# 系统标准库
import os          # 路径处理、目录创建、文件系统操作
import time        # API调用限流，避免请求频率过高
import re          # 正则表达式，匹配WGM消息头，实现消息精准分割
# 第三方库（正式模式使用）：openai（DS API兼容OpenAI接口规范）

# ======================== 路径配置区 ========================
BASE_DIR = "xxxxx"  # 项目根目录
SOURCE_FILE = os.path.join(BASE_DIR, "value.txt")         # WGM原始记录输入文件
PROMPT_FILE = os.path.join(BASE_DIR, "main/PromptV1.md")  # DS提示词文件（表格+类别总结）
OUTPUT_FILE = os.path.join(BASE_DIR, "summary_result.md") # 正式总结输出文件
TEST_OUTPUT_FILE = os.path.join(BASE_DIR, "test_summary_result.md") # 测试输出文件
TEST_CHUNK_FILE = os.path.join(BASE_DIR, "test_chunk_check.txt")   # 测试切片校验文件

# ======================== 滑动窗口参数（正式模式） ========================
CHUNK_SIZE = 1000    # 单窗口处理消息数
OVERLAP = 200        # 窗口重叠数，保证总结连贯性
STEP = CHUNK_SIZE - OVERLAP  # 窗口滑动步长

# ======================== 测试模式配置 ========================
TEST_MODE = False                # 测试模式开关
TEST_MESSAGES_LIMIT = 50         # 测试模式最大处理消息数
TEST_CHUNK_SIZE = 10             # 测试窗口大小
TEST_OVERLAP = 3                 # 测试窗口重叠数

# ======================== 正则表达式配置 ========================
# 匹配WGM消息头：用户名(时间戳): 格式
MESSAGE_HEADER_PATTERN = re.compile(r'^.*?\(\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}\):')

def parse_chat_records(lines):
    """
    原始聊天记录行解析：转换为单条消息列表
    参数：lines (list): WGM原始行列表
    返回：list: 完整单条消息列表
    """
    messages = []          # 最终消息列表
    current_message = []   # 临时拼接当前消息
    
    for line in lines:
        clean_line_for_match = line.lstrip() # 去除行首空白，保证正则匹配
        
        # 匹配到新消息头 → 保存上一条，开始新消息
        if MESSAGE_HEADER_PATTERN.match(clean_line_for_match):
            if current_message:
                messages.append("".join(current_message))
            current_message = [line]
        else:
            # 未匹配到 → 追加为当前消息的跨行内容
            if current_message:
                current_message.append(line)
    
    # 保存最后一条消息
    if current_message:
        messages.append("".join(current_message))
    
    return messages

def get_api_response(prompt, content, is_test=False):
    """
    调用DS API生成WGM记录总结
    参数：prompt(提示词), content(待总结内容), is_test(测试模式)
    返回：str: 总结文本
    """
    if is_test:
        # 测试模式：模拟总结结果，无需调用DS API
        return f"[模拟总结] 本批次包含 {len(content.splitlines())} 行内容，消息解析正常。"
    
    # 正式模式：初始化DS客户端（兼容OpenAI接口）
    from openai import OpenAI
    client = OpenAI(
        api_key="你的API Key",
        base_url="https://api.deepseek.com"
    )
    
    # 发送请求：系统提示词（表格+类别）+ WGM聊天内容
    response = client.chat.completions.create(
        model="deepseek-chat",
        messages=[{"role": "system", "content": prompt}, {"role": "user", "content": content}],
        temperature=0.3 
    )
    return response.choices[0].message.content

def main():
    """主执行流程：加载配置→解析WGM→切片→调用API→写入结果"""
    # 步骤1：加载DS提示词文件（要求表格+类别总结）
    try:
        with open(PROMPT_FILE, 'r', encoding='utf-8') as f:
            system_prompt = f.read()
    except (FileNotFoundError, Exception) as e:
        print(f"提示词加载失败：{e}")
        return

    # 步骤2：读取并解析WGM原始记录
    try:
        with open(SOURCE_FILE, 'r', encoding='utf-8') as f:
            raw_lines = f.readlines()
    except (FileNotFoundError, Exception) as e:
        print(f"WGM文件加载失败：{e}")
        return
    all_messages = parse_chat_records(raw_lines) # 解析消息
    total_messages = len(all_messages)
    print(f"✅ 检测到WGM总消息数: {total_messages}")

    # 步骤3：测试/正式模式参数切换
    if TEST_MODE:
        test_messages = all_messages[:TEST_MESSAGES_LIMIT]
        current_chunk_size = TEST_CHUNK_SIZE
        current_overlap = TEST_OVERLAP
        current_step = current_chunk_size - current_overlap
        current_output_file = TEST_OUTPUT_FILE
        chunk_check_file = open(TEST_CHUNK_FILE, 'w', encoding='utf-8')
    else:
        test_messages = all_messages
        current_chunk_size = CHUNK_SIZE
        current_overlap = OVERLAP
        current_step = STEP
        current_output_file = OUTPUT_FILE
        chunk_check_file = None

    # 步骤4：滑动窗口处理 + 写入总结结果
    with open(current_output_file, 'w', encoding='utf-8') as out_f:
        # 测试模式写入头部说明
        if TEST_MODE:
            out_f.write("# 测试模式WGM总结结果\n\n")
        
        # 滑动窗口循环切片
        for i in range(0, len(test_messages), current_step):
            chunk_messages = test_messages[i : i + current_chunk_size]
            chunk_text = "\n\n".join(chunk_messages)
            
            start_idx = i + 1
            end_idx = i + len(chunk_messages)
            print(f"🔄 处理WGM第 {start_idx} - {end_idx} 条消息")

            # 测试模式保存切片内容
            if TEST_MODE and chunk_check_file:
                chunk_check_file.write(f"=== WGM消息批次 {start_idx}-{end_idx} ===\n{chunk_text}\n\n")

            # 调用DS API生成总结
            try:
                summary = get_api_response(system_prompt, chunk_text, is_test=TEST_MODE)
                # 写入Markdown文件
                out_f.write(f"### WGM批次 {start_idx} - {end_idx}\n{summary}\n\n{'-'*50}\n\n")
                out_f.flush()
            except Exception as e:
                print(f"❌ 批次处理失败：{e}")
                continue

            # 结束循环条件
            if i + current_chunk_size >= len(test_messages):
                break
            
            # 正式模式限流1秒
            if not TEST_MODE:
                time.sleep(1)

    # 关闭测试文件
    if TEST_MODE and chunk_check_file:
        chunk_check_file.close()
    print(f"\n✅ WGM总结完成！文件：{current_output_file}")

if __name__ == "__main__":
    """程序入口：初始化目录 → 执行主流程"""
    # 自动创建所有输出文件的父目录，避免目录不存在报错
    for path in [OUTPUT_FILE, TEST_OUTPUT_FILE, TEST_CHUNK_FILE]:
        dir_path = os.path.dirname(path)
        os.makedirs(dir_path, exist_ok=True)
    
    # 启动主程序
    main()