"""会话目标自动持久化服务。

从用户消息自动持久化会话目标到 goals 表。
如果已有该 character 的活跃 session 级目标，判断内容是否显著不同：
- 内容相似 → 复用
- 内容不同 → 完成旧目标，创建新目标
"""

from app.models.governance import Goal
from app.repositories import governance as governance_repo

_TOPIC_SHIFT_MIN_LENGTH = 20


async def auto_persist_session_goal(
    session,
    character_id: str,
    content: str,
) -> Goal | None:
    """自动持久化会话目标：复用活跃的 session 目标或新建。

    内容过短（<10 字）不产生目标。
    如果已有活跃 session 目标但内容显著不同，先完成旧目标再新建。

    Returns:
        持久化后的 Goal（可能为 None）。
    """
    goal_text = (content or "").strip()
    if not goal_text or len(goal_text) < 10:
        return None

    existing = await governance_repo.list_active_session_goal(session, character_id)
    if existing is not None:
        # 判断是否话题转换：内容不再相似时完成旧目标
        if _is_topic_shift(existing.title, goal_text):
            existing.status = "completed"
            existing.progress = 1.0
            # 创建新目标
            goal = Goal(
                character_id=character_id,
                title=goal_text[:120],
                description=f"从对话自动提取的会话目标：{goal_text[:500]}",
                level="session",
                type="mission",
                status="active",
                priority=50,
            )
            await governance_repo.add_goal(session, goal)
            return goal
        return existing

    # 新建 session 目标
    goal = Goal(
        character_id=character_id,
        title=goal_text[:120],
        description=f"从对话自动提取的会话目标：{goal_text[:500]}",
        level="session",
        type="mission",
        status="active",
        priority=50,
    )
    await governance_repo.add_goal(session, goal)
    return goal


def _is_topic_shift(existing_title: str, new_content: str) -> bool:
    """判断新内容与旧目标标题是否话题转换。

    简单规则：如果新内容较长且与旧标题共享的关键词很少，视为话题转换。
    这是启发式方法，后续可升级为 LLM 判断。
    """
    if len(new_content) < _TOPIC_SHIFT_MIN_LENGTH:
        return False
    # 提取关键词集合（按空格/标点分割）
    import re
    old_words = set(re.split(r"[\s，。！？、；：,\.!?;:]+", existing_title.lower()))
    new_words = set(re.split(r"[\s，。！？、；：,\.!?;:]+", new_content.lower()))
    # 排除常见的助词
    stop_words = {"的", "了", "是", "在", "有", "和", "就", "不", "人", "都", "一",
                  "一个", "上", "也", "很", "到", "说", "要", "去", "你", "我", "他",
                  "这", "那", "吗", "啊", "呢", "吧", "地", "得"}
    old_keywords = old_words - stop_words
    new_keywords = new_words - stop_words

    if not old_keywords or not new_keywords:
        return False

    overlap = old_keywords & new_keywords
    # 如果旧关键词的一个都没有出现在新内容中，或者重合率极低
    if not overlap:
        return True
    return len(overlap) / max(len(old_keywords), len(new_keywords)) < 0.15