Two weeks ago, I was playing with some ML ideas in a Laravel application. I didn't quite like the accuracy of a certain cloud provider's ML product, so I wanted to try another. To do this, I had to re-write the client implementation in my app for each provider.

There must be a better way, right?

Other projects like Laravel Cashier and various Laravel facades allow you to interact with a single API, but you can swap out the underlying drivers under the hood. I thought, why don't we have something similar for ML tasks?

Introducing: Dream, an AI / ML toolbox for Laravel

Dream

What does Dream do? Dream is simply a Facade to an underlying class that allow you to do common ML tasks and return consistent responses, without having to re-write your application code. This allows you to write your application code once.

$meme = Storage::get('meme.jpg');

$sentiment = Dream::image($meme)->text()->sentiment();

The above code detects the text in an image with OCR, then analyzes the sentiment, all in a single line of code.

The $sentiment response also has helpers to make taking action on the sentiment easier:

$sentiment->positive(); // true

Dream can also do text related NLP tasks, like determining key phrases:

Dream::text('Laravel is a web application framework with expressive, 
elegant syntax. We’ve already laid the foundation — freeing you to create 
without sweating the small things.')
  ->phrases()
  ->pluck('text')
  ->toArray();

// [
   //   "Laravel",
   //   "a web application framework",
   //   "expressive, elegant syntax",
   //   "the foundation —",
   //   "the small things",
// ]

Or just determining what language a text is written in:

    Dream::text('¿Cuál es tu película favorita?')->language(); // 'es'

You may also combine these interchangeably. So the example meme from earlier, we can detect what language is in the image:

$meme = Storage::get('meme.jpg');

$sentiment = Dream::image($meme)->text()->language();

See more

Dream is still very early. As of writing this, I only have one driver for AWS Comprehend & AWS Rekgonition, but I will be adding one for Azure's products. You can also create your own and mix & match services.

To see the repo: github.com/christopherarter/dream