OpenAI 重组基础设施团队 核心数据中心高管离职

pip install openai requests

import time
import logging
import requests

# 配置日志输出
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')

class ResilientLLMClient:
    def __init__(self, api_key, base_url="https://api.n1n.ai/v1"):
        self.api_key = api_key
        self.base_url = base_url
        self.headers = {
            "Authorization": f"Bearer {api_key}",
            "Content-Type": "application/json"
        }

    def generate_completion(self, messages, model_fallback_list, temperature=0.7, max_tokens=1000):
        """
        顺序尝试调用模型列表，实现自动容灾切换
        """
        for model in model_fallback_list:
            logging.info(f"正在尝试调用模型: {model}")
            payload = {
                "model": model,
                "messages": messages,
                "temperature": temperature,
                "max_tokens": max_tokens
            }

            start_time = time.time()
            try:
                # 设置 15 秒的严格超时时间，防止因单个通道挂起导致业务长时间阻塞
                response = requests.post(
                    f"{self.base_url}/chat/completions",
                    json=payload,
                    headers=self.headers,
                    timeout=15.0
                )

                # 检查 HTTP 状态码
                if response.status_code == 200:
                    duration = time.time() - start_time
                    logging.info(f"成功使用 {model} 生成响应，耗时: {duration:.2f} 秒")
                    return response.json()
                else:
                    logging.warning(f"模型 {model} 返回异常状态码 {response.status_code}: {response.text}")

            except requests.exceptions.Timeout:
                logging.error(f"调用模型 {model} 超时 (响应时间 > 15秒)")
            except requests.exceptions.RequestException as e:
                logging.error(f"调用模型 {model} 时发生网络错误: {str(e)}")

            # 当前模型失败，准备进入下一次循环尝试备用模型
            logging.info(f"正在切换至备用路由管线中的下一个模型...")

        raise RuntimeError("备用管道中的所有 LLM 供应商均未能成功响应。")

# 使用示例
if __name__ == "__main__":
    # 请将此处替换为您的 n1n.ai API 密钥
    N1N_API_KEY = "your_n1n_api_key_here"

    client = ResilientLLMClient(api_key=N1N_API_KEY)

    user_messages = [
        {"role": "system", "content": "你是一位资深的系统架构师。"},
        {"role": "user", "content": "用少于 100 字简述 InfiniBand 和 RoCE v2 的区别。"}
    ]

    # 定义优先级模型列表
    # 当 GPT-4o 发生故障时，自动切换至 Claude 3.5 Sonnet，最后切换至 DeepSeek-V3
    models_pipeline = [
        "openai/gpt-4o",
        "anthropic/claude-3.5-sonnet",
        "deepseek/deepseek-v3"
    ]

    try:
        result = client.generate_completion(user_messages, models_pipeline)
        print("\n--- LLM 返回结果 ---")
        print(result['choices'][0]['message']['content'])
    except Exception as e:
        print(f"执行 LLM 容灾管线失败: {str(e)}")