@dataclass
class Message:
    uuid: str                # 消息唯一标识
    parent_uuid: Optional[str] # 父消息标识
    role: str                # 角色：'user' (用户), 'assistant' (助手)
    content: str             # 消息文本内容
    timestamp: str           # 时间戳
    tool_uses: list          # 工具使用记录列表
    tool_results: list       # 工具执行结果列表

@dataclass
class Conversation:
    session_id: str          # 会话 ID
    file_path: str           # 文件存储路径
    summary: Optional[str]   # 会话摘要（如果有）
    messages: list           # 消息对象列表
    project_path: str        # 项目路径
    git_branch: Optional[str]# Git 分支
    timestamp: str           # 会话时间

@dataclass
class SearchResult:
    conversation: Conversation # 关联的会话对象
    score: float             # 相关性得分
    matched_messages: list   # 匹配到的消息列表
    problem_excerpt: str     # 问题摘录
    solution_excerpt: str    # 解决方案摘录
    commands_run: list       # 执行过的命令列表

def calculate_relevance_score(query: str, conversation: Conversation) -> tuple:
    """计算会话与查询的相关性得分"""
    # 1. 对查询语句进行简单的分词（转小写，提取单词）
    query_tokens = tokenize(query)
    if not query_tokens:
        return 0.0, []

    total_score = 0.0
    matched_messages = []

    # 2. 检查会话摘要 (Summary) - 权重最高 (3.0)
    if conversation.summary:
        summary_tokens = tokenize(conversation.summary)
        # 计算重叠词比例
        summary_overlap = len(query_tokens & summary_tokens) / len(query_tokens)
        total_score += summary_overlap * 3.0

    # 3. 遍历每一条消息
    for msg in conversation.messages:
        msg_tokens = tokenize(msg.content)
        overlap = len(query_tokens & msg_tokens)

        if overlap > 0:
            # 基础分：重叠词比例
            msg_score = overlap / len(query_tokens)

            # 策略：提升用户消息的权重 (1.5x)
            # 原因：用户通常在消息中描述具体的问题或报错信息
            if msg.role == 'user':
                msg_score *= 1.5

            # 策略：提升包含工具使用的消息权重 (1.3x)
            # 原因：包含代码执行或文件修改的消息通常是解决方案所在
            if msg.tool_uses:
                msg_score *= 1.3

            total_score += msg_score
            matched_messages.append(msg)

    return total_score, matched_messages

def extract_bash_commands(conversation: Conversation) -> list:
    """提取会话期间运行的所有 Bash 命令"""
    commands = []
    for msg in conversation.messages:
        for tool in msg.tool_uses:
            # 过滤出 'Bash' 类型的工具调用
            if tool.get('name') == 'Bash':
                cmd = tool.get('input', {}).get('command', '')
                if cmd:
                    commands.append(cmd)
    return commands

def extract_files_touched(conversation: Conversation) -> list:
    """提取被读取、写入或编辑的文件列表"""
    files = set()
    for msg in conversation.messages:
        for tool in msg.tool_uses:
            name = tool.get('name', '')
            inp = tool.get('input', {})

            # 关注文件操作类的工具
            if name in ('Read', 'Write', 'Edit'):
                path = inp.get('file_path', '')
                if path:
                    # 仅保留文件名，保持输出简洁
                    files.add(Path(path).name)
    return sorted(files)[:10]

def format_digest(conversations: list, target_date: datetime, project_filter: Optional[str]) -> str:
    """格式化每日摘要报告"""
    date_str = target_date.strftime('%B %d, %Y')
    
    # ... (省略头部构建)

    lines = []
    for i, conv in enumerate(conversations, 1):
        # 自动提取问题描述作为标题
        problem = extract_problem_excerpt(conv)
        commands = extract_bash_commands(conv)
        files = extract_files_touched(conv)

        # 截取前60个字符作为标题
        title = problem[:60].replace('\n', ' ')
        
        # 构建 Markdown 输出
        lines.append(f"### {i}. {title}")
        lines.append(f"   Session: `{conv.session_id[:8]}`")
        if conv.git_branch:
            lines.append(f"   Branch: `{conv.git_branch}`")
        
        # 列出涉及的关键文件和命令数量
        if files:
            lines.append(f"   Files: {', '.join(files[:5])}")
        if commands:
            lines.append(f"   Commands: {len(commands)} executed")
        
        lines.append("")

    return '\n'.join(lines)