Fulltext Search on Firebase with Meilisearch

A common issue that developers face with NoSQL databases (and SQL to a lesser extent) is a lack of full-text search capabilities. If you want to build a performant typeahead search box or handle multi-property filtering on a collection, you’ll find it frustratingly difficult to implement in Firestore - it’s just not the right database for the job. There are many good solutions, like Algolia and ElasticSearch, but they can be expensive and/or complex to manage. Today we’ll look at a super-fast Rust-based open-source search engine called MeiliSearch. We’ll use it to convert a Firestore collection into a fully searchable index (great for autocomplete search), then deploy it to a VM on Google Cloud.

Initial Setup

If you’re new to Docker, get familiar with my Docker Tutorial

Run MeiliSearch with Docker

docker run -it --rm \
    -p 7700:7700 \
    -v $(pwd)/data.ms:/data.ms \
    getmeili/meilisearch

The following command will run a MeiliSearch container on http://localhost:7700

Create an Index

Create an index for searchable content. You can run a cURL command or use a HTTP client like Postman or Insomnia.

Initialize Firebase

Initialize Firebase with Firestore, Cloud Functions, and the Emulator Suite.

firebase init

cd functions
npm install meilisearch

MeiliSearch Cloud Functions

Initialize the connection to meilisearch.

const functions = require('firebase-functions');

const MeiliSearch = require('meilisearch');

const client = new MeiliSearch({
    host: 'http://localhost:7700',
    apiKey: '',
  });

Index on Firestore Create

The purpose of this function is to index every newly created Firestore document in Meilisearch.

exports.meilisearchIndex = functions.firestore
  .document('movies/{id}')
  .onCreate(async (snapshot, context) => {
    const index = client.getIndex('movies');

    const id = snapshot.id;
    const { title, year, description } = snapshot.data();

    const response = await index.addDocuments([
      { id, title, year, description }
    ])
    console.log(response)

  });

HTTP Proxy for Searches

You can proxy searches through your Cloud Functions or make requests directly from a frontend app (using the meilisearch public key).

exports.meilisearchQuery = functions.https.onRequest(async (req, res) => {
  const index = client.getIndex('movies');

  const search = await index.search(req.body.q);

  res.send(search);
});

Deploy MeiliSearch to Google Cloud

At this point, we need a way to deploy MeiliSearch to the cloud. Because it is a stateful service, we must mount a persistent disk volume to any container that runs it.

The minimum monthly price on GCP you will pay is $4.53 for 1 micro VM and $1.70 10GB of SSD disk space. If pricing is important, I would recommend looking into Digital Ocean for small VMs like this.

Create a Disk

Create a persistent disk to hold your data. SSD is more expensive, but will deliver faster I/O and performance is usually critical for fulltext search.

Create a VM

Create a VM with the following settings:

• Allow HTTP and HTTPS traffic
• Attach the disk from the previous step
• Launch from container. getmeili/meilisearch:v0.13.0
• Set production env variables (advanced container settings)
• Mount disk as volume (advanced container settings)

MEILI_HTTP_ADDR=0.0.0.0:80 
MEILI_MASTER_KEY="super_secret" 
MEILI_ENV="production"
MEILI_DB_PATH="./meili-data"

This will give you an external IP that you can access over http, i.e http://35.232.183.124

Get the API keys

In production, MeiliSearch requires requests to be authenticated with an API key.