import json
import logging
import os
import asyncio
from concurrent.futures import ThreadPoolExecutor, as_completed
from typing import Callable
from typing import List, Dict

from langchain.agents.middleware import ModelFallbackMiddleware
from langchain.tools import tool
from langchain_core.runnables import RunnableConfig

from deepinsight.core.utils.research_utils import parse_research_config
from deepinsight.core.utils.tool_utils import CoerceToolOutput


import os
from langchain.tools import tool

from deepinsight.utils.tavily_manager import tavily_key_manager


@tool
async def person_image_search_tool(person_name: str, person_background: str, config: RunnableConfig) -> str:
    """Tool to search for an image of a person based on their name and background information."""

    # 输入参数:
    # person_name (str): 人员的姓名。例如："John Doe"。
    # person_background (str): 人员的背景信息，通常包括其工作地和从事的行业或领域。
    #                          例如："Harvard University, Professor of Physics"。

    rc = parse_research_config(config)

    tool_instance = tavily_key_manager().tool(
        max_results=10,
        topic="general",
        include_answer=True,
        include_raw_content=False,
        include_images=True,
        include_image_descriptions=False,
        search_depth="advanced"
    )

    from langchain.agents import create_agent
    user_prompt = f"""
    帮我搜索找下如下人员的头像或生活照：人名：{person_name} 背景信息：{person_background}，输出格式参考如下(不要输出其它任何内容，仅输出json)： 
    {{
        "name": "<人员姓名>",  # 人员的姓名
        "image": "<图片URL>"  # 图片链接，可以是头像照片或生活照的URL
    }}
    """

    agent = create_agent(rc.default_model, tools=[tool_instance])
    input_messages = [
        {
            "role": "user",
            "content": user_prompt
        }]
    result = await agent.ainvoke({"messages": input_messages})
    result_text = result["messages"][-1].content
    return result_text


# ----------------- 单篇论文解析函数 -----------------

async def analyze_single_keynote(keynote_info: str, output_dir: str, config: RunnableConfig) -> bool:
    """
    对keynotes进行解析，并将结果保存到文件

    Args:
        keynote_info: keynotes信息
        output_dir: 保存解析结果的文件夹路径

    Returns:
        bool: True表示解析成功并保存，False表示解析失败
    """
    try:
        rc = parse_research_config(config)

        tavily_instance = tavily_key_manager().tool(
            max_results=2,
            topic="general",
            include_answer=True,
            include_raw_content=False,
            include_images=False,
            include_image_descriptions=True,
            search_depth="advanced",
        )
        # Step 2: Generate structured research brief from user messages
        prompt_content = rc.prompt_manager.get_prompt(
            name="analyze_keynote_system_prompt",
            group=rc.prompt_group,
        ).format(output_dir=output_dir)

        tools = [tavily_instance, person_image_search_tool]
        from deepagents import create_deep_agent
        middleware = [
            CoerceToolOutput(),
            ModelFallbackMiddleware(
                rc.default_model,
                rc.default_model,
            )
        ]
        # Create the deep agent
        summary_subagent_system_prompt = rc.prompt_manager.get_prompt(
            name="review_keynote_prompt",
            group=rc.prompt_group,
        ).format()
        summary_subagent = {
            "name": "summary-agent",
            "description": "顶会keynote洞察战略分析与点评助手，你需要获取全部洞察相关资料（必须包括1、演讲嘉宾的全部信息（300字）2、演讲嘉宾演讲的详细内容（500字），3、对演讲内容的 why、what、how的多角度总结（500字），请注意这些内容是你需要预先获取的！！不是你要生成的任务）；通过这些洞察资料分析并分别给出具有商业价值、技术领域高度"
                           "技术突破的分析与点评内容。",
            "system_prompt": summary_subagent_system_prompt,
            "tools": []
        }
        # Create the deep agent
        agent = create_deep_agent(
            model=rc.default_model,
            tools=tools,
            system_prompt=prompt_content,
            middleware=middleware,
            backend=rc.file_system.deep_agent_backend(),
            subagents=[summary_subagent]
        )
        input_messages = [
            {
                "role": "user",
                "content": f"请分析以下学术会议的keyote，并输出高质量结果, keynote信息：{keynote_info}, 并将最终结果输出到目录：{output_dir}"
            }]

        # Invoke the agent
        config_dict = dict(config) if not isinstance(config, dict) else config
        config_dict = {**config_dict, "recursion_limit": 300}

        result = await agent.ainvoke({"messages": input_messages}, config=config_dict)
    except Exception as e:
        logging.error(f"keynote分析失败: {keynote_info}, 错误: {e}")
        import traceback
        traceback.print_exc()  # 打印堆栈信息


# ----------------- 批量论文解析工具 -----------------
@tool
async def batch_analyze_keynotes(
    keynotes_info: List[str],
    output_dir: str,
    config: RunnableConfig
) -> Dict[str, bool]:
    """
    批量并行分析 keynotes，将分析后的结果保存到指定文件夹中，
    并返回每篇论文是否分析成功的 Map。

    Args:
        keynotes_info (List[str]):
            每个元素是一个 keynotes 的的自然语言描述或 JSON字符串,不能是json格式
            示例：
            [
                '{"keynote_name": "AI for the Future of Computing", "speaker": "Dr. Jane Doe", "conference": "ICCAD 2025"}',
                '{"keynote_name": "Semiconductor Innovation in the Post-Moore Era", "speaker": "Prof. John Smith", "conference": "ICCAD 2025"}'
            ]
        output_dir (str): 保存 keynotes 分析结果的文件夹路径
    Returns:
        Dict[str, bool]: 每个 keynote 对应的分析成功状态(True/False)
    """
    logging.info(f"接收到 keynotes_info，共 {len(keynotes_info)} 个")

    result_map: Dict[str, bool] = {}

    tasks = [
        analyze_single_keynote(keynote, output_dir, config)
        for keynote in keynotes_info
    ]

    # Execute tasks concurrently
    results = await asyncio.gather(*tasks, return_exceptions=True)
    for keynote, result in zip(keynotes_info, results):
        if isinstance(result, Exception):
            logging.error(f"Failed to analyze keynote {keynote}: {result}")
            result_map[keynote] = False
        else:
            result_map[keynote] = result
            
    return result_map