Michael Lefkowitz

Build an Image Search Microservice with the Imgur API

November 01, 2016

It’s been a while since I’ve had time to touch Free Code Camp backend projects, but during my week off I decided to build out the next API project I had yet to attempt, the Image Search Abstraction Layer microservice. I love building and working with APIs, so this was a fun project to build. Let’s get started!

First, let’s initialize our git repository and jump into it. git init image-search-microservice cd image-search-microservice

Now, we’ll want to setup our project’s package.json. If you haven’t tested out Yarn yet, I highly recommend giving it a try. yarn init

Speaking of which, make sure you’ve updated your copy of Node for this tutorial, as I’ll be using some ES6.

Whether you’re using npm or yarn, you can skip through most of the default settings. Personally though, I’m changing my entry point to app.js.

Next, let’s set up our .gitignore to ignore the node_modules folder and our .env file, which we’ll introduce a little later. Drop these two lines in your terminal to do this quickly. echo node_modules >> .gitignore echo .env >> .gitignore

Now let’s install the node modules we’ll need for this project. The list is pretty short for this one: yarn add express mongoose request yarn add dotenv --dev

Go ahead and create your app.js and setup the basic scaffolding for an express app.

const express = require('express');

const app = express();

const port = process.env.PORT '' 3000;
const server = app.listen(port, function() {
    console.log(`Server listening on port ${port}`);

At this point, you should be able to run nodemon (assuming you have it installed globally) in your terminal and see your server running on port 3000. If not, go ahead and compare your project files.

I’m all about writing modular code, so we’re going to set up some folders and files like so: config/db.js models/history.js routes/index.js services/imgur.js

First we’ll set up the routing. Add const routes = require('./routes/index'); and app.use('/', routes); in our app.js to be used as middleware. Then, setup your routes/index.js like so:

const express = require("express");
const router = express.Router();

router.get("/", (req, res) => {

router.get("/latest", (req, res) => {});

router.get("/search/:q", (req, res) => {});

module.exports = router;

Now, we’ve moved the job of routing out of our app.js. The advantage to doing this is that if our app were to grow, we could add additional route files and keep our project modular and clean. I’ve also setup the basic routing our application will need: a root endpoint where we can eventually display a landing page if we so choose, a /latest endpoint where our app will display the ten latest searches, and a /search/:q endpoint where we can pass a string in to search.

Next, head over to the services/imgur.js file and drop this in:

const request = require("request");

exports.getImage = function(search, page = 1) {
  return new Promise((resolve, reject) => {});

We’ll be making a request to the Imgur API through our app, and the request module makes that really simple. Also, we’re going to be calling this function back in our route file, so I’ve set it up as an exported function that takes two parameters. The first is the search term, the second is the pagination option which we can default to ‘1’ if nothing is passed. And finally, I’m a big fan of promises over callbacks, so I’ve set this function to return a promise.

Now, per the documentation, the request module can be invoked with two parameters - an options object and a callback. In our options object, we’ll need to pass in our unique client ID from Imgur. Swing over to api.imgur.com if you haven’t done so already and register your application. You can choose ‘Anonymous usage without user authorization’ when prompted.

Now that we’ve got our code, we can build out an options object within our promise statement like so:

let options = {
  url: `https://api.imgur.com/3/gallery/search/${page}?q=${search}`,
  headers: { Authorization: "Client-ID kfbr392kfbr392" },
  json: true

In here, we’re setting up the URL that we’ll be connecting to Imgur with. Here, we’ve used ES6 template strings to cleanly drop in the page and search parameters from our getImage function. Next, we set the headers as requested by Imgur to allow proper authorization of our API call (but, make sure to replace my ‘Client-ID with your own). And finally, we’ve specified that we want our response to be in JSON format.

Next, we’ll build out our callback. First, we can build a function called getPics that will take three parameters: error, response, and body. I’ll then make a quick error check inside the body of the function that ensures no errors occurred and the response received from Imgur has a status code of 200. Then, we want to take the body that was returned, and filter out items in the array that are albums. We’ll do this to best mirror the demo app shared by Free Code Camp, as we’ll want to provide a response with both a direct image link and a link to the image’s context - and unfortunately this wouldn’t work well if we included albums.

Then, after we’ve filtered out albums, we then want to map over the response and cut out all the extra information we’re not using. All we want to return is a url, a snippet, and the context. Finally, we’re going to resolve our promise with our newly transformed data, jump out of the callback and then setup our request function with its two newly created parameters. In the end, it will look like this:

function getPics(err, response, body) {
  if (!err && response.statusCode == 200) {
    body = body.data
      .filter(image => {
        if (!image.is_album) {
          return image;
      .map(image => {
        return {
          url: image.link,
          snippet: image.title,
          context: `https://imgur.com/${image.id}`

We’re almost ready to test it out, but first we have to jump back into our routes file and make sure to require our new imgur service file const imgur = require('../services/imgur'); and to call it in our /search/:q endpoint.

Here, we’ll pass the ‘q’ parameters as the search string of the function, and then we’ll let our second parameter req.query.offset handle the optional ?offset=42 flag. Then, once that promise returns, we’re going to output the response to our browser as JSON.

imgur.getImage(req.params.q, req.query.offset).then(ans => {

Now fire up nodemon and give it a test. If all is working so far, you should be able to search Imgur by pointing your browser to localhost:3000/search/puppies. If not, do some debugging or a quick code comparison at this point.

Now that this piece is working, all that we have left to add is the history component. I personally enjoy working with mLab in development just as much as in production, so I’m going to proceed with this tutorial as such. If you’d prefer to run MongoDB on your local machine prior to deployment, or you want to learn more about how to get setup on mLab - checkout my previous tutorial on building a URL shortener for more information.

The first thing we’re going to want to do is setup our environment variables. This way, we’ll able to connect to our mLab database in either production or development, without worrying about sharing our credentials on Github. In your app.js file, drop in the following at the very top:

if (process.env.NODE_ENV !== "production") {

Here, we’re setting up our server to require the dotenv package if we’re not running in a production environment. Later when we push this up to Heroku, their platform will automatically set an environment variable of NODE_ENV as ‘production’, but on our development machines - we won’t worry about that.

Next you’ll need to create a .env file in the root of your project. This file will then assign your environment variables every time you spin up your server. In that, drop in the following as it relates to the mLab database you set up:


Now that we have our environment variables setup, we can dive into setting up our connection to the database. Back in your app.js, you’ll want to require your /config/db.js file with a const db = require(‘./config/db’);

Then, jump into your config/db.js file and drop in the following:

const connection = `mongodb://${process.env.DB_HOST}/${process.env.DB_NAME}`;
const mongoose = require("mongoose");
mongoose.Promise = global.Promise;

exports.db = mongoose.connect(connection);

Here, we are setting up a connection to our mongo database and then using mongoose as an ORM for interacting with it. Also, I’ll be using some promises, which were recently deprecated in mongoose, so I’ve set the native promise library to be used instead. And finally, we export the connection for use elsewhere in our application.

Next, in our models/history.js file, I want to set up a schema for our database. First, we’ll need to require an instance of mongoose. Next, we want to setup a schema for our database. This way, when we pass data into the model later in our routes, we’ll be able to let MongoDB know what we want each new document to look like, and specifically in this case - to always attach a timestamp to each new entry. After that, we export our model for use.

const mongoose = require("mongoose");

const historySchema = new mongoose.Schema({
  term: String,
  when: { type: Date, default: Date.now }

const History = mongoose.model("History", historySchema);
module.exports = History;

At this point, our database is setup and we’re ready to start building out the queries we’ll need.

First, let’s make sure to require our new model in our router with const History = require('../models/history');

Now, we want to add a new entry into our database every time a search is made. Right before we send back a response with the results, drop in the following to add that search query into the database: new History({ term: req.params.q }).save();

And finally, we need our /latest route to return the most recent 10 entries. For that, we’re going to query the database like so:

History.find({}, "term when -_id")
  .then(results => {

Here, we first pass in an empty object, which will return all documents. Then, we specify that in those documents, we only are interested in the search term and the date, and we’d like to specifically exclude the unique _id field from the results. Afterwards, we instruct the query to sort the results in descending order and to only return the ten most recent documents. Finally, we pass those results through a promise and return them as JSON.

Go ahead and run your app with Nodemon and verify that everything is working. If so, you’re ready for deployment. Since we’re using ES6, you’ll need to define your node engine in your package.json, otherwise Heroku will, at the time of this writing, default to version 5.1.1 which won’t support ES6 unless you use strict mode. Also, you’ll need to make sure to add a start script to your package.json. To see what those look like, check out the file on my GitHub. And finally, make sure to drop your environment variables into your Heroku app, otherwise you definitely won’t be able to connect to your database.

That’s all there is to it. Here’s my final code for the project. Personally I went back and threw in pug as a template engine and a favicon server. The final deployed version can be seen here. Feel free to leave any questions or comments below.