Method stream_invoke

QueryEngine/llms/base.py:86–131

流式调用LLM，逐步返回响应内容 Args: system_prompt: 系统提示词 user_prompt: 用户提示词 **kwargs: 额外参数（temperature, top_p等） Yields: 响应文本块（str）

(self, system_prompt: str, user_prompt: str, **kwargs)

        return ""

    def stream_invoke(self, system_prompt: str, user_prompt: str, **kwargs) -> Generator[str, None, None]:
        """
        流式调用LLM，逐步返回响应内容
        
        Args:
            system_prompt: 系统提示词
            user_prompt: 用户提示词
            **kwargs: 额外参数（temperature, top_p等）
            
        Yields:
            响应文本块（str）
        """
        current_time = datetime.now().strftime("%Y年%m月%d日%H时%M分")
        time_prefix = f"今天的实际时间是{current_time}"
        if user_prompt:
            user_prompt = f"{time_prefix}\n{user_prompt}"
        else:
            user_prompt = time_prefix
        messages = [
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": user_prompt},
        ]

        allowed_keys = {"temperature", "top_p", "presence_penalty", "frequency_penalty"}
        extra_params = {key: value for key, value in kwargs.items() if key in allowed_keys and value is not None}
        # 强制使用流式
        extra_params["stream"] = True

        timeout = kwargs.pop("timeout", self.timeout)

        try:
            stream = self.client.chat.completions.create(
                model=self.model_name,
                messages=messages,
                timeout=timeout,
                **extra_params,
            )
            
            for chunk in stream:
                if chunk.choices and len(chunk.choices) > 0:
                    delta = chunk.choices[0].delta
                    if delta and delta.content:
                        yield delta.content
        except Exception as e:
            logger.error(f"流式请求失败: {str(e)}")
            raise e
    
    @with_retry(LLM_RETRY_CONFIG)
    def stream_invoke_to_string(self, system_prompt: str, user_prompt: str, **kwargs) -> str: