deepseek_chatbot/api/PDFReader.php

<?php

/**
 * PDFReader class to read PDF files from a specified directory.
 * It reads all PDF files, extracts their text, and returns an array of rows (one per file).
 */
class PDFReader {
    /**
     * Directory where PDF files are stored.
     * @var string
     */
    private $dir;

    /**
     * Constructor to initialize the PDFReader with a directory.
     *
     * @param string $directory The directory containing PDF files.
     */
    public function __construct($directory) {
        $this->dir = rtrim($directory, '/') . '/';
    }

    /**
     * Reads all PDF files in the specified directory and returns an array of extracted texts.
     *
     * @return array An array of associative arrays with filename and extracted text.
     */
    public function readAll() {
        $rows = array();
        foreach (glob($this->dir . '*.pdf') as $file) {
            $text = $this->readFile($file);
            $rows[] = [
                'filename' => basename($file),
                'text' => $text
            ];
        }
        return $rows;
    }

    /**
     * Extract text from a PDF file using pdftotext command line tool.
     *
     * @param string $file Path to the PDF file (relative or absolute).
     * @return string Extracted text from the PDF.
     */
    public function readFile($file) {
        $filePath = strpos($file, $this->dir) === 0 ? $file : $this->dir . ltrim($file, '/');
        $tmp = tempnam(sys_get_temp_dir(), 'pdftext');
        exec("pdftotext " . escapeshellarg($filePath) . " " . escapeshellarg($tmp));
        $text = file_get_contents($tmp);
        unlink($tmp);
        return $text ?: '';
    }

    /**
     * Converts an array of PDF rows into a string format.
     *
     * @param array $pdfRows An array of associative arrays representing the PDF rows.
     * @return string A string representation of the PDF rows.
     */
    public function toString($pdfRows) {
        $pdfString = '';
        foreach ($pdfRows as $pdfRow) {
            $pdfString .= "- " . $pdfRow['filename'] . ":\n" . trim($pdfRow['text']) . "\n\n";
        }
        return $pdfString;
    }
}