pdf-ai is a simple PHP library that makes extracting data from PDFs for large language models easy. It uses a single dependency, the Symfony Process Component, to interface with the Poppler command line tools from the xpdf library.

Install the library using Composer:

composer require 1tomany/pdf-ai

Before beginning, ensure the pdfinfo, pdftoppm, and pdftotext binaries are installed and located in the $PATH environment variables.

brew install poppler

apt-get install poppler-utils

This library has three main features:

• Read PDF metadata such as the number of pages
• Rasterize one or more pages to JPEG or PNG images
• Extract text from one or more pages

Extracted data is stored in memory and can be written to the filesystem or converted to a data: URI. Because extracted data is stored in memory, this library returns a \Generator object for each page that is extracted or rasterized.

Using the library is easy, and you have two ways to interact with it:

1. Direct Instantiate the OneToMany\PDFAI\Client\Poppler\PopplerExtractorClient class and call the methods directly. This method is easier to use, but comes with the cost that your application will be less flexible and testable.
2. Actions Create a container of OneToMany\PDFAI\Contract\Client\ExtractorClientInterface objects, and use the OneToMany\PDFAI\Factory\ExtractorClientFactory class to instantiate them.

Note: A Symfony bundle is available if you wish to integrate this library into your Symfony applications with autowiring and configuration support.

<?php

require_once __DIR__ . '/vendor/autoload.php';

use OneToMany\PDFAI\Client\Poppler\PopplerExtractorClient;
use OneToMany\PDFAI\Contract\Enum\OutputType;
use OneToMany\PDFAI\Request\ExtractDataRequest;
use OneToMany\PDFAI\Request\ExtractTextRequest;
use OneToMany\PDFAI\Request\ReadMetadataRequest;

$filePath = '/path/to/file.pdf';

// Construct the Poppler wrapper
$client = new PopplerExtractorClient();

// Construct and execute a request to read the PDF metadata
$metadata = $client->readMetadata(new ReadMetadataRequest($filePath));

vprintf("The PDF '%s' has %d page(s).\n", [
    $filePath, $metadata->getPages(),
]);

// Construct a request to rasterize all pages as 150 DPI JPEGs
$request = new ExtractDataRequest($filePath, 1, null, OutputType::Jpg, 150);

foreach ($client->extractData($request) as $image) {
    // $image->getData() or $image->toDataUri()
    printf("MD5: %s\n", md5($image->getData()));
}

// Extract text from pages 3 and 4
$request = new ExtractTextRequest($filePath, 3, 4);

foreach ($client->extractData($request) as $text) {
    // $text->getData()
    printf("Length: %d\n", strlen($text->getData()));
}

Run the test suite with PHPUnit:

./vendor/bin/phpunit

Run static analysis with PHPStan:

./vendor/bin/phpstan

• Vic Cherubini, 1:N Labs, LLC

The MIT License