math_cms/app/Services/OCR/Drivers/AliyunOCRDriver.php

<?php

namespace App\Services\OCR\Drivers;

use App\Services\OCR\OCRInterface;
use AlibabaCloud\SDK\Ocrapi\V20210707\Ocrapi;
use AlibabaCloud\Tea\Utils\Utils\RuntimeOptions;
use Darabonba\OpenApi\Models\Config;
use AlibabaCloud\SDK\Ocrapi\V20210707\Models\RecognizeEduPaperOcrRequest;
use Illuminate\Support\Facades\Log;

class AliyunOCRDriver implements OCRInterface
{
    protected $client;

    public function __construct(array $config, $client = null)
    {
        if ($client) {
            $this->client = $client;
            return;
        }

        $apiConfig = new Config([
            'accessKeyId' => $config['access_key_id'],
            'accessKeySecret' => $config['access_key_secret'],
            'endpoint' => $config['endpoint'],
        ]);

        $this->client = new Ocrapi($apiConfig);
    }

    public function recognize(string $imagePath, array $options = []): array
    {
        try {
            // Check if file exists
            if (!file_exists($imagePath)) {
                throw new \Exception("Image file not found: {$imagePath}");
            }

            // Get cutType from options (default: "question")
            $cutType = $options['cutType'] ?? 'question';
            $subject = $options['subject'] ?? 'Math';

            // Read file content
            $fileStream = fopen($imagePath, 'rb');
            $stream = \GuzzleHttp\Psr7\Utils::streamFor($fileStream);

            $request = new \AlibabaCloud\SDK\Ocrapi\V20210707\Models\RecognizeEduPaperCutRequest([
                'body' => $stream,
                'cutType' => $cutType,
                'imageType' => 'photo',
                'subject' => $subject,
                'outputOricoord' => false
            ]);

            $runtime = new RuntimeOptions([]);
            
            // Call Aliyun API
            $response = $this->client->recognizeEduPaperCutWithOptions($request, $runtime);

            // Close stream
            if (is_resource($fileStream)) {
                fclose($fileStream);
            }

            // Parse response
            $body = json_decode(json_encode($response->body), true);
            
            // Detailed logging
            Log::info('Aliyun EduPaperCut Full Response', [
                'cutType' => $cutType,
                'has_data' => isset($body['data']),
                'request_id' => $body['requestId'] ?? null,
                'code' => $body['code'] ?? null,
                'message' => $body['message'] ?? null,
                'body_keys' => array_keys($body ?? [])
            ]);
            
            // Log raw data if exists
            if (isset($body['data'])) {
                $dataPreview = is_string($body['data']) 
                    ? substr($body['data'], 0, 500) 
                    : json_encode($body['data']);
                Log::info('Aliyun Data Preview', ['data' => $dataPreview]);
            }
            
            // Extract data from Aliyun response
            $questions = [];
            
            if (isset($body['data'])) {
                // The data field is a JSON string
                $data = is_string($body['data']) ? json_decode($body['data'], true) : $body['data'];
                
                // Extract page_list -> subject_list OR answer_list
                if (isset($data['page_list']) && is_array($data['page_list'])) {
                    foreach ($data['page_list'] as $page) {
                        // Determine which list to use based on cutType
                        $itemList = null;
                        if ($cutType === 'answer' && isset($page['answer_list'])) {
                            $itemList = $page['answer_list'];
                        } elseif (isset($page['subject_list'])) {
                            $itemList = $page['subject_list'];
                        }
                        
                        if ($itemList && is_array($itemList)) {
                            foreach ($itemList as $item) {
                                // Extract question/answer data
                                $questionNumber = null;
                                if (isset($item['ids']) && is_array($item['ids']) && !empty($item['ids'])) {
                                    $questionNumber = $item['ids'][0];
                                } else {
                                    $questionNumber = count($questions) + 1;
                                }
                                
                                // Get text - if not provided, build from prism_wordsInfo
                                $text = $item['text'] ?? '';
                                if (empty($text) && isset($item['prism_wordsInfo']) && is_array($item['prism_wordsInfo'])) {
                                    $words = [];
                                    foreach ($item['prism_wordsInfo'] as $wordInfo) {
                                        if (isset($wordInfo['word'])) {
                                            $words[] = $wordInfo['word'];
                                        }
                                    }
                                    $text = implode('', $words);
                                }
                                
                                // Calculate confidence from prism_wordsInfo
                                $confidence = 0.0;
                                if (isset($item['prism_wordsInfo']) && is_array($item['prism_wordsInfo'])) {
                                    $totalProb = 0;
                                    $count = 0;
                                    foreach ($item['prism_wordsInfo'] as $wordInfo) {
                                        if (isset($wordInfo['prob'])) {
                                            $totalProb += $wordInfo['prob'];
                                            $count++;
                                        }
                                    }
                                    $confidence = $count > 0 ? ($totalProb / $count) / 100 : 0.0;
                                }
                                
                                $questions[] = [
                                    'question_number' => $questionNumber,
                                    'content' => $text,
                                    'cut_type' => $cutType,
                                    'confidence' => $confidence,
                                    'raw_data' => $item
                                ];
                            }
                        }
                    }
                }
            }

            return [
                'raw' => $body,
                'questions' => $questions,
                'cut_type' => $cutType
            ];

        } catch (\Exception $e) {
            Log::error('Aliyun OCR Error', [
                'message' => $e->getMessage(),
                'trace' => $e->getTraceAsString(),
            ]);
            throw $e;
        }
    }
}