dobetternorge-tools/includes/AzureDocIntelligence.php

<?php
declare(strict_types=1);

/**
 * Azure Document Intelligence (formerly Form Recognizer) — Read API for OCR + layout.
 *
 * Endpoint: https://bnl-doc-intelligence.cognitiveservices.azure.com/
 * Auth: Ocp-Apim-Subscription-Key header
 *
 * Read API is async: POST to /documentintelligence/documentModels/prebuilt-read:analyze
 *   → returns 202 + Operation-Location header with poll URL
 *   → poll until status == "succeeded", then parse analyzeResult.content
 */
final class AzureDocIntelligence
{
    private string $endpoint;
    private string $key;

    public function __construct(?string $endpoint = null, ?string $key = null)
    {
        $cfg = self::loadConfig();
        $this->endpoint = rtrim($endpoint ?? ($cfg['endpoint'] ?? ''), '/');
        $this->key      = $key ?? ($cfg['key'] ?? '');
        if ($this->endpoint === '' || $this->key === '') {
            throw new RuntimeException('AzureDocIntelligence: endpoint or key not configured.');
        }
    }

    private static function loadConfig(): array
    {
        $path = '/etc/bnl/azure.php';
        if (is_readable($path)) {
            $cfg = require $path;
            return [
                'endpoint' => (string)($cfg['DOC_INTELLIGENCE_ENDPOINT'] ?? ''),
                'key'      => (string)($cfg['DOC_INTELLIGENCE_KEY'] ?? ''),
            ];
        }
        return [
            'endpoint' => (string)(getenv('AZURE_DOC_INTELLIGENCE_ENDPOINT') ?: ''),
            'key'      => (string)(getenv('AZURE_DOC_INTELLIGENCE_KEY') ?: ''),
        ];
    }

    /**
     * OCR a local PDF file using the prebuilt-read model.
     * Returns: ['content' => string, 'pages' => array, 'languages' => array]
     */
    public function readPdf(string $localPath, int $pollTimeoutSeconds = 120): array
    {
        if (!is_readable($localPath)) {
            throw new InvalidArgumentException("Unreadable file: {$localPath}");
        }
        $url = $this->endpoint . '/documentintelligence/documentModels/prebuilt-read:analyze?api-version=2024-11-30';
        $body = file_get_contents($localPath);

        $ch = curl_init();
        curl_setopt_array($ch, [
            CURLOPT_URL => $url,
            CURLOPT_POST => true,
            CURLOPT_POSTFIELDS => $body,
            CURLOPT_HTTPHEADER => [
                'Content-Type: application/pdf',
                'Ocp-Apim-Subscription-Key: ' . $this->key,
            ],
            CURLOPT_RETURNTRANSFER => true,
            CURLOPT_HEADER => true,
            CURLOPT_TIMEOUT => 60,
        ]);
        $response = curl_exec($ch);
        $headerSize = (int)curl_getinfo($ch, CURLINFO_HEADER_SIZE);
        $status = (int)curl_getinfo($ch, CURLINFO_RESPONSE_CODE);
        curl_close($ch);

        if ($status !== 202 || !is_string($response)) {
            throw new RuntimeException("DocIntelligence analyze failed: HTTP {$status}");
        }
        $headers = substr($response, 0, $headerSize);
        if (!preg_match('/Operation-Location:\s*(.+?)\r?\n/i', $headers, $m)) {
            throw new RuntimeException('DocIntelligence: missing Operation-Location header.');
        }
        $pollUrl = trim($m[1]);

        $deadline = time() + $pollTimeoutSeconds;
        while (time() < $deadline) {
            usleep(1500_000);
            $pollCh = curl_init();
            curl_setopt_array($pollCh, [
                CURLOPT_URL => $pollUrl,
                CURLOPT_HTTPHEADER => ['Ocp-Apim-Subscription-Key: ' . $this->key],
                CURLOPT_RETURNTRANSFER => true,
                CURLOPT_TIMEOUT => 30,
            ]);
            $pollResp = curl_exec($pollCh);
            $pollStatus = (int)curl_getinfo($pollCh, CURLINFO_RESPONSE_CODE);
            curl_close($pollCh);
            if ($pollStatus !== 200 || !is_string($pollResp)) {
                throw new RuntimeException("DocIntelligence poll failed: HTTP {$pollStatus}");
            }
            $data = json_decode($pollResp, true);
            $st = (string)($data['status'] ?? '');
            if ($st === 'succeeded') {
                $result = $data['analyzeResult'] ?? [];
                return [
                    'content' => (string)($result['content'] ?? ''),
                    'pages' => $result['pages'] ?? [],
                    'languages' => $result['languages'] ?? [],
                    'page_count' => count($result['pages'] ?? []),
                ];
            }
            if ($st === 'failed') {
                $err = $data['error']['message'] ?? 'unknown';
                throw new RuntimeException("DocIntelligence analysis failed: {$err}");
            }
            // 'running' or 'notStarted' — continue polling
        }
        throw new RuntimeException("DocIntelligence poll timeout after {$pollTimeoutSeconds}s.");
    }
}