# 安装最新版本
pip install zai-sdk
# 或指定版本
pip install zai-sdk==0.0.1

import zai
print(zai.__version__)

import wave
import base64
from zai import ZhipuAiClient

def save_audio_as_wav(audio_data, filepath):
    """保存音频数据为WAV文件（模型返回的语音用）"""
    with wave.open(filepath, 'wb') as wav_file:
        wav_file.setnchannels(1)
        wav_file.setsampwidth(2)
        wav_file.setframerate(44100)
        wav_file.writeframes(audio_data)
    print(f"Audio saved to {filepath}")

def get_base64_from_wav(wav_path):
    """将WAV文件转为Base64编码字符串"""
    with open(wav_path, "rb") as f:
        audio_bytes = f.read()
    return base64.b64encode(audio_bytes).decode("utf-8")

client = ZhipuAiClient(api_key="your_api_key")  # 请填写您自己的APIKey

input_wav_path = "your_voice.wav"  # 你的WAV文件路径
base64_voice = get_base64_from_wav(input_wav_path)

response = client.chat.completions.create(
    model="glm-4-voice",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "text",
                    "text": "你好，这是我的语音输入测试，请慢速复述一遍"
                },
                {
                    "type": "input_audio",
                    "input_audio": {
                        "data": base64_voice,
                        "format": "wav"
                    }
                }
            ]
        }
    ],
    stream=False
)

print(response.choices[0].message.content)

# 解析并保存模型返回的语音
try:
    audio_data = response.choices[0].message.audio['data']
    decoded_data = base64.b64decode(audio_data)
    save_audio_as_wav(decoded_data, "output.wav")
except Exception as e:
    print("处理音频失败：", e)