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 AlibabaCloud\SDK\Ocrapi\V20210707\Models\RecognizeEduPaperCutRequest;
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 parameters from options
            $cutType = $options['cutType'] ?? 'question';
            $subject = $options['subject'] ?? 'Math';
            $ocrRecordId = $options['ocr_record_id'] ?? null;

            // 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
                                $questionNumber = count($questions) + 1; // 默认使用索引
                                if (isset($item['ids']) && is_array($item['ids']) && !empty($item['ids'])) {
                                    $idValue = $item['ids'][0];
                                    // 只有当 ids[0] 是数字时才使用它作为题号
                                    if (is_numeric($idValue)) {
                                        $questionNumber = (int) $idValue;
                                    }
                                }
                                
                                // 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
                                ];
                            }
                        }
                    }
                }
            }

            // 保存完整的API响应到ocr_raw_data表
            if (isset($body['requestId']) && !empty($questions)) {
                try {
                    \App\Models\OCRRawData::saveRawResponse($ocrRecordId ?? 0, $body);
                    \Log::info('Aliyun OCR: 原始数据已保存', [
                        'request_id' => $body['requestId'],
                        'questions_count' => count($questions),
                        'ocr_record_id' => $ocrRecordId ?? null
                    ]);
                } catch (\Exception $e) {
                    \Log::error('Aliyun OCR: 保存原始数据失败', [
                        'error' => $e->getMessage(),
                        'request_id' => $body['requestId'] ?? null
                    ]);
                }
            }

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

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

    /**
     * 识别手写内容(使用RecognizeEduPaperOcr接口)
     * 
     * @param string $imagePath 图片路径
     * @param array $options 选项参数
     * @return array 识别结果
     */
    public function recognizeHandwriting(string $imagePath, array $options = []): array
    {
        try {
            // Check if file exists
            if (!file_exists($imagePath)) {
                throw new \Exception("Image file not found: {$imagePath}");
            }

            // Get parameters from options
            $subject = $options['subject'] ?? 'Math';
            $ocrRecordId = $options['ocr_record_id'] ?? null;

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

            $request = new RecognizeEduPaperOcrRequest([
                'body' => $stream,
                'imageType' => 'photo',
                'subject' => $subject,
                'textType' => '2',  // 2 = 手写体
                'outputOricoord' => true  // 输出坐标信息
            ]);

            $runtime = new RuntimeOptions([]);
            
            // Call Aliyun API
            $response = $this->client->recognizeEduPaperOcrWithOptions($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 EduPaperOcr (Handwriting) Response', [
                'has_data' => isset($body['data']),
                'request_id' => $body['requestId'] ?? null,
                'code' => $body['code'] ?? null,
                'message' => $body['message'] ?? null,
                'body_keys' => array_keys($body ?? [])
            ]);
            
            // Extract recognized text
            $recognizedTexts = [];
            
            if (isset($body['data'])) {
                $data = is_string($body['data']) ? json_decode($body['data'], true) : $body['data'];
                
                // Extract content from data structure
                if (isset($data['content']) && is_string($data['content'])) {
                    // Simple content string
                    $recognizedTexts[] = [
                        'text' => $data['content'],
                        'confidence' => 1.0
                    ];
                } elseif (isset($data['prism_wordsInfo']) && is_array($data['prism_wordsInfo'])) {
                    // Detailed word-level information
                    $allWords = [];
                    $totalProb = 0;
                    $count = 0;
                    
                    foreach ($data['prism_wordsInfo'] as $wordInfo) {
                        if (isset($wordInfo['word'])) {
                            $allWords[] = $wordInfo['word'];
                            if (isset($wordInfo['prob'])) {
                                $totalProb += $wordInfo['prob'];
                                $count++;
                            }
                        }
                    }
                    
                    if (!empty($allWords)) {
                        $recognizedTexts[] = [
                            'text' => implode('', $allWords),
                            'confidence' => $count > 0 ? ($totalProb / $count) / 100 : 0.0
                        ];
                    }
                } elseif (isset($data['page_list']) && is_array($data['page_list'])) {
                    // Page-based structure (similar to EduPaperCut)
                    foreach ($data['page_list'] as $page) {
                        if (isset($page['prism_wordsInfo']) && is_array($page['prism_wordsInfo'])) {
                            $words = [];
                            foreach ($page['prism_wordsInfo'] as $wordInfo) {
                                if (isset($wordInfo['word'])) {
                                    $words[] = $wordInfo['word'];
                                }
                            }
                            if (!empty($words)) {
                                $recognizedTexts[] = [
                                    'text' => implode('', $words),
                                    'confidence' => 1.0
                                ];
                            }
                        }
                    }
                }
            }
            
            Log::info('Handwriting recognition result', [
                'texts_count' => count($recognizedTexts),
                'preview' => !empty($recognizedTexts) ? mb_substr($recognizedTexts[0]['text'], 0, 100) : 'N/A'
            ]);

            return [
                'raw' => $body,
                'texts' => $recognizedTexts,
                'type' => 'handwriting'
            ];

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