Ai_GirlFriend/test_real_audio.py

#!/usr/bin/env python3
"""
使用真实音频文件测试 DashScope ASR
"""
import sys
import os
sys.path.append('.')

import dashscope
from dashscope.audio.asr import Transcription
from lover.config import settings
from lover.oss_utils import upload_audio_file, delete_audio_file
import logging
import wave
import struct

# 设置日志
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

def create_test_wav():
    """创建一个简单的 WAV 测试文件"""
    # 创建 1 秒的 16kHz 单声道 WAV 文件
    sample_rate = 16000
    duration = 1  # 1 秒
    frequency = 440  # A4 音符
    
    # 生成正弦波
    samples = []
    import math
    for i in range(sample_rate * duration):
        t = i / sample_rate
        # 生成简单的正弦波，幅度控制在合理范围内
        sample = int(16000 * math.sin(2 * math.pi * frequency * t))
        # 确保在 16-bit 范围内
        sample = max(-32767, min(32767, sample))
        samples.append(sample)
    
    # 写入 WAV 文件
    wav_file = "test_audio.wav"
    with wave.open(wav_file, 'wb') as wav:
        wav.setnchannels(1)  # 单声道
        wav.setsampwidth(2)  # 16-bit
        wav.setframerate(sample_rate)  # 16kHz
        
        # 写入样本数据
        for sample in samples:
            wav.writeframes(struct.pack('<h', sample))
    
    logger.info(f"创建测试 WAV 文件: {wav_file}")
    return wav_file

def test_real_audio_asr():
    """使用真实音频文件测试 ASR"""
    
    # 设置 API Key
    dashscope.api_key = settings.DASHSCOPE_API_KEY
    logger.info(f"DashScope API Key: {settings.DASHSCOPE_API_KEY[:10]}***")
    
    # 创建测试 WAV 文件
    wav_file = create_test_wav()
    
    try:
        # 读取 WAV 文件
        with open(wav_file, 'rb') as f:
            wav_data = f.read()
        
        logger.info(f"WAV 文件大小: {len(wav_data)} 字节")
        
        # 上传到 OSS
        logger.info("上传 WAV 文件到 OSS...")
        file_url = upload_audio_file(wav_data, "wav")
        logger.info(f"上传成功: {file_url}")
        
        # 调用 DashScope ASR
        logger.info("调用 DashScope ASR...")
        
        task_response = Transcription.async_call(
            model='paraformer-v2',
            file_urls=[file_url],
            parameters={
                'format': 'wav',  # 使用 WAV 格式
                'sample_rate': 16000,
                'enable_words': False
            }
        )
        
        logger.info(f"ASR 任务响应: status_code={task_response.status_code}")
        
        if task_response.status_code == 200:
            task_id = task_response.output.task_id
            logger.info(f"任务创建成功: {task_id}")
            
            # 等待结果
            logger.info("等待 ASR 结果...")
            try:
                result = Transcription.wait(task=task_id)
                logger.info(f"ASR 结果: status_code={result.status_code}")
                
                if result.status_code == 200:
                    logger.info(f"任务状态: {result.output.task_status}")
                    if result.output.task_status == "FAILED":
                        logger.info(f"失败原因: {getattr(result.output, 'code', 'Unknown')}")
                        logger.info(f"失败消息: {getattr(result.output, 'message', 'Unknown')}")
                    elif result.output.task_status == "SUCCEEDED":
                        logger.info("✅ ASR 识别成功！")
                        logger.info(f"结果: {result.output}")
                else:
                    logger.error(f"获取结果失败: {result.status_code}")
                    
            except Exception as wait_error:
                logger.error(f"等待结果失败: {wait_error}")
        else:
            logger.error(f"任务创建失败: {task_response.status_code}")
        
        # 清理文件
        logger.info("清理文件...")
        delete_audio_file(file_url)
        os.remove(wav_file)
        
        return True
        
    except Exception as e:
        logger.error(f"测试失败: {e}")
        import traceback
        logger.error(f"错误堆栈: {traceback.format_exc()}")
        
        # 清理文件
        try:
            os.remove(wav_file)
        except:
            pass
        
        return False

if __name__ == "__main__":
    success = test_real_audio_asr()
    if success:
        logger.info("🎉 真实音频 ASR 测试完成")
    else:
        logger.error("💥 真实音频 ASR 测试失败")
        sys.exit(1)