Tolan/backend/analyze_current_log.py

#!/usr/bin/env python3
"""
分析当前的日志片段
从终端选中的日志中分析性能
"""

# 当前终端日志
log_content = """
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] 开始接收音频数据
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] 启动流式STT识别
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] STT连接已打开
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] STT识别会话已启动
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] STT实时结果: 您
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] STT实时结果: 你好，你

[85c570c0-2346-44d2-8a6b-49aed1aa0b67] 音频录制结束
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] 停止STT识别
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] STT实时结果: 你好，你好，今天
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] STT实时结果: 你好你好，今天天气怎么样？
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] STT连接已关闭
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] STT识别已停止
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] STT识别结果: 你好你好，今天天气怎么样？
[85c570c0-2346-44d2-8a6b-49aed1aa0b67] 步骤 2/3: LLM 生成回复

========== LLM 开始 ==========
用户输入: 你好你好，今天天气怎么样？
调用 LLM: qwen-plus
LLM 原始响应: Hello! It's great to hear from you! I can't check the actual weather, but if you tell me where you are, I can help describe what the weather might feel like or how to talk about it in English! ☀️🌧️⛅  
For example, you could say: "It's sunny and warm today!" or "It's a bit cloudy but nice outside."  

Keep up the good work practicing English!
[EMOTION:encouraging]
回复文本: Hello! It's great to hear from you! I can't check the actual weather, but if you tell me where you are, I can help describe what the weather might feel like or how to talk about it in English! ☀️🌧️⛅  
For example, you could say: "It's sunny and warm today!" or "It's a bit cloudy but nice outside."  

Keep up the good work practicing English!
情感: encouraging
========== LLM 成功 ==========

[85c570c0-2346-44d2-8a6b-49aed1aa0b67] 步骤 3/3: TTS 文本转语音

========== TTS 开始 ==========
要合成的文本: Hello! It's great to hear from you! I can't check the actual weather, but if you tell me where you are, I can help describe what the weather might feel like or how to talk about it in English! ☀️🌧️⛅  
For example, you could say: "It's sunny and warm today!" or "It's a bit cloudy but nice outside."  

Keep up the good work practicing English!
情感: encouraging
使用音色: Cherry
TTS模型: qwen3-tts-flash
调用 DashScope TTS API...
TTS API 响应类型: <class 'dashscope.api_entities.dashscope_response.MultiModalConversationResponse'>
output 类型: <class 'dashscope.api_entities.dashscope_response.TextToSpeechOutput'>
finish_reason: stop
从URL下载音频: http://dashscope-result-bj.oss-cn-beijing.aliyuncs.com/1d/a0/20251015/d2033070/6ecda767-06ca-45fa-a312-fc391c32265c.wav?Expires=1760609218&OSSAccessKeyId=LTAI5tKPD3TMqf2Lna1fASuh&Signature=1%2FJLVFPvrvh2rqUTv2YZEQSeR6g%3D
音频下载成功，大小: 1105964 字节
将分为 346 个数据块发送
========== TTS 成功 ==========

[85c570c0-2346-44d2-8a6b-49aed1aa0b67] 请求处理完成

[85c570c0-2346-44d2-8a6b-49aed1aa0b67] 资源已清理

[72ee5620-33f3-4d99-b56c-bbf17bf186d7] 开始接收音频数据
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] 启动流式STT识别
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] STT连接已打开
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] STT识别会话已启动
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] STT实时结果: 你
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] STT实时结果: 你好，你
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] STT实时结果: 你好，你好。你今

[72ee5620-33f3-4d99-b56c-bbf17bf186d7] 音频录制结束
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] 停止STT识别
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] STT实时结果: 你好，你好。你今年多
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] STT实时结果: 你好你好，你今年多大了？
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] STT连接已关闭
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] STT识别已停止
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] STT识别结果: 你好你好，你今年多大了？
[72ee5620-33f3-4d99-b56c-bbf17bf186d7] 步骤 2/3: LLM 生成回复

========== LLM 开始 ==========
用户输入: 你好你好，你今年多大了？
调用 LLM: qwen-plus
LLM 原始响应: 你好呀！我是AI助手，没有具体的年龄哦～但我很年轻有活力呢！✨ 很高兴认识你！
[EMOTION:happy]
回复文本: 你好呀！我是AI助手，没有具体的年龄哦～但我很年轻有活力呢！✨ 很高兴认识你！
情感: happy
========== LLM 成功 ==========

[72ee5620-33f3-4d99-b56c-bbf17bf186d7] 步骤 3/3: TTS 文本转语音

========== TTS 开始 ==========
要合成的文本: 你好呀！我是AI助手，没有具体的年龄哦～但我很年轻有活力呢！✨ 很高兴认识你！
情感: happy
使用音色: Cherry
TTS模型: qwen3-tts-flash
调用 DashScope TTS API...
TTS API 响应类型: <class 'dashscope.api_entities.dashscope_response.MultiModalConversationResponse'>
output 类型: <class 'dashscope.api_entities.dashscope_response.TextToSpeechOutput'>
finish_reason: stop
从URL下载音频: http://dashscope-result-bj.oss-cn-beijing.aliyuncs.com/1d/aa/20251015/d2033070/e32c33e7-91df-4088-8fa2-e998898e82d2.wav?Expires=1760609249&OSSAccessKeyId=LTAI5tKPD3TMqf2Lna1fASuh&Signature=I%2F2MWpKZf5piIIg%2B%2BPoy%2BWSCmwM%3D
音频下载成功，大小: 345644 字节
将分为 109 个数据块发送
========== TTS 成功 ==========

[72ee5620-33f3-4d99-b56c-bbf17bf186d7] 请求处理完成

[72ee5620-33f3-4d99-b56c-bbf17bf186d7] 资源已清理

[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] 开始接收音频数据
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] 启动流式STT识别
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] STT连接已打开
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] STT识别会话已启动
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] STT实时结果: 你
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] STT实时结果: 您好，你好

[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] 音频录制结束
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] 停止STT识别
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] STT实时结果: 您好，你好。今天星
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] STT实时结果: 您好，你好，今天星期几？
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] STT实时结果: 你好你好，今天星期几呀？
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] STT连接已关闭
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] STT识别已停止
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] STT识别结果: 你好你好，今天星期几呀？
[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] 步骤 2/3: LLM 生成回复

========== LLM 开始 ==========
用户输入: 你好你好，今天星期几呀？
调用 LLM: qwen-plus
LLM 原始响应: Hello! Today is Tuesday. How can I help you today?  
[EMOTION:happy]
回复文本: Hello! Today is Tuesday. How can I help you today?
情感: happy
========== LLM 成功 ==========

[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] 步骤 3/3: TTS 文本转语音

========== TTS 开始 ==========
要合成的文本: Hello! Today is Tuesday. How can I help you today?
情感: happy
使用音色: Cherry
TTS模型: qwen3-tts-flash
调用 DashScope TTS API...
TTS API 响应类型: <class 'dashscope.api_entities.dashscope_response.MultiModalConversationResponse'>
output 类型: <class 'dashscope.api_entities.dashscope_response.TextToSpeechOutput'>
finish_reason: stop
从URL下载音频: http://dashscope-result-sh.oss-cn-shanghai.aliyuncs.com/1d/8e/20251015/e9ee724c/038ed2c1-7065-4e17-9b3f-07f069128486.wav?Expires=1760609263&OSSAccessKeyId=LTAI5tKPD3TMqf2Lna1fASuh&Signature=Lil%2FuwlWgpGA3sKvxFqIrYO7nq4%3D
音频下载成功，大小: 222764 字节
将分为 70 个数据块发送
========== TTS 成功 ==========

[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] 请求处理完成

[068fdf74-2c2d-48ee-8455-72d5c73cc9bf] 资源已清理
"""

import re
from datetime import datetime


def analyze_manual_log():
    """手动分析当前日志（无时间戳版本）"""
    print("\n" + "=" * 70)
    print("当前日志分析 (基于日志内容估算)".center(70))
    print("=" * 70)
    
    # 按请求ID分组
    requests = {}
    current_request = None
    
    lines = log_content.strip().split('\n')
    
    for line in lines:
        # 提取请求ID
        match = re.search(r'\[([a-f0-9\-]{36})\]', line)
        if match:
            request_id = match.group(1)
            if request_id not in requests:
                requests[request_id] = {
                    'id': request_id[:8],
                    'user_input': '',
                    'response': '',
                    'emotion': '',
                    'audio_size': 0,
                    'chunks': 0,
                    'lines': []
                }
            current_request = requests[request_id]
            current_request['lines'].append(line)
        elif current_request and ('用户输入:' in line or '要合成的文本:' in line or '情感:' in line):
            current_request['lines'].append(line)
        
        # 提取关键信息
        if '用户输入:' in line:
            current_request['user_input'] = line.split('用户输入:')[1].strip()
        elif '回复文本:' in line:
            current_request['response'] = line.split('回复文本:')[1].strip()
        elif re.match(r'^情感:', line):
            current_request['emotion'] = line.split('情感:')[1].strip()
        elif '音频下载成功，大小:' in line:
            size_match = re.search(r'(\d+)\s*字节', line)
            if size_match:
                current_request['audio_size'] = int(size_match.group(1))
        elif '将分为' in line and '个数据块发送' in line:
            chunks_match = re.search(r'(\d+)\s*个数据块', line)
            if chunks_match:
                current_request['chunks'] = int(chunks_match.group(1))
    
    # 打印每个请求的分析
    print(f"\n发现 {len(requests)} 个完整请求\n")
    
    total_audio_size = 0
    total_chunks = 0
    
    for i, req in enumerate(requests.values(), 1):
        print(f"\n{'=' * 70}")
        print(f"请求 #{i} [{req['id']}...]")
        print(f"{'=' * 70}")
        print(f"📝 用户输入: {req['user_input'][:60]}..." if len(req['user_input']) > 60 else f"📝 用户输入: {req['user_input']}")
        print(f"💬 AI回复: {req['response'][:60]}..." if len(req['response']) > 60 else f"💬 AI回复: {req['response']}")
        print(f"😊 情感: {req['emotion']}")
        print(f"🎵 音频大小: {req['audio_size']:,} 字节 ({req['audio_size']/1024:.1f} KB)")
        print(f"📦 数据块数: {req['chunks']}")
        
        total_audio_size += req['audio_size']
        total_chunks += req['chunks']
        
        # 流程分析
        print(f"\n流程分析:")
        print(f"  ✓ STT: 流式识别成功")
        print(f"  ✓ LLM: 生成回复成功")
        print(f"  ✓ TTS: 合成音频成功")
    
    # 汇总统计
    print(f"\n{'=' * 70}")
    print("汇总统计".center(70))
    print(f"{'=' * 70}")
    print(f"总请求数: {len(requests)}")
    print(f"总音频大小: {total_audio_size:,} 字节 ({total_audio_size/1024:.1f} KB)")
    print(f"总数据块数: {total_chunks}")
    print(f"平均音频大小: {total_audio_size/len(requests):,.0f} 字节 ({total_audio_size/len(requests)/1024:.1f} KB)")
    print(f"平均数据块数: {total_chunks/len(requests):.0f}")
    
    print(f"\n{'=' * 70}")
    print("⚠️  注意事项".center(70))
    print(f"{'=' * 70}")
    print("当前日志没有时间戳，无法统计各阶段精确耗时。")
    print("请重新运行测试，使用更新后的代码，将获得详细的性能统计。")
    print(f"\n新代码已添加以下时间统计:")
    print("  - STT 识别耗时")
    print("  - LLM 生成耗时")
    print("  - TTS 合成耗时")
    print("  - 总体处理耗时")
    print(f"\n重新测试后，运行以下命令分析:")
    print("  python analyze_performance.py backend.log")
    print(f"{'=' * 70}\n")


if __name__ == "__main__":
    analyze_manual_log()