sbnews/server/app/common/service/chat/WhisperAsrService.php

<?php

namespace app\common\service\chat;

use app\common\model\ai\AiConfig;

class WhisperAsrService
{
    private const DEFAULT_BASE_URL = 'http://127.0.0.1:8791';
    private const DEFAULT_TIMEOUT = 120;
    private const DEFAULT_MAX_SIZE_MB = 10;
    private const ALLOWED_EXTENSIONS = ['mp3', 'm4a', 'aac', 'wav', 'amr', 'mp4', '3gp'];

    public static function transcribeUploadedFile($file): array
    {
        if ((int) AiConfig::getVal('whisper_asr_enabled', '1') !== 1) {
            return self::fail('语音转文字服务暂未启用');
        }
        if (empty($file)) {
            return self::fail('请先录制语音');
        }

        $filePath = self::uploadedFilePath($file);
        $originalName = self::uploadedFileName($file);
        $mime = self::uploadedFileMime($file);
        $maxSizeMb = self::maxSizeMb();

        $valid = self::validateAudioFile($filePath, $originalName, $maxSizeMb);
        if (empty($valid['success'])) {
            return $valid;
        }

        return self::requestAsr($filePath, $originalName, $mime);
    }

    public static function validateAudioFile(string $filePath, string $originalName = '', int $maxSizeMb = self::DEFAULT_MAX_SIZE_MB): array
    {
        if ($filePath === '' || !is_file($filePath) || !is_readable($filePath)) {
            return self::fail('录音文件不存在');
        }

        $maxSizeMb = max(1, min(100, $maxSizeMb));
        $size = (int) filesize($filePath);
        if ($size <= 0) {
            return self::fail('录音文件为空');
        }
        if ($size > $maxSizeMb * 1024 * 1024) {
            return self::fail('录音文件不能超过' . $maxSizeMb . 'MB');
        }

        $extension = self::resolveAudioExtension($originalName, $filePath);
        if ($extension === '') {
            return self::fail('无法识别录音文件格式');
        }
        if (!in_array($extension, self::ALLOWED_EXTENSIONS, true)) {
            return self::fail('仅支持 mp3、m4a、aac、wav、amr、mp4、3gp 格式语音');
        }

        return self::success([
            'extension' => $extension,
            'size' => $size,
            'max_size_mb' => $maxSizeMb,
        ]);
    }

    public static function normalizeAsrResponse(array $payload): array
    {
        $nested = is_array($payload['data'] ?? null) ? $payload['data'] : [];
        $text = trim((string) ($payload['text'] ?? $payload['result'] ?? $nested['text'] ?? ''));
        if ($text === '') {
            return self::fail('未识别到语音内容');
        }

        $durationMs = (int) ($payload['duration_ms'] ?? $nested['duration_ms'] ?? 0);
        if ($durationMs <= 0) {
            $duration = (float) ($payload['duration'] ?? $nested['duration'] ?? 0);
            $durationMs = $duration > 0 ? (int) round($duration * 1000) : 0;
        }

        return self::success([
            'text' => $text,
            'duration_ms' => max(0, $durationMs),
        ]);
    }

    private static function requestAsr(string $filePath, string $originalName, string $mime): array
    {
        if (!function_exists('curl_init')) {
            return self::fail('服务器未启用 cURL，无法连接语音识别服务');
        }

        $baseUrl = rtrim((string) AiConfig::getVal('whisper_asr_base_url', self::DEFAULT_BASE_URL), '/');
        if ($baseUrl === '') {
            return self::fail('语音转文字服务地址未配置');
        }

        $headers = ['Accept: application/json'];
        $apiToken = trim((string) AiConfig::getVal('whisper_asr_api_token', ''));
        if ($apiToken !== '') {
            $headers[] = 'Authorization: Bearer ' . $apiToken;
        }

        $timeout = self::timeoutSeconds();
        $startTime = microtime(true);
        $postFields = [
            'file' => new \CURLFile(
                $filePath,
                $mime !== '' ? $mime : 'application/octet-stream',
                self::asrUploadFileName($originalName, $filePath)
            ),
        ];

        $ch = curl_init($baseUrl . '/v1/transcribe');
        curl_setopt_array($ch, [
            CURLOPT_POST => true,
            CURLOPT_HTTPHEADER => $headers,
            CURLOPT_POSTFIELDS => $postFields,
            CURLOPT_RETURNTRANSFER => true,
            CURLOPT_TIMEOUT => $timeout,
            CURLOPT_CONNECTTIMEOUT => 5,
            CURLOPT_SSL_VERIFYPEER => false,
        ]);

        $response = curl_exec($ch);
        $httpCode = (int) curl_getinfo($ch, CURLINFO_HTTP_CODE);
        $error = curl_error($ch);
        curl_close($ch);

        $costMs = (int) ((microtime(true) - $startTime) * 1000);
        if ($error !== '') {
            return self::fail('语音识别服务连接失败：' . $error, ['duration_ms' => $costMs]);
        }

        $payload = json_decode((string) $response, true);
        if (!is_array($payload)) {
            return self::fail('语音识别服务返回异常：响应不是有效JSON', ['duration_ms' => $costMs]);
        }
        if ($httpCode !== 200) {
            $message = $payload['detail'] ?? $payload['message'] ?? $payload['error'] ?? ('HTTP ' . $httpCode);
            return self::fail('语音识别服务返回异常：' . (is_string($message) ? $message : json_encode($message, JSON_UNESCAPED_UNICODE)), ['duration_ms' => $costMs]);
        }

        $normalized = self::normalizeAsrResponse($payload);
        if (empty($normalized['success'])) {
            return $normalized;
        }
        if (empty($normalized['data']['duration_ms'])) {
            $normalized['data']['duration_ms'] = $costMs;
        }
        return $normalized;
    }

    private static function uploadedFilePath($file): string
    {
        foreach (['getRealPath', 'getPathname'] as $method) {
            if (is_object($file) && method_exists($file, $method)) {
                $path = (string) $file->{$method}();
                if ($path !== '') {
                    return $path;
                }
            }
        }
        return '';
    }

    private static function uploadedFileName($file): string
    {
        if (is_object($file) && method_exists($file, 'getOriginalName')) {
            return (string) $file->getOriginalName();
        }
        return '';
    }

    private static function uploadedFileMime($file): string
    {
        if (is_object($file) && method_exists($file, 'getMime')) {
            return (string) $file->getMime();
        }
        return '';
    }

    private static function resolveAudioExtension(string $originalName, string $filePath): string
    {
        $extension = strtolower((string) pathinfo($originalName, PATHINFO_EXTENSION));
        if ($extension === '') {
            $extension = strtolower((string) pathinfo($filePath, PATHINFO_EXTENSION));
        }
        return $extension;
    }

    private static function asrUploadFileName(string $originalName, string $filePath): string
    {
        if ($originalName !== '' && pathinfo($originalName, PATHINFO_EXTENSION) !== '') {
            return $originalName;
        }

        $extension = self::resolveAudioExtension($originalName, $filePath);
        return 'voice.' . ($extension !== '' ? $extension : 'mp3');
    }

    private static function timeoutSeconds(): int
    {
        return max(5, min(300, (int) AiConfig::getVal('whisper_asr_timeout', (string) self::DEFAULT_TIMEOUT)));
    }

    private static function maxSizeMb(): int
    {
        return max(1, min(100, (int) AiConfig::getVal('whisper_asr_max_size_mb', (string) self::DEFAULT_MAX_SIZE_MB)));
    }

    private static function success(array $data): array
    {
        return ['success' => true, 'data' => $data];
    }

    private static function fail(string $error, array $data = []): array
    {
        return ['success' => false, 'error' => $error, 'data' => $data];
    }
}