collection and the schema using schema definition.

src/database/models/note.ts

typescript

import mongoose from "mongoose";
export interface INote extends mongoose.Document {
  title: string;
  content: string;
  date: Date;
}
const schema: mongoose.SchemaDefinition = {
  title: { type: mongoose.SchemaTypes.String, required: true },
  content: { type: mongoose.SchemaTypes.String, required: true },
  date: { type: mongoose.SchemaTypes.Date, required: true },
};
const collectionName: string = "note";
const noteSchema: mongoose.Schema = new mongoose.Schema(schema);

Since we’re building a serverless API, we can’t depend on a persistent connection to database. When creating a Model we need to create it using the database connection.

To compensate for using serverless we’ll generate the Model at runtime using a function. After declaring the function, make it the default export.

src/database/models/note.ts

typescript

import mongoose from "mongoose";
export interface INote extends mongoose.Document {
  title: string;
  content: string;
  date: Date;
}
const schema: mongoose.SchemaDefinition = {
  title: { type: mongoose.SchemaTypes.String, required: true },
  content: { type: mongoose.SchemaTypes.String, required: true },
  date: { type: mongoose.SchemaTypes.Date, required: true },
};
const collectionName: string = "note";
const noteSchema: mongoose.Schema = new mongoose.Schema(schema);
const Note = (conn: mongoose.Connection): mongoose.Model<INote> =>
  conn.model(collectionName, noteSchema);
export default Note;

After creating the Note Model, let’s look into the database connection.

Since we’re using serverless, we can’t have a persistent connection to database for every invocation of the serverless function as this would create issues with scaling.

First create the .env file in the project root to store the MongoDB URI as DB_PATH.

Never commit the .env file. Make sure to add it to.gitignore.

.env

DB_PATH=mongodb://user:password@ds654321.mlab.com:12345/example-db

Next create the the index.js file in src/database and import mongoose.

src/database/index.ts

typescript

import mongoose from "mongoose";

Then get the MongoDB URI that we included in the .env file. It can be accessed as a key in process.env.

src/database/index.ts

typescript

import mongoose from "mongoose";
const uri: string = process.env.DB_PATH;

After that declare a variable that will be used to cache the database connection between function invocations to prevent overloading the database.

src/database/index.ts

typescript

import mongoose from "mongoose";
const uri: string = process.env.DB_PATH;
let conn: mongoose.Connection = null;

Finally we need to create a function that will return a database connection. The function will first check if there is a cached connection. If there is one it will return it or else it will create a new connection and cache and return it. Be sure to export the function.

src/database/index.ts

typescript

import mongoose from "mongoose";
const uri: string = process.env.DB_PATH;
let conn: mongoose.Connection = null;
export const getConnection = async (): Promise<mongoose.Connection> => {
  if (conn == null) {
    conn = await mongoose.createConnection(uri, {
      bufferCommands: false, // Disable mongoose buffering
      bufferMaxEntries: 0, // and MongoDB driver buffering
      useNewUrlParser: true,
      useUnifiedTopology: true,
      useCreateIndex: true,
    });
  }
  return conn;
};

Setup the GraphQL API

First let’s install the dependencies.

shell

npm install apollo-server-micro dayjs graphql graphql-iso-date

Next create the folder api in the project root and in it create the file graphql.ts. Vercel will expose the serverless function in the file as the endpoint /api/graphql. The GraphQL function will be created using Apollo Server. Since we’re doing a serverless version we’ll be using apollo-server-micro.

Then import ApolloServer from apollo-server-micro and the function to create the database connection.

api/graphql.ts

typescript

import { ApolloServer } from "apollo-server-micro";
import { getConnection } from "../src/database";

After that import the GraphQL Schemas and Resolvers. We’ll create these later.

api/graphql.ts

typescript

import { ApolloServer } from "apollo-server-micro";
import { getConnection } from "../src/database";
import typeDefs from "../src/graphql/schema";
import resolvers from "../src/graphql/resolvers";

Then initialize the Apollo Server and export it.

api/graphql.ts

typescript

import { ApolloServer } from "apollo-server-micro";
import { getConnection } from "../src/database";
import typeDefs from "../src/graphql/schema";
import resolvers from "../src/graphql/resolvers";
const apolloServer = new ApolloServer({
  typeDefs,
  resolvers,
});
export default apolloServer.createHandler({ path: "/api/graphql" });

When creating a new ApolloServer, we can define the context option. The context is an object which can be accessed by any of the resolvers.

This will be a good place to initialize the database connection and automatically pass it down to the resolvers.

api/graphql.ts

typescript

import { ApolloServer } from "apollo-server-micro";
import { getConnection } from "../src/database";
import typeDefs from "../src/graphql/schema";
import resolvers from "../src/graphql/resolvers";
const apolloServer = new ApolloServer({
  typeDefs,
  resolvers,
  context: async () => {
    const dbConn = await getConnection();
    return { dbConn };
  },
});
export default apolloServer.createHandler({ path: "/api/graphql" });

To finish configuring we’ll set the introspection and playground options.

api/graphql.ts

typescript

import { ApolloServer } from "apollo-server-micro";
import { getConnection } from "../src/database";
import typeDefs from "../src/graphql/schema";
import resolvers from "../src/graphql/resolvers";
const apolloServer = new ApolloServer({
  typeDefs,
  resolvers,
  context: async () => {
    const dbConn = await getConnection();
    return { dbConn };
  },
  playground: true,
  introspection: true,
});
export default apolloServer.createHandler({ path: "/api/graphql" });

Setup the GraphQL Schema

In the src folder create the graphql folder and in it create the schema folder.

GraphQL doesn’t have a type for date and time so we’ll need a schema for those types. Create custom.ts in the schema folder.

src/graphql/schema/custom.ts

typescript

import { gql } from "apollo-server-micro";
export default gql`
  scalar Date
  scalar Time
  scalar DateTime
`;

Next we’ll setup the schema for Note.

First create the Note schema file, note.ts in the schema folder.

src/graphql/schema/note.ts

typescript

import { gql } from "apollo-server-micro";
export default gql``;

Then add the Note type.

src/graphql/schema/note.ts

typescript

import { gql } from "apollo-server-micro";
export default gql`
  type Note {
    _id: ID!
    title: String!
    content: String!
    date: DateTime!
  }
`;

After that add two Queries, one to get all Notes and one to get a specific Note.

src/graphql/schema/note.ts

typescript

import { gql } from "apollo-server-micro";
export default gql`
  extend type Query {
    getAllNotes: [Note!]
    getNote(_id: ID!): Note
  }
  type Note {
    _id: ID!
    title: String!
    content: String!
    date: DateTime!
  }
`;

Next add two Mutations to create new Notes and delete existing ones.

src/graphql/schema/note.ts

typescript

import { gql } from "apollo-server-micro";
export default gql`
  extend type Query {
    getAllNotes: [Note!]
    getNote(_id: ID!): Note
  }
  extend type Mutation {
    saveNote(title: String!, content: String!): Note!
    deleteNote(_id: ID!): Note
  }
  type Note {
    _id: ID!
    title: String!
    content: String!
    date: DateTime!
  }
`;

Note that we add the keyword extend to the Query and Mutation type in the two schemas. This is because we’ll be joining all the schemas into one later.

Finally we’ll need a schema a join all the others schemas together. Create index.ts in the schema folder and declare and export the Link Schema in an array.

src/graphql/schema/index.ts

typescript

import { gql } from "apollo-server-micro";
const linkSchema = gql`
  type Query {
    _: Boolean
  }
  type Mutation {
    _: Boolean
  }
  type Subscription {
    _: Boolean
  }
`;
export default [linkSchema];

Then import the other Schemas and add them to the array.

src/graphql/schema/index.ts

typescript

import { gql } from "apollo-server-micro";
import noteSchema from "./note";
import customSchema from "./custom";
const linkSchema = gql`
  type Query {
    _: Boolean
  }
  type Mutation {
    _: Boolean
  }
  type Subscription {
    _: Boolean
  }
`;
export default [linkSchema, noteSchema, customSchema];

Setup the resolvers

Since we added a custom type for dates and times in our schema, the first resolver we’ll add is for those types. Create a folder called resolvers in the graphql folder and in it create custom.ts.

src/graphql/resolvers/custom.ts

typescript

import { GraphQLDate, GraphQLTime, GraphQLDateTime } from "graphql-iso-date";
export default {
  Date: GraphQLDate,
  Time: GraphQLTime,
  DateTime: GraphQLDateTime,
};

Next we’ll start working on the resolver for Notes.

Start by creating note.ts in graphql, importing the required dependencies and exporting an empty object which is going to be our resolver.

src/graphql/resolvers/note.ts

typescript

import mongoose from "mongoose";
import dayjs from "dayjs";
import NoteModel, { INote } from "../../database/models/note";
import { ApolloError } from "apollo-server-micro";
export default {};

Then define an object for the Queries.

src/graphql/resolvers/note.ts

typescript

import mongoose from "mongoose";
import dayjs from "dayjs";
import NoteModel, { INote } from "../../database/models/note";
import { ApolloError } from "apollo-server-micro";
export default {
  Query: {},
};

When defining the function for each of the queries (or mutations or fields) there are three important parameters.

parent - If you a resolving a field of an object this parameter will contain it. We won’t be needing it in this post.
args - The arguments passed to the query or mutation.
context - The context object created when setting up the API. In our case it’ll contain the database connection.

First let’s write the resolver for getAllNotes. Start by declaring the function.

src/graphql/resolvers/note.ts

typescript

import mongoose from "mongoose";
import dayjs from "dayjs";
import NoteModel, { INote } from "../../database/models/note";
import { ApolloError } from "apollo-server-micro";
export default {
  Query: {getAllNotes: async (
      parent,
      args,
      context
    ): Promise<INote[]> => {}
};

Let’s destructure the context object.

src/graphql/resolvers/note.ts

typescript

import mongoose from "mongoose";
import dayjs from "dayjs";
import NoteModel, { INote } from "../../database/models/note";
import { ApolloError } from "apollo-server-micro";
export default {
  Query: {getAllNotes: async (
      parent,
      args,
      { dbConn }: { dbConn: mongoose.Connection }
    ): Promise<INote[]> => {}
};

In the function we should create the Note model using the database connection from the context and retrieve all the Notes using it.

src/graphql/resolvers/note.ts

typescript

import mongoose from "mongoose";
import dayjs from "dayjs";
import NoteModel, { INote } from "../../database/models/note";
import { ApolloError } from "apollo-server-micro";
export default {
  Query: {
    getAllNotes: async (
      parent,
      args,
      { dbConn }: { dbConn: mongoose.Connection }
    ): Promise<INote[]> => {
      const Note: mongoose.Model<INote> = NoteModel(dbConn);
      let list: INote[];
      try {
        list = await Note.find().exec();
      } catch (error) {
        console.error("> getAllNotes error: ", error);
        throw new ApolloError("Error retrieving all notes");
      }
      return list;
    },
  },
};

After that let’s add the resolver for getNote.

src/graphql/resolvers/note.ts

typescript

import mongoose from "mongoose";
import dayjs from "dayjs";
import NoteModel, { INote } from "../../database/models/note";
import { ApolloError } from "apollo-server-micro";
export default {
  Query: {
    getAllNotes: async (
      parent,
      args,
      { dbConn }: { dbConn: mongoose.Connection }
    ): Promise<INote[]> => {
      const Note: mongoose.Model<INote> = NoteModel(dbConn);
      let list: INote[];
      try {
        list = await Note.find().exec();
      } catch (error) {
        console.error("> getAllNotes error: ", error);
        throw new ApolloError("Error retrieving all notes");
      }
      return list;
    },
    getNote: async (
      parent,
      { _id }: { _id: INote["_id"] },
      { dbConn }: { dbConn: mongoose.Connection }
    ): Promise<INote> => {
      const Note: mongoose.Model<INote> = NoteModel(dbConn);
      try {
        const note = await Note.findById(_id).exec();
        return note;
      } catch (error) {
        console.error("> getNote error: ", error);
        throw new ApolloError("Error retrieving all notes");
      }
    },
  },
};

We can retrieve the id argument from the second parameter.

Let’s finish of the Note resolvers by adding the resolvers for the mutations.

src/graphql/resolvers/note.ts

typescript

import mongoose from "mongoose";
import dayjs from "dayjs";
import NoteModel, { INote } from "../../database/models/note";
import { ApolloError } from "apollo-server-micro";
export default {
  Query: {
    getAllNotes: async (
      parent,
      args,
      { dbConn }: { dbConn: mongoose.Connection }
    ): Promise<INote[]> => {
      const Note: mongoose.Model<INote> = NoteModel(dbConn);
      let list: INote[];
      try {
        list = await Note.find().exec();
      } catch (error) {
        console.error("> getAllNotes error: ", error);
        throw new ApolloError("Error retrieving all notes");
      }
      return list;
    },
    getNote: async (
      parent,
      { _id }: { _id: INote["_id"] },
      { dbConn }: { dbConn: mongoose.Connection }
    ): Promise<INote> => {
      const Note: mongoose.Model<INote> = NoteModel(dbConn);
      try {
        const note = await Note.findById(_id).exec();
        return note;
      } catch (error) {
        console.error("> getNote error: ", error);
        throw new ApolloError("Error retrieving all notes");
      }
    },
  },
  Mutation: {
    saveNote: async (
      parent,
      { title, content }: { title: INote["title"]; content: INote["content"] },
      { dbConn }: { dbConn: mongoose.Connection }
    ): Promise<INote> => {
      const Note: mongoose.Model<INote> = NoteModel(dbConn);
      try {
        const note = await Note.create({
          title,
          content,
          date: dayjs().toDate(),
        });
        return note;
      } catch (error) {
        console.error("> saveNote error: ", error);
        throw new ApolloError("Error creating note");
      }
    },
    deleteNote: async (
      parent,
      { _id }: { _id: INote["_id"] },
      { dbConn }: { dbConn: mongoose.Connection }
    ): Promise<INote> => {
      const Note: mongoose.Model<INote> = NoteModel(dbConn);
      try {
        const note = await Note.findByIdAndDelete(_id).exec();
        return note;
      } catch (error) {
        console.error("> getNote error: ", error);
        throw new ApolloError("Error retrieving all notes");
      }
    },
  },
};

Now we have to connect all the resolvers together using an index.ts file created in the resolvers folder.

src/graphql/resolvers/index.ts

typescript

import noteResolver from "./note";
import customResolver from "./custom";
export default [noteResolver, customResolver];

Running the API locally

To run the project locally first you need to link it to a Vercel project.

shell

vercel link

Then run vercel dev to start the API locally.

shell

vercel dev

If you visit http://localhost:3000/api/graphql you can see the GraphQL Playground.

Deploying to Vercel

First we need to upload the database path as an Environment Variable.

shell

vercel env add

Name the variable DB_PATH and make sure you make to available for all three environments (Production, Preview and Development).

Then all that’s left to do is to deploy to Vercel.

shell

vercel

The GraphQL Playground should be visible in the /api/graphql route of the URL returned.

Wrapping Up

I made a sample deployment which you can check out below or here. The source code is available on GitHub.

If you want a more detailed explanation into GraphQL and making a server for it, you can check out the excellent guide that I learnt from here.

Bonus: Using it in Next.js

You can create a GraphQL API endpoint in your Next.js project by putting the graphql.ts inside the api folder in the pages folder.

The only extra step you need to do is to add this bit of code to the end of the graphql.ts file.

pages/api/graphql.ts

typescript

// ...
export default apolloServer.createHandler({ path: "/api/graphql" });
export const config = {
  api: {
    bodyParser: false,
  },
};