src/model_interface.py

# -*- coding: utf-8 -*-
"""
模型接口模块
用于与两个大模型API进行交互的封装
"""

import os
import sys
import requests
import json
import time
import logging
import random
from typing import Dict, Any, Optional, Callable
from urllib3.exceptions import InsecureRequestWarning

# 在代码开头强制设置终端编码为UTF-8（仅在Windows执行）
if os.name == 'nt':
    os.system('chcp 65001 > nul')

# 获取当前脚本文件所在目录的绝对路径
current_script_dir = os.path.dirname(os.path.abspath(__file__))

# 项目根目录 (即 '20250907_大模型辩论' 目录, 是 'src' 的上一级)
project_root = os.path.dirname(current_script_dir)

# 定义数据输入、输出和日志目录
DATA_DIR = os.path.join(project_root, 'data')
OUTPUT_DIR = os.path.join(project_root, 'output')
LOGS_DIR = os.path.join(project_root, 'logs')

# 确保目录存在
os.makedirs(OUTPUT_DIR, exist_ok=True)
os.makedirs(LOGS_DIR, exist_ok=True)

# 配置日志
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
    handlers=[
        logging.FileHandler(os.path.join(LOGS_DIR, '模型接口日志.log'), encoding='utf-8'),
        logging.StreamHandler(sys.stdout)
    ]
)
logger = logging.getLogger(__name__)

# 捕获警告并记录到日志
logging.captureWarnings(True)

class ModelInterface:
    """模型接口基类"""
    
    def __init__(self, api_key: str, base_url: str):
        self.api_key = api_key
        self.base_url = base_url
        self.headers = {
            'Authorization': f'Bearer {api_key}',
            'Content-Type': 'application/json'
        }
        logger.info(f"模型接口初始化完成，基础URL: {base_url}")
    
    def _should_retry(self, status_code: Optional[int]) -> bool:
        return status_code == 429 or (status_code is not None and 500 <= status_code < 600)
    
    def _compute_backoff_seconds(self, attempt: int, retry_after: Optional[str]) -> float:
        # 优先使用服务端的 Retry-After
        if retry_after:
            try:
                return float(retry_after)
            except Exception:
                pass
        # 指数退避 + 抖动：1, 2, 4, 8 ... 上限10，并加入 0~300ms 抖动
        base = min(10, 2 ** attempt)
        return base + random.random() * 0.3
    
    def send_request(self, model: str, messages: list, **kwargs) -> Dict[str, Any]:
        payload = {
            'model': model,
            'messages': messages,
            **kwargs
        }
        max_retries = 3
        last_exc = None
        for attempt in range(max_retries + 1):
            try:
                logger.info(f"向模型 {model} 发送请求")
                response = requests.post(
                    f"{self.base_url}/chat/completions",
                    headers=self.headers,
                    json=payload,
                    timeout=60
                )
                if self._should_retry(response.status_code):
                    wait_s = self._compute_backoff_seconds(attempt, response.headers.get('Retry-After'))
                    logger.warning(f"模型 {model} 返回 {response.status_code}，{wait_s:.2f}s 后重试 (attempt={attempt})")
                    time.sleep(wait_s)
                    continue
                response.raise_for_status()
                result = response.json()
                logger.info(f"模型 {model} 响应成功")
                return result
            except requests.exceptions.RequestException as e:
                last_exc = e
                status = getattr(e.response, 'status_code', None)
                if self._should_retry(status) and attempt < max_retries:
                    wait_s = self._compute_backoff_seconds(attempt, getattr(e.response, 'headers', {}).get('Retry-After'))
                    logger.warning(f"请求异常 {status}，{wait_s:.2f}s 后重试 (attempt={attempt})")
                    time.sleep(wait_s)
                    continue
                logger.error(f"请求模型 {model} 失败: {str(e)}")
                raise
        # 如果走到这里，说明重试仍失败
        if last_exc:
            raise last_exc
    
    def send_stream_request(self, model: str, messages: list, callback: Callable[[str], None], **kwargs) -> str:
        payload = {
            'model': model,
            'messages': messages,
            'stream': True,
            **kwargs
        }
        max_retries = 3
        last_exc = None
        for attempt in range(max_retries + 1):
            full_response = ""
            try:
                logger.info(f"向模型 {model} 发送流式请求")
                response = requests.post(
                    f"{self.base_url}/chat/completions",
                    headers=self.headers,
                    json=payload,
                    timeout=60,
                    stream=True
                )
                if self._should_retry(response.status_code):
                    wait_s = self._compute_backoff_seconds(attempt, response.headers.get('Retry-After'))
                    logger.warning(f"模型 {model} 流式返回 {response.status_code}，{wait_s:.2f}s 后重试 (attempt={attempt})")
                    time.sleep(wait_s)
                    continue
                response.raise_for_status()
                # 处理流式响应
                for line in response.iter_lines():
                    if line:
                        decoded_line = line.decode('utf-8')
                        if decoded_line.startswith("data: "):
                            data = decoded_line[6:]
                            if data != "[DONE]":
                                try:
                                    json_data = json.loads(data)
                                    content = json_data["choices"][0]["delta"].get("content", "")
                                    if content:
                                        full_response += content
                                        callback(content)
                                except json.JSONDecodeError:
                                    pass
                logger.info(f"模型 {model} 流式响应完成")
                return full_response
            except requests.exceptions.RequestException as e:
                last_exc = e
                status = getattr(e.response, 'status_code', None)
                if self._should_retry(status) and attempt < max_retries:
                    wait_s = self._compute_backoff_seconds(attempt, getattr(e.response, 'headers', {}).get('Retry-After'))
                    logger.warning(f"流式请求异常 {status}，{wait_s:.2f}s 后重试 (attempt={attempt})")
                    time.sleep(wait_s)
                    continue
                logger.error(f"流式请求模型 {model} 失败: {str(e)}")
                raise
            except Exception as e:
                last_exc = e
                logger.error(f"处理流式响应时出错: {str(e)}")
                raise
        if last_exc:
            raise last_exc
    
    def get_response_text(self, response: Dict[str, Any]) -> str:
        try:
            return response['choices'][0]['message']['content']
        except (KeyError, IndexError) as e:
            logger.error(f"提取回复文本失败: {str(e)}")
            raise

class GLM45Interface(ModelInterface):
    """GLM-4.5模型接口"""
    
    def __init__(self, api_key: str):
        super().__init__(api_key, "https://api-inference.modelscope.cn/v1")
        self.model_name = "ZhipuAI/GLM-4.5"
        logger.info(f"GLM-4.5模型接口初始化完成")
    
    def chat(self, messages: list, temperature: float = 0.7, max_tokens: int = 8000) -> str:
        try:
            response = self.send_request(
                model=self.model_name,
                messages=messages,
                temperature=temperature,
                max_tokens=max_tokens
            )
            return self.get_response_text(response)
        except Exception as e:
            logger.error(f"GLM-4.5对话失败: {str(e)}")
            return f"GLM-4.5对话失败: {str(e)}"
    
    def chat_stream(self, messages: list, callback: Callable[[str], None], temperature: float = 0.7, max_tokens: int = 8000) -> str:
        try:
            return self.send_stream_request(
                model=self.model_name,
                messages=messages,
                callback=callback,
                temperature=temperature,
                max_tokens=max_tokens
            )
        except Exception as e:
            logger.error(f"GLM-4.5流式对话失败: {str(e)}")
            error_msg = f"GLM-4.5流式对话失败: {str(e)}"
            callback(f"\n{error_msg}\n")
            return error_msg

class DeepSeekV31Interface(ModelInterface):
    """DeepSeek-V3.1模型接口"""
    
    def __init__(self, api_key: str):
        super().__init__(api_key, "https://api-inference.modelscope.cn/v1")
        self.model_name = "deepseek-ai/DeepSeek-V3.1"
        logger.info(f"DeepSeek-V3.1模型接口初始化完成")
    
    def chat(self, messages: list, temperature: float = 0.7, max_tokens: int = 8000) -> str:
        try:
            response = self.send_request(
                model=self.model_name,
                messages=messages,
                temperature=temperature,
                max_tokens=max_tokens
            )
            return self.get_response_text(response)
        except Exception as e:
            logger.error(f"DeepSeek-V3.1对话失败: {str(e)}")
            return f"DeepSeek-V3.1对话失败: {str(e)}"
    
    def chat_stream(self, messages: list, callback: Callable[[str], None], temperature: float = 0.7, max_tokens: int = 8000) -> str:
        try:
            return self.send_stream_request(
                model=self.model_name,
                messages=messages,
                callback=callback,
                temperature=temperature,
                max_tokens=max_tokens 
            )
        except Exception as e:
            logger.error(f"DeepSeek-V3.1流式对话失败: {str(e)}")
            error_msg = f"DeepSeek-V3.1流式对话失败: {str(e)}"
            callback(f"\n{error_msg}\n")
            return error_msg

class QwenInterface(ModelInterface):
    """Qwen模型接口"""
    
    def __init__(self, api_key: str):
        super().__init__(api_key, "https://api-inference.modelscope.cn/v1")
        self.model_name = "Qwen/Qwen3-235B-A22B-Thinking-2507"
        logger.info(f"Qwen模型接口初始化完成")

    def chat(self, messages: list, temperature: float = 0.7, max_tokens: int = 8000) -> str:
        try:
            response = self.send_request(
                model=self.model_name,
                messages=messages,
                temperature=temperature,
                max_tokens=max_tokens
            )
            return self.get_response_text(response)
        except Exception as e:
            logger.error(f"Qwen对话失败: {str(e)}")
            return f"Qwen对话失败: {str(e)}"

    def chat_stream(self, messages: list, callback: Callable[[str], None], temperature: float = 0.7, max_tokens: int = 8000) -> str:
        try:
            return self.send_stream_request(
                model=self.model_name,
                messages=messages,
                callback=callback,
                temperature=temperature,
                max_tokens=max_tokens
            )
        except Exception as e:
            logger.error(f"Qwen流式对话失败: {str(e)}")
            error_msg = f"Qwen流式对话失败: {str(e)}"
            callback(f"\n{error_msg}\n")
            return error_msg

# 新增：Qwen Instruct 非思考版
class QwenInstructInterface(ModelInterface):
    """Qwen Instruct模型接口（非思考版）"""
    
    def __init__(self, api_key: str):
        super().__init__(api_key, "https://api-inference.modelscope.cn/v1")
        self.model_name = "Qwen/Qwen3-235B-A22B-Instruct-2507"
        logger.info(f"Qwen Instruct模型接口初始化完成")

    def chat(self, messages: list, temperature: float = 0.7, max_tokens: int = 8000) -> str:
        try:
            response = self.send_request(
                model=self.model_name,
                messages=messages,
                temperature=temperature,
                max_tokens=max_tokens
            )
            return self.get_response_text(response)
        except Exception as e:
            logger.error(f"Qwen Instruct对话失败: {str(e)}")
            return f"Qwen Instruct对话失败: {str(e)}"

    def chat_stream(self, messages: list, callback: Callable[[str], None], temperature: float = 0.7, max_tokens: int = 8000) -> str:
        try:
            return self.send_stream_request(
                model=self.model_name,
                messages=messages,
                callback=callback,
                temperature=temperature,
                max_tokens=max_tokens
            )
        except Exception as e:
            logger.error(f"Qwen Instruct流式对话失败: {str(e)}")
            error_msg = f"Qwen Instruct流式对话失败: {str(e)}"
            callback(f"\n{error_msg}\n")
            return error_msg

class ModelManager:
    """模型管理器，统一管理两个模型接口"""
    
    def __init__(self, api_key: str):
        self.api_key = api_key
        self.glm45 = GLM45Interface(api_key)
        self.deepseek_v31 = DeepSeekV31Interface(api_key)
        self.qwen = QwenInterface(api_key)
        self.qwen_instruct = QwenInstructInterface(api_key)
        logger.info("模型管理器初始化完成")
    
    def get_model(self, model_name: str) -> 'ModelInterface':
        if model_name.lower() == 'glm45':
            return self.glm45
        elif model_name.lower() == 'deepseek_v31':
            return self.deepseek_v31
        elif model_name.lower() == 'qwen':
            return self.qwen
        elif model_name.lower() == 'qwen_instruct':
            return self.qwen_instruct
        else:
            logger.error(f"不支持的模型名称: {model_name}")
            raise ValueError(f"不支持的模型名称: {model_name}")

# 测试代码
if __name__ == "__main__":
    # 测试模型接口
    api_key = "ms-b4690538-3224-493a-8f5b-4073d527f788"
    manager = ModelManager(api_key)
    
    # 测试GLM-4.5
    glm45 = manager.get_model('glm45')
    messages = [{"role": "user", "content": "你好，请简单介绍一下自己"}]
    
    print("=== 测试GLM-4.5普通对话 ===")
    response = glm45.chat(messages)
    print(f"GLM-4.5回复: {response}")
    
    print("\n=== 测试GLM-4.5流式对话 ===")
    def stream_callback(content):
        print(content, end='', flush=True)
    
    response = glm45.chat_stream(messages, stream_callback)
    print(f"\n完整回复: {response}")
    
    # 测试DeepSeek-V3.1
    deepseek_v31 = manager.get_model('deepseek_v31')
    messages = [{"role": "user", "content": "你好，请简单介绍一下自己"}]
    
    print("\n=== 测试DeepSeek-V3.1普通对话 ===")
    response = deepseek_v31.chat(messages)
    print(f"DeepSeek-V3.1回复: {response}")
    
    print("\n=== 测试DeepSeek-V3.1流式对话 ===")
    response = deepseek_v31.chat_stream(messages, stream_callback)
    print(f"\n完整回复: {response}")