curl --location --request POST 'http://prod-cn.your-api-server.com/v1/responses' \
--header 'Accept: application/json' \
--header 'Authorization: Bearer {{YOUR_API_KEY}}' \
--header 'Content-Type: application/json' \
--data-raw '{
    "model": "gpt-5",
    "stream": true,
    "messages": [
        {
            "role": "user",
            "content": "Write a one-sentence bedtime story about a unicorn."
        }
    ],
    "tools": ["format_story_output"],
    "tool_choice": {
        "type": "function",
        "function": {"name": "format_story_output"}
    },
    "temperature": 0.6,
    "max_tokens": 512
}'

import requests
import json

# 1. 配置基础参数
API_URL = "http://prod-cn.your-api-server.com/v1/responses"
API_KEY = "Bearer {{YOUR_API_KEY}}"  # 替换为实际Token
HEADERS = {
    "Content-Type": "application/json",
    "Accept": "application/json",
    "Authorization": API_KEY
}

# 2. 构造请求体
PAYLOAD = {
    "model": "gpt-5",
    "stream": True,  # 启用流式返回
    "messages": [
        {"role": "user", "content": "Write a one-sentence bedtime story about a unicorn."}
    ],
    "tools": ["format_story_output"],
    "tool_choice": {
        "type": "function",
        "function": {"name": "format_story_output"}
    },
    "temperature": 0.6,
    "max_tokens": 512
}

# 3. 发送请求并处理流式响应
try:
    response = requests.post(
        API_URL,
        headers=HEADERS,
        json=PAYLOAD,
        stream=True  # 关键：启用流式接收
    )
    response.raise_for_status()  # 捕获HTTP错误（如401授权失败）

    # 迭代处理每一条流式数据
    for chunk in response.iter_lines(decode_unicode=True):
        if chunk:
            # 检查是否为结束信号
            if chunk == "data: [DONE]":
                print("\n流式返回结束")
                break
            # 解析JSON数据（需去除前缀"data: "，若存在）
            if chunk.startswith("data: "):
                chunk = chunk[6:]
            data = json.loads(chunk)
            # 提取助手回复内容
            if "choices" in data and len(data["choices"]) > 0:
                content = data["choices"][0]["message"].get("content", "")
                if content:
                    print(content, end="", flush=True)  # 实时输出

    # 解析最终使用的Token统计
    usage = data.get("usage", {})
    print(f"\nToken使用统计：")
    print(f"请求Token: {usage.get('prompt_tokens', 0)}")
    print(f"响应Token: {usage.get('completion_tokens', 0)}")
    print(f"总Token: {usage.get('total_tokens', 0)}")

except requests.exceptions.RequestException as e:
    print(f"请求异常：{e}")
except json.JSONDecodeError as e:
    print(f"JSON解析异常：{e}")