"""학습 워크스페이스(study_topic) API — 필기 세션 + 자료(library document) 묶음 컨테이너

핵심 개념:
  - study_topic 은 단순 폴더/태그가 아니라 학습 자산 1차 컨테이너 (학습 워크스페이스).
  - 향후 단어장/오디오/문제 세트 같은 자산이 같은 컨테이너 아래로 들어올 수 있도록
    응답 구조를 sections + stats 형태로 설계 (현재는 sessions / documents 두 키만).
  - documents.category(자료실 UI 축) 와 직교한 별도 분류 축. 자료실 facet/카테고리는 미터치.
  - StudySession.certification/subject/topic 은 보존 — 본 컨테이너 와 직교한 세부 메타.

설계 제약:
  - study_type 은 강한 enum 미사용. VALID_STUDY_TYPES 는 단순 권장값 (DB/Pydantic 강제 안 함).
  - soft delete (deleted_at). active 행끼리만 (user_id, name) partial unique.
  - 권한: 모든 쿼리에 user_id 강제. 다른 사용자 자료를 묶지 못함.
  - 문서 매핑은 N:M (study_topic_documents). 세션 매핑은 1:N (study_sessions.study_topic_id).
  - polymorphic 단일 study_topic_items 테이블은 만들지 않는다 (영구 금지).
"""

import asyncio
import logging
import random as _random
from datetime import datetime, timezone
from enum import Enum
from pathlib import Path as _Path
from typing import Annotated, Any

from fastapi import APIRouter, Depends, HTTPException, Query
from pydantic import BaseModel, Field
from sqlalchemy import and_, case, delete, func, or_ as sql_or, select, text as sql_text, update
from sqlalchemy.exc import IntegrityError
from sqlalchemy.ext.asyncio import AsyncSession

from ai.client import AIClient, strip_thinking
from eid.ai import EidAIClient
from eid.compose import compose
from core.auth import get_current_user
from core.database import get_session
from core.library import LIBRARY_PREFIX, normalize_library_path
from models.document import Document
from models.study_session import StudySession
from models.study_topic import StudyTopic, StudyTopicDocument
from models.study_question import StudyQuestion, StudyQuestionAttempt
from models.study_question_image import StudyQuestionImage
from models.study_quiz_session import StudyQuizSession
from models.study_topic_subject_note import StudyTopicSubjectNote
from models.eid_study_weakness import EidStudyWeakness
from models.eid_review_set_draft import EidReviewSetDraft
from models.user import User
from services.search.llm_gate import Priority, acquire_mlx_gate
from services.study.subject_note_rag import (
    SubjectNoteContext,
    gather_subject_note_context,
    render_evidence_block,
)
from services.study.weakness_compute import format_habit_block, format_weakness_block

logger = logging.getLogger(__name__)
router = APIRouter()


# ─── 권장값 (강제 enum 아님, UI 안내용) ───

RECOMMENDED_STUDY_TYPES: set[str] = {
    "certification", "language", "school", "work", "general",
}


# ─── Pydantic 스키마 ───


class StudyTopicCreate(BaseModel):
    name: str = Field(min_length=1, max_length=120)
    description: str | None = None
    color: str | None = Field(default=None, max_length=20)
    study_type: str | None = Field(default=None, max_length=40)
    sort_order: int = 0
    # PR-6: 시험 메타
    exam_round_size: int | None = Field(default=None, ge=1, le=300)
    exam_subjects: list[str] = []


class StudyTopicUpdate(BaseModel):
    name: str | None = Field(default=None, min_length=1, max_length=120)
    description: str | None = None
    color: str | None = Field(default=None, max_length=20)
    study_type: str | None = Field(default=None, max_length=40)
    sort_order: int | None = None
    # PR-6: 시험 메타
    exam_round_size: int | None = Field(default=None, ge=1, le=300)
    exam_subjects: list[str] | None = None
    # 공부 암기노트: 공부중 토글 (true=focused_at=now, false=clear)
    focused: bool | None = None


class StudyTopicResponse(BaseModel):
    """주제 목록 응답 — 집계 카운트 포함."""

    id: int
    name: str
    description: str | None
    color: str | None
    study_type: str | None
    sort_order: int
    session_count: int = 0
    document_count: int = 0
    question_count: int = 0
    # PR-6: 시험 메타
    exam_round_size: int | None = None
    exam_subjects: list[str] = []
    # 공부 암기노트: 공부중 태그 상태
    focused: bool = False
    created_at: datetime
    updated_at: datetime


class StudyTopicListResponse(BaseModel):
    items: list[StudyTopicResponse]
    total: int


class StudyTopicSessionSummary(BaseModel):
    """상세 뷰의 세션 카드 페이로드 — 통합 뷰 렌더용 최소 필드."""

    id: int
    study_type: str
    certification: str | None
    language_code: str | None
    learning_level: str | None
    subject: str | None
    topic: str | None
    mode: str
    repetition_count: int
    review_state: str | None
    created_at: datetime
    updated_at: datetime


class StudyTopicDocumentSummary(BaseModel):
    """상세 뷰의 자료 카드 페이로드.

    library_paths: 통합뷰 자료 섹션의 카테고리 트리 그룹핑용. user_tags 의
    `@library/<path>` 태그에서 prefix 제거한 path 들. 한 자료가 여러 카테고리에
    속할 수 있으나(다중 분류), 트리 그룹핑 시 첫 번째 path 만 사용. 빈 리스트면
    "분류 없음" 그룹.
    """

    id: int
    title: str | None
    file_format: str
    file_type: str
    category: str | None
    ai_domain: str | None
    importance: str | None
    sort_order: int
    linked_at: datetime
    library_paths: list[str] = []


class StudyTopicQuestionSummary(BaseModel):
    """상세 뷰의 문제 카드 페이로드 — 정답·해설 비공개, 본문 80자 truncate."""

    id: int
    question_text: str
    subject: str | None
    scope: str | None
    exam_name: str | None
    exam_round: str | None
    exam_question_number: int | None  # PR-11: 회차별 그룹 안 정렬 키
    is_active: bool
    attempt_count: int
    last_correct: bool | None
    created_at: datetime


class StudyTopicSections(BaseModel):
    """확장 친화 dict — 향후 audio_assets / vocab_decks 키 추가 가능."""

    sessions: list[StudyTopicSessionSummary]
    documents: list[StudyTopicDocumentSummary]
    questions: list[StudyTopicQuestionSummary] = []


class StudyTopicStats(BaseModel):
    """자산 카운트 — 0 으로 미리 노출해서 후속 PR 에서 필드만 채움.
    question_count = PR-2 단일 문제 수. question_set_count = 후속 PR 의 회차/모의고사 묶음 수 (둘은 의미 다름).
    """

    session_count: int
    document_count: int
    question_count: int = 0
    audio_count: int = 0
    vocab_count: int = 0
    question_set_count: int = 0


class StudyTopicMeta(BaseModel):
    id: int
    name: str
    description: str | None
    color: str | None
    study_type: str | None
    sort_order: int
    # PR-6: 시험 메타
    exam_round_size: int | None = None
    exam_subjects: list[str] = []
    # 공부 암기노트: 공부중 태그 상태
    focused: bool = False
    created_at: datetime
    updated_at: datetime


class StudyTopicDetailResponse(BaseModel):
    topic: StudyTopicMeta
    sections: StudyTopicSections
    stats: StudyTopicStats


class StudyTopicDocumentLinkRequest(BaseModel):
    document_ids: list[int] = Field(min_length=1, max_length=100)


class StudyTopicDocumentLinkResponse(BaseModel):
    linked: list[int]
    skipped_existing: list[int]
    skipped_not_found: list[int]


class StudyTopicDocumentByPathRequest(BaseModel):
    """카테고리 트리 노드 일괄 추가용. path prefix 매칭으로 하위 자료 모두 매핑.
    include_subtree=True 면 path 와 그 하위 카테고리 전체, False 면 path 정확 일치만.
    """
    path: str = Field(min_length=1, max_length=500)
    include_subtree: bool = True


class StudyTopicDocumentByPathResponse(BaseModel):
    linked_count: int
    skipped_existing_count: int
    total_in_path: int
    path: str


# ─── Helpers ───


def _verify_topic_ownership(topic: StudyTopic | None, user: User) -> StudyTopic:
    """소유자 검증 + soft-deleted 행 차단. mismatch 도 404 (정보 누설 방지)."""
    if topic is None or topic.user_id != user.id or topic.deleted_at is not None:
        raise HTTPException(status_code=404, detail="학습 주제를 찾을 수 없습니다")
    return topic


def _meta_from_topic(t: StudyTopic) -> StudyTopicMeta:
    return StudyTopicMeta(
        id=t.id,
        name=t.name,
        description=t.description,
        color=t.color,
        study_type=t.study_type,
        sort_order=t.sort_order,
        exam_round_size=t.exam_round_size,
        exam_subjects=t.exam_subjects or [],
        created_at=t.created_at,
        updated_at=t.updated_at,
    )


# ─── 엔드포인트 ───


@router.get("/", response_model=StudyTopicListResponse)
async def list_study_topics(
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
    limit: int = Query(100, ge=1, le=500),
    offset: int = Query(0, ge=0),
):
    """사용자의 active 주제 목록. 세션 수·자료 수 집계 포함."""

    # active 주제 + 세션 수 (LEFT JOIN study_sessions)
    sess_count_sub = (
        select(
            StudySession.study_topic_id.label("topic_id"),
            func.count().label("c"),
        )
        .where(StudySession.user_id == user.id)
        .where(StudySession.study_topic_id.is_not(None))
        .group_by(StudySession.study_topic_id)
        .subquery()
    )

    doc_count_sub = (
        select(
            StudyTopicDocument.study_topic_id.label("topic_id"),
            func.count().label("c"),
        )
        .where(StudyTopicDocument.user_id == user.id)
        .group_by(StudyTopicDocument.study_topic_id)
        .subquery()
    )

    # PR-2 — 활성 문제 카운트 (soft-deleted 제외)
    from models.study_question import StudyQuestion as _SQ
    q_count_sub = (
        select(
            _SQ.study_topic_id.label("topic_id"),
            func.count().label("c"),
        )
        .where(_SQ.user_id == user.id, _SQ.deleted_at.is_(None))
        .group_by(_SQ.study_topic_id)
        .subquery()
    )

    base = (
        select(
            StudyTopic,
            func.coalesce(sess_count_sub.c.c, 0).label("session_count"),
            func.coalesce(doc_count_sub.c.c, 0).label("document_count"),
            func.coalesce(q_count_sub.c.c, 0).label("question_count"),
        )
        .outerjoin(sess_count_sub, sess_count_sub.c.topic_id == StudyTopic.id)
        .outerjoin(doc_count_sub, doc_count_sub.c.topic_id == StudyTopic.id)
        .outerjoin(q_count_sub, q_count_sub.c.topic_id == StudyTopic.id)
        .where(StudyTopic.user_id == user.id, StudyTopic.deleted_at.is_(None))
        .order_by(StudyTopic.sort_order.asc(), StudyTopic.id.desc())
    )

    total_query = select(func.count()).select_from(
        select(StudyTopic.id)
        .where(StudyTopic.user_id == user.id, StudyTopic.deleted_at.is_(None))
        .subquery()
    )
    total = (await session.execute(total_query)).scalar() or 0

    rows = (await session.execute(base.offset(offset).limit(limit))).all()

    items = [
        StudyTopicResponse(
            id=t.id,
            name=t.name,
            description=t.description,
            color=t.color,
            study_type=t.study_type,
            sort_order=t.sort_order,
            session_count=int(sc),
            document_count=int(dc),
            question_count=int(qc),
            exam_round_size=t.exam_round_size,
            exam_subjects=t.exam_subjects or [],
            created_at=t.created_at,
            updated_at=t.updated_at,
        )
        for t, sc, dc, qc in rows
    ]
    return StudyTopicListResponse(items=items, total=total)


class ExamRoundProgress(BaseModel):
    exam_round: str
    question_count: int
    max_question_number: int  # 0 = exam_question_number 가 모두 NULL
    next_question_number: int | None  # 도달 시 None
    is_complete: bool


class ExamRoundsResponse(BaseModel):
    exam_round_size: int | None  # 토픽 메타. NULL = 미설정
    items: list[ExamRoundProgress]


@router.get("/{topic_id}/exam-rounds", response_model=ExamRoundsResponse)
async def list_exam_rounds(
    topic_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """토픽 안 회차별 진행률 집계. 진행률 표시 + 재개 UX 용.

    is_complete = (exam_round_size IS NOT NULL AND question_count >= exam_round_size).
    next_question_number = max + 1 (도달 시 NULL).
    """
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    from models.study_question import StudyQuestion as _SQ

    rows = (
        await session.execute(
            select(
                _SQ.exam_round,
                func.count().label("question_count"),
                func.coalesce(func.max(_SQ.exam_question_number), 0).label("max_qnum"),
            )
            .where(
                _SQ.user_id == user.id,
                _SQ.study_topic_id == topic_id,
                _SQ.deleted_at.is_(None),
                _SQ.exam_round.is_not(None),
            )
            .group_by(_SQ.exam_round)
            .order_by(_SQ.exam_round.desc())
        )
    ).all()

    size = topic.exam_round_size
    items: list[ExamRoundProgress] = []
    for r in rows:
        cnt = int(r.question_count)
        mx = int(r.max_qnum)
        is_complete = bool(size is not None and cnt >= size)
        # 도달했으면 next None, 아니면 max+1 (max=0 이면 1번부터)
        next_n = None if is_complete else (mx + 1 if mx > 0 else 1)
        items.append(ExamRoundProgress(
            exam_round=r.exam_round,
            question_count=cnt,
            max_question_number=mx,
            next_question_number=next_n,
            is_complete=is_complete,
        ))

    return ExamRoundsResponse(exam_round_size=size, items=items)


@router.get("/by-document/{document_id}", response_model=list[StudyTopicMeta])
async def list_topics_for_document(
    document_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """이 자료가 매핑된 active 주제 목록.
    /study/sources 카드의 "이 자료가 속한 주제" 배지·컨텍스트 메뉴용.
    """
    rows = (
        await session.execute(
            select(StudyTopic)
            .join(StudyTopicDocument, StudyTopicDocument.study_topic_id == StudyTopic.id)
            .where(
                StudyTopicDocument.document_id == document_id,
                StudyTopicDocument.user_id == user.id,
                StudyTopic.deleted_at.is_(None),
            )
            .order_by(StudyTopic.sort_order.asc(), StudyTopic.id.desc())
        )
    ).scalars().all()
    return [_meta_from_topic(t) for t in rows]


@router.post("/", response_model=StudyTopicResponse, status_code=201)
async def create_study_topic(
    body: StudyTopicCreate,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """주제 생성. 같은 이름의 active 주제가 이미 있으면 409.
    soft-deleted 동명 주제는 partial unique index 에서 빠지므로 신규 생성 가능.
    """
    topic = StudyTopic(
        user_id=user.id,
        name=body.name.strip(),
        description=body.description,
        color=body.color,
        study_type=body.study_type,
        sort_order=body.sort_order,
        exam_round_size=body.exam_round_size,
        exam_subjects=body.exam_subjects or [],
    )
    session.add(topic)
    try:
        await session.flush()
        await session.commit()
    except IntegrityError:
        await session.rollback()
        raise HTTPException(status_code=409, detail="이미 같은 이름의 학습 주제가 있습니다")

    return StudyTopicResponse(
        id=topic.id,
        name=topic.name,
        description=topic.description,
        color=topic.color,
        study_type=topic.study_type,
        sort_order=topic.sort_order,
        session_count=0,
        document_count=0,
        question_count=0,
        exam_round_size=topic.exam_round_size,
        exam_subjects=topic.exam_subjects or [],
        created_at=topic.created_at,
        updated_at=topic.updated_at,
    )


@router.get("/{topic_id}", response_model=StudyTopicDetailResponse)
async def get_study_topic(
    topic_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """통합 뷰: 주제 메타 + 세션 목록 + 자료 목록 + stats.
    응답 형태는 후속 PR(오디오/단어장/문제세트 추가) 에서 sections / stats 키만 늘리면 되도록 dict 구조.
    """
    topic = await session.get(StudyTopic, topic_id)
    topic = _verify_topic_ownership(topic, user)

    # 세션 목록 — 최근순
    sess_rows = (
        await session.execute(
            select(StudySession)
            .where(
                StudySession.user_id == user.id,
                StudySession.study_topic_id == topic_id,
            )
            .order_by(StudySession.created_at.desc(), StudySession.id.desc())
        )
    ).scalars().all()

    sessions_payload = [
        StudyTopicSessionSummary(
            id=s.id,
            study_type=s.study_type,
            certification=s.certification,
            language_code=s.language_code,
            learning_level=s.learning_level,
            subject=s.subject,
            topic=s.topic,
            mode=s.mode,
            repetition_count=s.repetition_count,
            review_state=s.review_state,
            created_at=s.created_at,
            updated_at=s.updated_at,
        )
        for s in sess_rows
    ]

    # 자료 목록 — 매핑 sort_order → created_at desc
    doc_rows = (
        await session.execute(
            select(Document, StudyTopicDocument)
            .join(
                StudyTopicDocument,
                and_(
                    StudyTopicDocument.document_id == Document.id,
                    StudyTopicDocument.study_topic_id == topic_id,
                    StudyTopicDocument.user_id == user.id,
                ),
            )
            .where(Document.deleted_at.is_(None))
            .order_by(
                StudyTopicDocument.sort_order.asc(),
                StudyTopicDocument.created_at.desc(),
            )
        )
    ).all()

    def _extract_library_paths(tags: list | None) -> list[str]:
        if not tags:
            return []
        out: list[str] = []
        for t in tags:
            if isinstance(t, str) and t.startswith(LIBRARY_PREFIX):
                out.append(t[len(LIBRARY_PREFIX):])
        return out

    documents_payload = [
        StudyTopicDocumentSummary(
            id=d.id,
            title=d.title,
            file_format=d.file_format,
            file_type=str(d.file_type) if d.file_type is not None else "immutable",
            category=str(d.category) if d.category is not None else None,
            ai_domain=d.ai_domain,
            importance=d.importance,
            sort_order=link.sort_order,
            linked_at=link.created_at,
            library_paths=_extract_library_paths(d.user_tags),
        )
        for d, link in doc_rows
    ]

    # 문제 목록 (PR-2) — 본문 truncate, 정답·해설 비공개. attempt 통계 batch 로 끌어옴.
    from models.study_question import StudyQuestion, StudyQuestionAttempt

    q_rows = (
        await session.execute(
            select(StudyQuestion)
            .where(
                StudyQuestion.user_id == user.id,
                StudyQuestion.study_topic_id == topic_id,
                StudyQuestion.deleted_at.is_(None),
            )
            .order_by(StudyQuestion.created_at.desc(), StudyQuestion.id.desc())
        )
    ).scalars().all()

    qids = [q.id for q in q_rows]
    q_attempt_count: dict[int, int] = {}
    q_last_correct: dict[int, bool] = {}
    if qids:
        from sqlalchemy import case as _case
        cnt_rows = (
            await session.execute(
                select(
                    StudyQuestionAttempt.study_question_id,
                    func.count().label("total"),
                )
                .where(
                    StudyQuestionAttempt.user_id == user.id,
                    StudyQuestionAttempt.study_question_id.in_(qids),
                )
                .group_by(StudyQuestionAttempt.study_question_id)
            )
        ).all()
        for r in cnt_rows:
            q_attempt_count[r.study_question_id] = int(r.total)

        latest_rows = (
            await session.execute(
                select(
                    StudyQuestionAttempt.study_question_id,
                    StudyQuestionAttempt.is_correct,
                )
                .where(
                    StudyQuestionAttempt.user_id == user.id,
                    StudyQuestionAttempt.study_question_id.in_(qids),
                )
                .order_by(
                    StudyQuestionAttempt.study_question_id,
                    StudyQuestionAttempt.answered_at.desc(),
                )
                .distinct(StudyQuestionAttempt.study_question_id)
            )
        ).all()
        for r in latest_rows:
            q_last_correct[r.study_question_id] = bool(r.is_correct)

    def _truncate_q(text: str, n: int = 80) -> str:
        return text if len(text) <= n else text[:n].rstrip() + "…"

    questions_payload = [
        StudyTopicQuestionSummary(
            id=q.id,
            question_text=_truncate_q(q.question_text, 80),
            subject=q.subject,
            scope=q.scope,
            exam_name=q.exam_name,
            exam_round=q.exam_round,
            exam_question_number=q.exam_question_number,
            is_active=q.is_active,
            attempt_count=q_attempt_count.get(q.id, 0),
            last_correct=q_last_correct.get(q.id),
            created_at=q.created_at,
        )
        for q in q_rows
    ]

    return StudyTopicDetailResponse(
        topic=_meta_from_topic(topic),
        sections=StudyTopicSections(
            sessions=sessions_payload,
            documents=documents_payload,
            questions=questions_payload,
        ),
        stats=StudyTopicStats(
            session_count=len(sessions_payload),
            document_count=len(documents_payload),
            question_count=len(questions_payload),
            # 후속 PR 에서 채움
            audio_count=0,
            vocab_count=0,
            question_set_count=0,
        ),
    )


@router.patch("/{topic_id}", response_model=StudyTopicResponse)
async def update_study_topic(
    topic_id: int,
    body: StudyTopicUpdate,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    topic = await session.get(StudyTopic, topic_id)
    topic = _verify_topic_ownership(topic, user)

    fields_set = body.model_fields_set
    if "name" in fields_set and body.name is not None:
        topic.name = body.name.strip()
    for fname in {"description", "color", "study_type", "sort_order"} & fields_set:
        setattr(topic, fname, getattr(body, fname))
    # PR-6: 시험 메타. exam_subjects 가 None 이면 (PATCH 미명시) 기존 유지, 빈 배열은 명시적 클리어
    if "exam_round_size" in fields_set:
        topic.exam_round_size = body.exam_round_size
    if "exam_subjects" in fields_set and body.exam_subjects is not None:
        topic.exam_subjects = body.exam_subjects
    # 공부 암기노트: 공부중 태그 토글 (focused_at IS NOT NULL = reminder/세션 대상)
    if "focused" in fields_set:
        topic.focused_at = datetime.now(timezone.utc) if body.focused else None

    topic.updated_at = datetime.now(timezone.utc)
    try:
        await session.commit()
    except IntegrityError:
        await session.rollback()
        raise HTTPException(status_code=409, detail="이미 같은 이름의 학습 주제가 있습니다")

    # 응답용 카운트 별도 조회
    sc = (await session.execute(
        select(func.count())
        .select_from(StudySession)
        .where(StudySession.user_id == user.id, StudySession.study_topic_id == topic.id)
    )).scalar() or 0
    dc = (await session.execute(
        select(func.count())
        .select_from(StudyTopicDocument)
        .where(StudyTopicDocument.study_topic_id == topic.id)
    )).scalar() or 0
    from models.study_question import StudyQuestion as _SQ2
    qc = (await session.execute(
        select(func.count())
        .select_from(_SQ2)
        .where(
            _SQ2.user_id == user.id,
            _SQ2.study_topic_id == topic.id,
            _SQ2.deleted_at.is_(None),
        )
    )).scalar() or 0

    return StudyTopicResponse(
        id=topic.id,
        name=topic.name,
        description=topic.description,
        color=topic.color,
        study_type=topic.study_type,
        sort_order=topic.sort_order,
        session_count=int(sc),
        document_count=int(dc),
        question_count=int(qc),
        exam_round_size=topic.exam_round_size,
        exam_subjects=topic.exam_subjects or [],
        focused=topic.focused_at is not None,
        created_at=topic.created_at,
        updated_at=topic.updated_at,
    )


@router.delete("/{topic_id}", status_code=204)
async def delete_study_topic(
    topic_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """주제 soft delete. 세션의 study_topic_id 는 SET NULL, 문서 매핑은 cascade 제거.
    같은 이름 재생성을 위해 partial unique index 가 deleted_at IS NULL 인 행만 본다.
    """
    topic = await session.get(StudyTopic, topic_id)
    topic = _verify_topic_ownership(topic, user)

    # 세션 FK SET NULL — 명시적으로 (DB ON DELETE SET NULL 은 hard delete 시에만 동작)
    await session.execute(
        update(StudySession)
        .where(
            StudySession.user_id == user.id,
            StudySession.study_topic_id == topic_id,
        )
        .values(study_topic_id=None)
    )
    # 문서 매핑 명시 제거 (soft delete 라 FK CASCADE 안 탐)
    await session.execute(
        delete(StudyTopicDocument).where(
            StudyTopicDocument.study_topic_id == topic_id,
        )
    )

    topic.deleted_at = datetime.now(timezone.utc)
    await session.commit()


# ─── 자료(문서) 매핑 ───


@router.post(
    "/{topic_id}/documents",
    response_model=StudyTopicDocumentLinkResponse,
    status_code=201,
)
async def link_documents_to_topic(
    topic_id: int,
    body: StudyTopicDocumentLinkRequest,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """자료 N개를 주제에 일괄 매핑. 이미 연결된 건 skipped_existing 으로 보고."""
    topic = await session.get(StudyTopic, topic_id)
    topic = _verify_topic_ownership(topic, user)

    requested = list(dict.fromkeys(body.document_ids))  # 중복 제거 + 순서 보존

    # 존재 + 미삭제 documents 만 통과 (single-user 시스템: documents.user_id 부재 가능)
    valid_rows = (
        await session.execute(
            select(Document.id).where(
                Document.id.in_(requested),
                Document.deleted_at.is_(None),
            )
        )
    ).scalars().all()
    valid_set = set(valid_rows)
    not_found = [d for d in requested if d not in valid_set]

    # 이미 매핑된 것
    existing_rows = (
        await session.execute(
            select(StudyTopicDocument.document_id).where(
                StudyTopicDocument.study_topic_id == topic_id,
                StudyTopicDocument.document_id.in_(valid_set),
            )
        )
    ).scalars().all()
    existing_set = set(existing_rows)
    to_link = [d for d in requested if d in valid_set and d not in existing_set]

    # 기존 max sort_order 다음부터
    max_sort = (
        await session.execute(
            select(func.coalesce(func.max(StudyTopicDocument.sort_order), -1))
            .where(StudyTopicDocument.study_topic_id == topic_id)
        )
    ).scalar() or -1

    for i, doc_id in enumerate(to_link, start=1):
        session.add(StudyTopicDocument(
            study_topic_id=topic_id,
            document_id=doc_id,
            user_id=user.id,
            sort_order=int(max_sort) + i,
        ))

    try:
        await session.commit()
    except IntegrityError:
        await session.rollback()
        raise HTTPException(status_code=409, detail="자료 매핑 중 충돌이 발생했습니다")

    return StudyTopicDocumentLinkResponse(
        linked=to_link,
        skipped_existing=sorted(existing_set),
        skipped_not_found=not_found,
    )


@router.post(
    "/{topic_id}/documents/by-path",
    response_model=StudyTopicDocumentByPathResponse,
    status_code=201,
)
async def link_documents_by_library_path(
    topic_id: int,
    body: StudyTopicDocumentByPathRequest,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """자료실 카테고리 path 기준으로 하위 자료를 일괄 매핑.
    /api/documents/library?path=... 와 동일한 user_tags(@library/...) prefix 매칭 사용.
    100건 limit 우회 — 한 카테고리에 자료가 많아도 한 번에 처리.
    """
    topic = await session.get(StudyTopic, topic_id)
    topic = _verify_topic_ownership(topic, user)

    try:
        normalized = normalize_library_path(body.path)
    except ValueError as e:
        raise HTTPException(status_code=400, detail=str(e))

    exact_tag = f"{LIBRARY_PREFIX}{normalized}"
    prefix_tag = f"{LIBRARY_PREFIX}{normalized}/%"

    # @library/<path> 정확 일치 또는 (subtree 옵션 시) 그 하위까지 prefix 매칭
    if body.include_subtree:
        tag_filter = sql_text("""
            EXISTS (
                SELECT 1 FROM jsonb_array_elements_text(documents.user_tags) AS t
                WHERE t = :exact OR t LIKE :prefix
            )
        """).bindparams(exact=exact_tag, prefix=prefix_tag)
    else:
        tag_filter = sql_text("""
            EXISTS (
                SELECT 1 FROM jsonb_array_elements_text(documents.user_tags) AS t
                WHERE t = :exact
            )
        """).bindparams(exact=exact_tag)

    candidate_rows = (
        await session.execute(
            select(Document.id).where(
                Document.deleted_at.is_(None),
                Document.category == "library",
                tag_filter,
            )
        )
    ).scalars().all()
    candidate_set = set(candidate_rows)
    total_in_path = len(candidate_set)

    if total_in_path == 0:
        return StudyTopicDocumentByPathResponse(
            linked_count=0,
            skipped_existing_count=0,
            total_in_path=0,
            path=normalized,
        )

    existing_rows = (
        await session.execute(
            select(StudyTopicDocument.document_id).where(
                StudyTopicDocument.study_topic_id == topic_id,
                StudyTopicDocument.document_id.in_(candidate_set),
            )
        )
    ).scalars().all()
    existing_set = set(existing_rows)
    to_link = [d for d in candidate_rows if d not in existing_set]

    max_sort = (
        await session.execute(
            select(func.coalesce(func.max(StudyTopicDocument.sort_order), -1))
            .where(StudyTopicDocument.study_topic_id == topic_id)
        )
    ).scalar() or -1

    for i, doc_id in enumerate(to_link, start=1):
        session.add(StudyTopicDocument(
            study_topic_id=topic_id,
            document_id=doc_id,
            user_id=user.id,
            sort_order=int(max_sort) + i,
        ))

    try:
        await session.commit()
    except IntegrityError:
        await session.rollback()
        raise HTTPException(status_code=409, detail="자료 매핑 중 충돌이 발생했습니다")

    return StudyTopicDocumentByPathResponse(
        linked_count=len(to_link),
        skipped_existing_count=len(existing_set),
        total_in_path=total_in_path,
        path=normalized,
    )


@router.delete("/{topic_id}/documents/{document_id}", status_code=204)
async def unlink_document_from_topic(
    topic_id: int,
    document_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    result = await session.execute(
        delete(StudyTopicDocument).where(
            StudyTopicDocument.study_topic_id == topic_id,
            StudyTopicDocument.document_id == document_id,
            StudyTopicDocument.user_id == user.id,
        )
    )
    await session.commit()
    if result.rowcount == 0:
        raise HTTPException(status_code=404, detail="매핑을 찾을 수 없습니다")


# ─── 세션 매핑 ───


@router.post("/{topic_id}/sessions/{session_id}", status_code=204)
async def attach_session_to_topic(
    topic_id: int,
    session_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """기존 세션을 주제에 연결. 다른 주제에 이미 연결돼 있으면 새 주제로 갱신 (1:N)."""
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    sess = await session.get(StudySession, session_id)
    if sess is None or sess.user_id != user.id:
        raise HTTPException(status_code=404, detail="학습 세션을 찾을 수 없습니다")
    sess.study_topic_id = topic_id
    sess.updated_at = datetime.now(timezone.utc)
    await session.commit()


@router.delete("/{topic_id}/sessions/{session_id}", status_code=204)
async def detach_session_from_topic(
    topic_id: int,
    session_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """세션 분리. 현재 study_topic_id 가 topic_id 가 아니면 404."""
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    sess = await session.get(StudySession, session_id)
    if sess is None or sess.user_id != user.id:
        raise HTTPException(status_code=404, detail="학습 세션을 찾을 수 없습니다")
    if sess.study_topic_id != topic_id:
        raise HTTPException(status_code=404, detail="해당 주제에 연결된 세션이 아닙니다")
    sess.study_topic_id = None
    sess.updated_at = datetime.now(timezone.utc)
    await session.commit()


# ─── PR-9: 분야 설명 (study_topic_subject_notes) ───

SUBJECT_NOTE_TIMEOUT_S = 30.0
_SUBJECT_NOTE_PROMPT_PATH = "study_subject_note.txt"
_subject_note_prompt_cache: str | None = None


def _load_subject_note_prompt() -> str:
    global _subject_note_prompt_cache
    if _subject_note_prompt_cache is None:
        prompts_dir = _Path(__file__).resolve().parent.parent / "prompts"
        _subject_note_prompt_cache = (prompts_dir / _SUBJECT_NOTE_PROMPT_PATH).read_text(encoding="utf-8")
    return _subject_note_prompt_cache


def _render_subject_note_prompt(subject: str, scope: str, doc_block: str, q_block: str) -> str:
    template = _load_subject_note_prompt()
    return (
        template
        .replace("{subject}", subject)
        .replace("{scope}", scope or "(미지정)")
        .replace("{documents_evidence_block}", doc_block)
        .replace("{questions_evidence_block}", q_block)
    )


class SubjectNoteRequest(BaseModel):
    subject: str = Field(min_length=1, max_length=120)
    scope: str = Field(default="", max_length=200)
    regenerate: bool = False


class SubjectNoteEvidence(BaseModel):
    source_type: str
    source_id: int
    title: str
    snippet: str


class SubjectNoteResponse(BaseModel):
    subject: str
    scope: str
    content: str | None
    status: str  # ready | failed | none | stale | pending
    generated_at: datetime | None
    model: str | None
    evidence: list[SubjectNoteEvidence] = []
    from_cache: bool = False
    can_regenerate: bool = True


def _note_cache_response(note: StudyTopicSubjectNote) -> SubjectNoteResponse:
    return SubjectNoteResponse(
        subject=note.subject,
        scope=note.scope,
        content=note.content,
        status=note.status,
        generated_at=note.generated_at,
        model=note.model,
        evidence=[],
        from_cache=True,
        can_regenerate=True,
    )


@router.get("/{topic_id}/subject-notes", response_model=SubjectNoteResponse)
async def get_subject_note(
    topic_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
    subject: str = Query(..., min_length=1, max_length=120),
    scope: str = Query("", max_length=200),
):
    """캐시 조회. 없으면 status='none' + content=null 응답."""
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)
    note = (
        await session.execute(
            select(StudyTopicSubjectNote).where(
                StudyTopicSubjectNote.user_id == user.id,
                StudyTopicSubjectNote.study_topic_id == topic_id,
                StudyTopicSubjectNote.subject == subject,
                StudyTopicSubjectNote.scope == scope,
            )
        )
    ).scalar_one_or_none()
    if note is None:
        return SubjectNoteResponse(
            subject=subject, scope=scope, content=None, status="none",
            generated_at=None, model=None, evidence=[], from_cache=True, can_regenerate=True,
        )
    return _note_cache_response(note)


@router.post("/{topic_id}/subject-notes/generate", response_model=SubjectNoteResponse)
async def generate_subject_note(
    topic_id: int,
    body: SubjectNoteRequest,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """분야 설명 AI 생성 + 캐시. PR-3 race-safe pending 패턴 동일.

    regenerate=false + status=ready → 캐시 반환.
    pending → 409.
    그 외 → 새 생성. 실패 시 status='failed', 직전 본문 보존.
    """
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    # upsert: 기존 행 있으면 사용, 없으면 신규
    note = (
        await session.execute(
            select(StudyTopicSubjectNote).where(
                StudyTopicSubjectNote.user_id == user.id,
                StudyTopicSubjectNote.study_topic_id == topic_id,
                StudyTopicSubjectNote.subject == body.subject,
                StudyTopicSubjectNote.scope == body.scope,
            )
        )
    ).scalar_one_or_none()

    if note is None:
        note = StudyTopicSubjectNote(
            user_id=user.id,
            study_topic_id=topic_id,
            subject=body.subject,
            scope=body.scope,
            status="none",
        )
        session.add(note)
        await session.flush()
        await session.commit()
    else:
        # 캐시 단축
        if not body.regenerate:
            if note.status == "ready":
                return _note_cache_response(note)
            if note.status == "pending":
                raise HTTPException(
                    status_code=409,
                    detail={"status": "pending", "detail": "이미 생성 중입니다"},
                )
            # none/failed/stale → 새로 생성
        else:
            if note.status == "pending":
                raise HTTPException(
                    status_code=409,
                    detail={"status": "pending", "detail": "이미 생성 중입니다"},
                )

    # race-safe pending 전이
    lock = await session.execute(
        update(StudyTopicSubjectNote)
        .where(
            StudyTopicSubjectNote.id == note.id,
            StudyTopicSubjectNote.status != "pending",
        )
        .values(status="pending", updated_at=datetime.now(timezone.utc))
        .returning(StudyTopicSubjectNote.id)
    )
    if lock.scalar_one_or_none() is None:
        raise HTTPException(
            status_code=409,
            detail={"status": "pending", "detail": "이미 생성 중입니다"},
        )
    await session.commit()

    # RAG
    try:
        ctx = await gather_subject_note_context(session, user.id, topic_id, body.subject, body.scope)
    except Exception as e:
        logger.warning("subject_note_rag_failed: %s: %s", type(e).__name__, e)
        ctx = SubjectNoteContext(documents=[], questions=[])

    doc_block = render_evidence_block(ctx.documents)
    q_block = render_evidence_block(ctx.questions)
    prompt = _render_subject_note_prompt(body.subject, body.scope, doc_block, q_block)

    ai_client = EidAIClient()  # 이드 표면 = egress 코드층 박탈(call_primary only, W4-1)
    raw_text: str | None = None
    try:
        async with acquire_mlx_gate(Priority.FOREGROUND):
            async with asyncio.timeout(SUBJECT_NOTE_TIMEOUT_S):
                # 이드 substrate(persona+rules)=system / 렌더 템플릿(지시+evidence)=user (W2-2)
                raw_text = await ai_client.call_primary(
                    prompt, system=compose("study_subject_note", task="")
                )
    except asyncio.TimeoutError:
        logger.warning("subject_note_mlx_timeout topic=%s subject=%s", topic_id, body.subject)
    except Exception:
        logger.exception("subject_note_mlx_failed topic=%s subject=%s", topic_id, body.subject)
    finally:
        await ai_client.close()

    note = await session.get(StudyTopicSubjectNote, note.id)
    if not raw_text or not raw_text.strip():
        note.status = "failed"
        note.updated_at = datetime.now(timezone.utc)
        await session.commit()
        return SubjectNoteResponse(
            subject=note.subject, scope=note.scope, content=note.content,
            status="failed", generated_at=note.generated_at, model=note.model,
            evidence=[e.to_dict() for e in ctx.all],
            from_cache=False, can_regenerate=True,
        )

    cleaned = strip_thinking(raw_text).strip()
    note.content = cleaned
    note.status = "ready"
    note.generated_at = datetime.now(timezone.utc)
    primary_name = ai_client.ai.primary.model if hasattr(ai_client.ai.primary, "model") else "primary"
    note.model = f"mlx:{primary_name}"
    note.updated_at = note.generated_at
    await session.commit()

    return SubjectNoteResponse(
        subject=note.subject, scope=note.scope, content=note.content,
        status="ready", generated_at=note.generated_at, model=note.model,
        evidence=[e.to_dict() for e in ctx.all],
        from_cache=False, can_regenerate=True,
    )


# ─── 이드 W3-2: 학습 약점 진단 (study_diagnosis surface) ───
#
# 워커(study_weakness)가 산출한 최신 eid_study_weakness 스냅샷을 '학습 진단 코치'(study overlay)
# 로 번역. 약점/태도 '판정'은 코드 derived(스냅샷) — LLM 은 스냅샷 블록 값만 인용(환각 약점 차단).
# compose("study_diagnosis") = persona+rules+study overlay(+{placeholder}) → 표면이 블록 substitute.
DIAGNOSIS_TIMEOUT_S = 40.0


class StudyDiagnosisResponse(BaseModel):
    status: str  # ready | none
    content: str | None = None
    model: str | None = None
    generated_at: datetime | None = None
    snapshot_at: datetime | None = None
    review_set_draft_id: int | None = None


@router.post("/diagnosis/generate", response_model=StudyDiagnosisResponse)
async def generate_study_diagnosis(
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """누적 학습 약점/태도 진단(학습 진단 코치). 최신 약점 스냅샷을 코치 언어로 번역만.

    워커 미가동(스냅샷 부재)이면 status='none' — '아직 진단 데이터 없음' 명시(빈약속/추측 회피).
    """
    snap = (
        await session.execute(
            select(EidStudyWeakness)
            .where(EidStudyWeakness.user_id == user.id, EidStudyWeakness.status == "active")
            .order_by(EidStudyWeakness.created_at.desc())
            .limit(1)
        )
    ).scalar_one_or_none()
    if snap is None:
        return StudyDiagnosisResponse(status="none")

    draft = (
        await session.execute(
            select(EidReviewSetDraft)
            .where(
                EidReviewSetDraft.user_id == user.id,
                EidReviewSetDraft.source_weakness_id == snap.id,  # 이 스냅샷이 산출한 draft만(W3 review #5)
            )
            .order_by(EidReviewSetDraft.created_at.desc())
            .limit(1)
        )
    ).scalar_one_or_none()

    weakness_block = format_weakness_block(
        snap.weaknesses or [], shallow_overall=snap.is_shallow_sample
    )
    if draft is not None and draft.question_ids:
        weakness_block += (
            f"\n《권장 복습세트 초안》 set #{draft.id} · {len(draft.question_ids)}문항 "
            f"(reason={draft.reason}) — 사용자 1클릭 확인 후에만 실제 편성. 자율 편성 금지."
        )
    habit_block = format_habit_block(snap.habit_signals or {})

    # compose 는 study overlay(placeholder 포함)를 system 에 넣음 → 표면이 placeholder 를 실데이터로 치환.
    composed = compose("study_diagnosis", task="")
    # fail-closed: overlay degrade(placeholder 부재)면 스냅샷 없이 LLM 돌릴 때 약점 날조 위험 →
    # 진단 생략(status='none'). weakness·habit 두 placeholder 다 확인(W3 review #4).
    if "{weakness_snapshot_block}" not in composed or "{habit_signal_block}" not in composed:
        logger.error(
            "study_diagnosis: study overlay degraded — placeholder 부재, 진단 생략(fail-closed) user=%s",
            user.id,
        )
        return StudyDiagnosisResponse(status="none")
    system = (
        composed
        .replace("{weakness_snapshot_block}", weakness_block)
        .replace("{habit_signal_block}", habit_block)
    )
    prompt = (
        "누적 학습 이력을 근거로 내 약점 토픽과 학습 태도를 진단해줘. "
        "위 《약점 스냅샷》·《태도 신호》 블록에 있는 값만 인용하고, 블록에 없는 토픽·수치·약점명은 "
        "만들지 마라. 약점 Top-N + 각 구체 근거 + (있으면) 권장 복습세트 초안을 제시하고, "
        "각 토픽의 tier 가 정한 강도를 넘기지 마라(라벨=방향, tier=긴급도)."
    )

    ai_client = EidAIClient()  # 이드 표면 = egress 코드층 박탈(call_primary only, W4-1)
    raw_text: str | None = None
    try:
        async with acquire_mlx_gate(Priority.FOREGROUND):
            async with asyncio.timeout(DIAGNOSIS_TIMEOUT_S):
                raw_text = await ai_client.call_primary(prompt, system=system)
    except asyncio.TimeoutError:
        logger.warning("study_diagnosis_mlx_timeout user=%s", user.id)
    except Exception:
        logger.exception("study_diagnosis_mlx_failed user=%s", user.id)
    finally:
        await ai_client.close()

    if not raw_text or not raw_text.strip():
        raise HTTPException(status_code=503, detail="진단 생성 실패 (LLM)")

    primary_name = ai_client.ai.primary.model if hasattr(ai_client.ai.primary, "model") else "primary"
    return StudyDiagnosisResponse(
        status="ready",
        content=strip_thinking(raw_text).strip(),
        model=f"mlx:{primary_name}",
        generated_at=datetime.now(timezone.utc),
        snapshot_at=snap.source_generated_at,
        review_set_draft_id=draft.id if draft else None,
    )


# ─── PR-10: 문제풀이 세션 (quiz_session) lifecycle ───
#
# 한 토픽당 in_progress 1개. 출제 시 session 행 생성 + question_ids 스냅샷.
# 풀이 중 attempt 마다 cursor + count 증가 (study_questions.submit_attempt 가 같은 트랜잭션에서).
# 마지막 문제 풀이 후 status='done' + finished_at 박힘.


class QuizMode(str, Enum):
    """PR-12-B: 출제 모드. 1차는 random 만. PR-12-C 에서 frequent_focus / wrong_variants 추가 예약."""
    random = "random"


class QuizSessionStartRequest(BaseModel):
    target_per_subject: int = Field(default=20, ge=1, le=100)
    subject: str | None = Field(default=None, max_length=120)
    scope: str | None = Field(default=None, max_length=200)
    wrong_only: bool = False
    abandon_existing: bool = False  # true 면 기존 in_progress 세션을 abandoned 로 마감 후 새로
    quiz_mode: QuizMode = QuizMode.random
    # Phase 1-E: bucket + stage 비율 기반 선별. stage 설정 시 기존 _select_questions_for_topic
    # 우회하고 select_questions_for_quiz 사용. size 가 명시되면 size 만큼 출제 (subject 무관).
    stage: str | None = Field(default=None, pattern="^(intro|learning|pre_exam)$")
    size: int | None = Field(default=None, ge=1, le=200)
    # Phase 2-E: 복습함에서 명시 선택된 문제로 세션. 우선순위 stage > question_ids > 기존 경로.
    # 출제 순서는 클라이언트가 보낸 그대로 보존 (사용자 인지 순서). 한도 200, 중복 제거됨.
    question_ids: list[int] | None = Field(default=None)


class QuizSessionSummary(BaseModel):
    """진행 카드 + 결과 카드 공통 (가벼운 메타). 문제 본문은 상세 endpoint 에서."""
    id: int
    status: str
    cursor: int
    total: int
    correct_count: int
    wrong_count: int
    unsure_count: int
    target_per_subject: int
    subject_filter: str | None
    wrong_only: bool
    quiz_mode: str  # PR-12-B: 세션 메타. 향후 모드별 정답률 통계용.
    subject_distribution: dict
    created_at: datetime
    updated_at: datetime
    finished_at: datetime | None
    # 결과 카드 헤더 "미확인 N건" 용 — done 세션에서만 의미.
    unreviewed_wrong_unsure_count: int = 0
    # Phase 2-B: finalize 시점 스냅샷 (DB 영속화 — 세션 종료 후 변하지 않음).
    newly_correct_count: int = 0
    relapsed_count: int = 0
    recovered_count: int = 0
    chronic_remaining_count: int = 0
    # Phase 2-B: 동적 (지금 시점) 할 일 카운트. 결과 화면에서 다른 세션이 끼어도 fresh.
    pending_review_count: int = 0
    chronic_count: int = 0
    regressed_count: int = 0


class QuizSessionListResponse(BaseModel):
    active: QuizSessionSummary | None
    recent_done: list[QuizSessionSummary]


async def _build_session_summary(
    s: StudyQuizSession,
    session: AsyncSession,
    *,
    include_progress_counts: bool = False,
) -> QuizSessionSummary:
    """status='done' 일 때 미확인 카운트 + (옵션) 지금 시점 progress 기반 할 일 카운트 계산.

    include_progress_counts=True 면 결과 화면 헤더용 동적 카운트 3종 (pending_review/chronic/regressed)
    SQL 3회 추가. 목록 endpoint 에서는 호출당 비용을 피하려고 default False.
    """
    unreviewed = 0
    if s.status == "done":
        row = (
            await session.execute(
                select(func.count())
                .select_from(StudyQuestionAttempt)
                .where(
                    StudyQuestionAttempt.quiz_session_id == s.id,
                    StudyQuestionAttempt.outcome.in_(("wrong", "unsure")),
                    StudyQuestionAttempt.reviewed_at.is_(None),
                )
            )
        ).scalar() or 0
        unreviewed = int(row)

    pending_review_count = 0
    chronic_count = 0
    regressed_count = 0
    if include_progress_counts:
        from models.study_question_progress import StudyQuestionProgress

        pr_row = (
            await session.execute(
                select(func.count())
                .select_from(StudyQuestionProgress)
                .where(
                    StudyQuestionProgress.user_id == s.user_id,
                    StudyQuestionProgress.study_topic_id == s.study_topic_id,
                    StudyQuestionProgress.last_outcome.in_(("wrong", "unsure")),
                    sql_or(
                        StudyQuestionProgress.last_reviewed_at.is_(None),
                        and_(
                            StudyQuestionProgress.last_reviewed_at.is_not(None),
                            StudyQuestionProgress.last_attempted_at.is_not(None),
                            StudyQuestionProgress.last_reviewed_at
                            < StudyQuestionProgress.last_attempted_at,
                        ),
                    ),
                )
            )
        ).scalar() or 0
        pending_review_count = int(pr_row)

        ch_row = (
            await session.execute(
                select(func.count())
                .select_from(StudyQuestionProgress)
                .where(
                    StudyQuestionProgress.user_id == s.user_id,
                    StudyQuestionProgress.study_topic_id == s.study_topic_id,
                    StudyQuestionProgress.pattern_state == "chronic_wrong",
                )
            )
        ).scalar() or 0
        chronic_count = int(ch_row)

        rg_row = (
            await session.execute(
                select(func.count())
                .select_from(StudyQuestionProgress)
                .where(
                    StudyQuestionProgress.user_id == s.user_id,
                    StudyQuestionProgress.study_topic_id == s.study_topic_id,
                    StudyQuestionProgress.pattern_state == "regressed",
                )
            )
        ).scalar() or 0
        regressed_count = int(rg_row)

    return QuizSessionSummary(
        id=s.id,
        status=s.status,
        cursor=s.cursor,
        total=len(s.question_ids or []),
        correct_count=s.correct_count,
        wrong_count=s.wrong_count,
        unsure_count=s.unsure_count,
        target_per_subject=s.target_per_subject,
        subject_filter=s.subject_filter,
        wrong_only=s.wrong_only,
        quiz_mode=s.quiz_mode,
        subject_distribution=s.subject_distribution or {},
        created_at=s.created_at,
        updated_at=s.updated_at,
        finished_at=s.finished_at,
        unreviewed_wrong_unsure_count=unreviewed,
        newly_correct_count=s.newly_correct_count,
        relapsed_count=s.relapsed_count,
        recovered_count=s.recovered_count,
        chronic_remaining_count=s.chronic_remaining_count,
        pending_review_count=pending_review_count,
        chronic_count=chronic_count,
        regressed_count=regressed_count,
    )


async def _select_questions_for_topic(
    session: AsyncSession,
    user: User,
    topic_id: int,
    *,
    subject: str | None,
    scope: str | None,
    target_per_subject: int,
    wrong_only: bool,
    apply_spacing: bool = False,
) -> tuple[list[int], dict[str, int]]:
    """과목별 균등 추출 + (옵션) PR-12-B type spacing.

    apply_spacing=True 면 subject bucket 별로 buffer × SPACING_BUFFER_RATIO 만큼
    뽑은 뒤 apply_type_spacing 으로 같은 유형 과밀 방지. 회차 무관.
    """
    from services.study.related_types import (
        SPACING_BUFFER_RATIO,
        apply_type_spacing,
        make_spacing_candidates_from_rows,
    )

    base_filter = and_(
        StudyQuestion.user_id == user.id,
        StudyQuestion.study_topic_id == topic_id,
        StudyQuestion.deleted_at.is_(None),
        StudyQuestion.is_active.is_(True),
    )

    wrong_qids: set[int] | None = None
    if wrong_only:
        latest = (
            await session.execute(
                select(
                    StudyQuestionAttempt.study_question_id,
                    StudyQuestionAttempt.is_correct,
                )
                .where(
                    StudyQuestionAttempt.user_id == user.id,
                    StudyQuestionAttempt.study_topic_id == topic_id,
                )
                .order_by(
                    StudyQuestionAttempt.study_question_id,
                    StudyQuestionAttempt.answered_at.desc(),
                )
                .distinct(StudyQuestionAttempt.study_question_id)
            )
        ).all()
        wrong_qids = {r.study_question_id for r in latest if r.is_correct is False}
        if not wrong_qids:
            return [], {}

    if subject is not None:
        subjects: list[str | None] = [subject]
    else:
        subj_query = (
            select(StudyQuestion.subject)
            .where(base_filter)
            .distinct()
        )
        if scope is not None:
            subj_query = subj_query.where(StudyQuestion.scope == scope)
        if wrong_qids is not None:
            subj_query = subj_query.where(StudyQuestion.id.in_(wrong_qids))
        subjects = [r[0] for r in (await session.execute(subj_query)).all()]
        if not subjects:
            return [], {}

    selected_ids: list[int] = []
    distribution: dict[str, int] = {}
    for subj in subjects:
        if apply_spacing:
            # spacing buffer × ratio. subject bucket 안에서만 cosine 비교.
            buffer_limit = max(target_per_subject, int(target_per_subject * SPACING_BUFFER_RATIO))
            sub_q = select(
                StudyQuestion.id,
                StudyQuestion.embedding,
                StudyQuestion.embedding_status,
            ).where(base_filter)
        else:
            sub_q = select(StudyQuestion.id).where(base_filter)
            buffer_limit = target_per_subject

        if subj is None:
            sub_q = sub_q.where(StudyQuestion.subject.is_(None))
        else:
            sub_q = sub_q.where(StudyQuestion.subject == subj)
        if scope is not None:
            sub_q = sub_q.where(StudyQuestion.scope == scope)
        if wrong_qids is not None:
            sub_q = sub_q.where(StudyQuestion.id.in_(wrong_qids))
        sub_q = sub_q.order_by(func.random()).limit(buffer_limit)
        rows = (await session.execute(sub_q)).all()

        if apply_spacing:
            # rows 가 _SpacingCandidate 의 row protocol 만족 (id/embedding/embedding_status).
            spacing_cands = make_spacing_candidates_from_rows(rows)
            spacing_res = await apply_type_spacing(
                spacing_cands,
                target=target_per_subject,
                subject_label=subj or "",
            )
            picked = spacing_res.selected_ids
        else:
            picked = [r[0] for r in rows[:target_per_subject]]

        if picked:
            selected_ids.extend(picked)
            distribution[subj or ""] = len(picked)

    _random.shuffle(selected_ids)
    return selected_ids, distribution


@router.get("/{topic_id}/quiz-sessions", response_model=QuizSessionListResponse)
async def list_quiz_sessions(
    topic_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
    limit: int = Query(20, ge=1, le=100),
):
    """진행 중 + 최근 완료 N건. 통합뷰 진행/결과 카드용."""
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    active_row = (
        await session.execute(
            select(StudyQuizSession).where(
                StudyQuizSession.user_id == user.id,
                StudyQuizSession.study_topic_id == topic_id,
                StudyQuizSession.status == "in_progress",
            )
        )
    ).scalar_one_or_none()
    active_summary = await _build_session_summary(active_row, session) if active_row else None

    done_rows = (
        await session.execute(
            select(StudyQuizSession)
            .where(
                StudyQuizSession.user_id == user.id,
                StudyQuizSession.study_topic_id == topic_id,
                StudyQuizSession.status == "done",
            )
            .order_by(StudyQuizSession.finished_at.desc().nulls_last())
            .limit(limit)
        )
    ).scalars().all()
    done = [await _build_session_summary(s, session) for s in done_rows]

    return QuizSessionListResponse(active=active_summary, recent_done=done)


@router.post("/{topic_id}/quiz-sessions", response_model=QuizSessionSummary, status_code=201)
async def start_quiz_session(
    topic_id: int,
    body: QuizSessionStartRequest,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """새 세션 시작. 기존 in_progress 가 있으면:
       - body.abandon_existing=false → 기존 세션 그대로 반환 (이어풀기).
       - body.abandon_existing=true → 기존을 abandoned 로 마감 후 새로 출제.
    """
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    existing = (
        await session.execute(
            select(StudyQuizSession).where(
                StudyQuizSession.user_id == user.id,
                StudyQuizSession.study_topic_id == topic_id,
                StudyQuizSession.status == "in_progress",
            )
        )
    ).scalar_one_or_none()

    if existing is not None:
        if not body.abandon_existing:
            # 이어풀기 — 기존 그대로 반환.
            return await _build_session_summary(existing, session)
        existing.status = "abandoned"
        existing.updated_at = datetime.now(timezone.utc)
        await session.flush()

    # Phase 1-E: stage 명시 시 bucket + 비율 기반 선별 (단일 풀이 진입점 vision).
    # Phase 2-E: question_ids 명시 시 복습함 multi-select 경로 — 검증만 + 순서 보존.
    # stage 미명시 시 기존 subject bucket + spacing (PR-12-B 호환).
    if body.stage is not None:
        from services.study.quiz_selection import select_questions_for_quiz
        size = body.size or 100
        try:
            qids, distribution = await select_questions_for_quiz(
                session,
                user_id=user.id,
                study_topic_id=topic_id,
                stage=body.stage,
                size=size,
            )
        except ValueError as e:
            raise HTTPException(status_code=400, detail=str(e))
        if not qids:
            raise HTTPException(status_code=400, detail="출제 가능한 문제가 없습니다")
    elif body.question_ids is not None:
        # 중복 제거 (순서 보존)
        seen: set[int] = set()
        wanted: list[int] = []
        for qid in body.question_ids:
            if qid not in seen:
                seen.add(qid)
                wanted.append(qid)
        if not wanted:
            raise HTTPException(status_code=400, detail="문제를 1개 이상 선택해주세요")
        if len(wanted) > 200:
            raise HTTPException(status_code=400, detail="한 세션에 최대 200문제까지 선택 가능합니다")
        # 유효성 검증 — user × topic 소속 + 미삭제. soft-deleted 또는 다른 user/topic 의 qid 는 제거.
        valid_rows = (
            await session.execute(
                select(StudyQuestion.id, StudyQuestion.subject)
                .where(
                    StudyQuestion.id.in_(wanted),
                    StudyQuestion.user_id == user.id,
                    StudyQuestion.study_topic_id == topic_id,
                    StudyQuestion.deleted_at.is_(None),
                )
            )
        ).all()
        valid_set = {r.id for r in valid_rows}
        qids = [qid for qid in wanted if qid in valid_set]
        if not qids:
            raise HTTPException(status_code=400, detail="선택한 문제가 이 주제에 존재하지 않습니다")
        # subject distribution 계산 (결과 카드 통계용)
        subject_by_id = {r.id: (r.subject or "(미분류)") for r in valid_rows}
        distribution: dict[str, int] = {}
        for qid in qids:
            sub = subject_by_id.get(qid, "(미분류)")
            distribution[sub] = distribution.get(sub, 0) + 1
    else:
        # 기존 PR-12-B 경로: subject bucket + type spacing.
        apply_spacing = body.quiz_mode == QuizMode.random
        qids, distribution = await _select_questions_for_topic(
            session,
            user,
            topic_id,
            subject=body.subject,
            scope=body.scope,
            target_per_subject=body.target_per_subject,
            wrong_only=body.wrong_only,
            apply_spacing=apply_spacing,
        )
        if not qids:
            raise HTTPException(status_code=400, detail="출제 가능한 문제가 없습니다")

    new_session = StudyQuizSession(
        user_id=user.id,
        study_topic_id=topic_id,
        target_per_subject=body.target_per_subject,
        subject_filter=body.subject,
        wrong_only=body.wrong_only,
        quiz_mode=body.quiz_mode.value,
        question_ids=qids,
        subject_distribution=distribution,
        status="in_progress",
        cursor=0,
    )
    session.add(new_session)
    try:
        await session.commit()
    except IntegrityError:
        # partial unique idx 충돌 — 동시에 다른 호출이 in_progress 만든 경우.
        await session.rollback()
        existing = (
            await session.execute(
                select(StudyQuizSession).where(
                    StudyQuizSession.user_id == user.id,
                    StudyQuizSession.study_topic_id == topic_id,
                    StudyQuizSession.status == "in_progress",
                )
            )
        ).scalar_one_or_none()
        if existing is None:
            raise HTTPException(status_code=409, detail="quiz_session 충돌 — 다시 시도하세요")
        return await _build_session_summary(existing, session)

    return await _build_session_summary(new_session, session)


class QuizSessionAttemptItem(BaseModel):
    """결과 카드 expand 시 카드별 메타 (attempt + 학습완료 상태)."""
    attempt_id: int
    question_id: int
    selected_choice: int | None
    correct_choice: int
    outcome: str
    answered_at: datetime
    reviewed_at: datetime | None


class QuizSessionAnalysisOut(BaseModel):
    """Phase 4-B v1: 결과 화면 헤더 카드용. summary_md 박혔으면 본문 표시,
    없으면 job_status / job_error_code 보고 placeholder 분기."""
    summary_md: str | None
    confidence: str | None
    generated_at: datetime | None
    is_stale: bool
    job_status: str | None       # 최신 job — 'pending'/'processing'/'completed'/'failed'/'skipped'
    job_error_code: str | None   # 최신 job 의 error_code (실패/skip 사유 노출용)


class QuizSessionDetailResponse(BaseModel):
    summary: QuizSessionSummary
    questions: list[dict]  # ReviewQuestionItem 호환 shape
    attempts: list[QuizSessionAttemptItem]  # 풀이된 것만 (cursor 까지)
    ai_session_analysis: QuizSessionAnalysisOut | None = None


def _attempt_stats_dict_default() -> dict:
    return {"attempt_count": 0, "correct_count": 0, "wrong_count": 0}


@router.get("/{topic_id}/quiz-sessions/{session_id}", response_model=QuizSessionDetailResponse)
async def get_quiz_session(
    topic_id: int,
    session_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """풀이 화면 / 결과 화면 공용. questions 는 출제 순서. attempts 는 cursor 까지 풀이된 것만."""
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    qs = await session.get(StudyQuizSession, session_id)
    if qs is None or qs.user_id != user.id or qs.study_topic_id != topic_id:
        raise HTTPException(status_code=404, detail="quiz_session 을 찾을 수 없습니다")

    qids = list(qs.question_ids or [])
    questions_payload: list[dict] = []
    if qids:
        rows = (
            await session.execute(
                select(StudyQuestion).where(StudyQuestion.id.in_(qids))
            )
        ).scalars().all()
        by_id = {q.id: q for q in rows}

        # 이미지 batch — 기존 study_questions._images_for_questions_batch 재사용 (sort_order 기준).
        from api.study_questions import _images_for_questions_batch
        images_typed = await _images_for_questions_batch(session, qids)
        images_map: dict[int, list[dict]] = {
            qid: [it.model_dump() for it in items] for qid, items in images_typed.items()
        }

        # attempt count batch (per question)
        stat_rows = (
            await session.execute(
                select(
                    StudyQuestionAttempt.study_question_id,
                    func.count().label("total"),
                    func.coalesce(
                        func.sum(case((StudyQuestionAttempt.is_correct.is_(True), 1), else_=0)),
                        0,
                    ).label("correct"),
                )
                .where(
                    StudyQuestionAttempt.user_id == user.id,
                    StudyQuestionAttempt.study_question_id.in_(qids),
                )
                .group_by(StudyQuestionAttempt.study_question_id)
            )
        ).all()
        stats_map: dict[int, dict] = {}
        for r in stat_rows:
            t = int(r.total)
            c = int(r.correct)
            stats_map[r.study_question_id] = {
                "attempt_count": t, "correct_count": c, "wrong_count": t - c
            }

        # 출제 순서 그대로 nest
        for qid in qids:
            q = by_id.get(qid)
            if q is None:
                continue
            questions_payload.append({
                "id": q.id,
                "question_text": q.question_text,
                "choices": [
                    {"number": 1, "text": q.choice_1},
                    {"number": 2, "text": q.choice_2},
                    {"number": 3, "text": q.choice_3},
                    {"number": 4, "text": q.choice_4},
                ],
                "subject": q.subject,
                "scope": q.scope,
                "exam_round": q.exam_round,
                "explanation": q.explanation,
                "stats": stats_map.get(q.id, _attempt_stats_dict_default()),
                "images": images_map.get(q.id, []),
                # done 세션 결과 화면용 — 정답 노출. in_progress 세션은 클라이언트가 cursor 까지만 표시.
                "correct_choice": q.correct_choice if qs.status == "done" else None,
                # Phase 4-A: 결과 헤더의 AI 풀이 진척 indicator 용. wrong/unsure 와 결합해 카운트.
                "ai_explanation_status": q.ai_explanation_status,
            })

    # 이 세션의 attempts (cursor 까지). 같은 question 에 여러 attempts 있을 수 있지만,
    # 이 세션 내에서는 정상 흐름상 question 당 1개. quiz_session_id 기준으로 가져옴.
    attempt_rows = (
        await session.execute(
            select(StudyQuestionAttempt)
            .where(
                StudyQuestionAttempt.user_id == user.id,
                StudyQuestionAttempt.quiz_session_id == qs.id,
            )
            .order_by(StudyQuestionAttempt.answered_at.asc())
        )
    ).scalars().all()
    attempts_payload = [
        QuizSessionAttemptItem(
            attempt_id=a.id,
            question_id=a.study_question_id,
            selected_choice=a.selected_choice,
            correct_choice=a.correct_choice,
            outcome=a.outcome,
            answered_at=a.answered_at,
            reviewed_at=a.reviewed_at,
        )
        for a in attempt_rows
    ]

    summary = await _build_session_summary(qs, session, include_progress_counts=True)

    # Phase 4-A: 결과 화면 GET fallback — finalize enqueue 누락 또는 worker 처리 전
    # 사용자가 결과 들어온 경우 같은 wrong/unsure qid 를 idempotent backfill enqueue.
    # 한 요청당 30 cap + non-blocking + debug 로그. 실패해도 GET 200 유지.
    if qs.status == "done":
        try:
            from services.study.explanation_enqueue import enqueue_explanation_for_qids
            wrong_unsure_qids = [
                a.study_question_id for a in attempt_rows
                if a.outcome in ("wrong", "unsure")
            ]
            if wrong_unsure_qids:
                res = await enqueue_explanation_for_qids(
                    session,
                    user_id=user.id,
                    qids=wrong_unsure_qids,
                    max_count=30,  # GET fallback cap
                )
                await session.commit()
                logger.debug(
                    "phase4a_get_backfill session=%s candidates=%s enqueued=%s skipped=%s",
                    qs.id, res["candidate_count"], res["enqueue_count"], res["skipped_count"],
                )
        except Exception as e:
            logger.warning(
                "phase4a_get_backfill_failed session=%s: %s: %s",
                qs.id, type(e).__name__, e,
            )

    # Phase 4-B v1: ai_session_analysis 응답 (결과 캐시 LEFT JOIN + 최신 job)
    ai_session_analysis: QuizSessionAnalysisOut | None = None
    if qs.status == "done":
        try:
            from models.study_quiz_session_analysis import StudyQuizSessionAnalysis
            from models.study_quiz_session_job import StudyQuizSessionJob

            an_row = (
                await session.execute(
                    select(StudyQuizSessionAnalysis).where(
                        StudyQuizSessionAnalysis.study_quiz_session_id == qs.id
                    )
                )
            ).scalar_one_or_none()

            latest_job = (
                await session.execute(
                    select(StudyQuizSessionJob)
                    .where(StudyQuizSessionJob.study_quiz_session_id == qs.id)
                    .order_by(StudyQuizSessionJob.id.desc())
                    .limit(1)
                )
            ).scalar_one_or_none()

            if an_row is not None or latest_job is not None:
                ai_session_analysis = QuizSessionAnalysisOut(
                    summary_md=an_row.summary_md if an_row else None,
                    confidence=an_row.confidence if an_row else None,
                    generated_at=an_row.generated_at if an_row else None,
                    is_stale=bool(an_row.is_stale) if an_row else False,
                    job_status=latest_job.status if latest_job else None,
                    job_error_code=latest_job.error_code if latest_job else None,
                )

            # GET fallback enqueue — 기존 analysis 또는 active job 없으면만 시도.
            # best-effort: 실패가 GET 응답 깨지 않게 try/except.
            try:
                if an_row is None and (latest_job is None or latest_job.status not in ("pending", "processing")):
                    from services.study.session_analysis_enqueue import enqueue_session_analysis_auto
                    res = await enqueue_session_analysis_auto(
                        session, user_id=user.id, study_quiz_session_id=qs.id,
                    )
                    await session.commit()
                    logger.debug(
                        "phase4b_get_backfill session=%s enqueued=%s",
                        qs.id, res["enqueued"],
                    )
            except Exception as e:
                logger.warning(
                    "phase4b_get_backfill_failed session=%s: %s: %s",
                    qs.id, type(e).__name__, e,
                )
        except Exception as e:
            logger.warning(
                "phase4b_get_analysis_failed session=%s: %s: %s",
                qs.id, type(e).__name__, e,
            )

    return QuizSessionDetailResponse(
        summary=summary,
        questions=questions_payload,
        attempts=attempts_payload,
        ai_session_analysis=ai_session_analysis,
    )


class QuizSessionPatchRequest(BaseModel):
    """abandon → status='abandoned'. 다른 필드 패치는 현재 안 받음."""
    action: str = Field(..., pattern="^(abandon)$")


@router.patch("/{topic_id}/quiz-sessions/{session_id}", response_model=QuizSessionSummary)
async def patch_quiz_session(
    topic_id: int,
    session_id: int,
    body: QuizSessionPatchRequest,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """현재 abandon 만 지원 — 진행 중 세션 포기 (다시 시작 시 사용)."""
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    qs = await session.get(StudyQuizSession, session_id)
    if qs is None or qs.user_id != user.id or qs.study_topic_id != topic_id:
        raise HTTPException(status_code=404, detail="quiz_session 을 찾을 수 없습니다")

    if body.action == "abandon":
        if qs.status != "in_progress":
            raise HTTPException(status_code=409, detail=f"이미 {qs.status} 상태입니다")
        qs.status = "abandoned"
        qs.updated_at = datetime.now(timezone.utc)
        await session.commit()
    return await _build_session_summary(qs, session)


# ─── Phase 4-B v1: 세션 분석 재생성 ───


class RegenerateSummaryResponse(BaseModel):
    enqueued: bool
    reason: str | None = None  # insufficient_attempts / already_active / not_done / not_found / race_lost


@router.post(
    "/{topic_id}/quiz-sessions/{session_id}/regenerate-summary",
    response_model=RegenerateSummaryResponse,
)
async def regenerate_session_summary(
    topic_id: int,
    session_id: int,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """사용자 [재생성] 버튼 — wrong/unsure < 5 즉시 차단 + active job 차단 + is_stale 처리.

    Plan ~/.claude/plans/nifty-sparking-spindle.md (Phase 4-B v1) 의 Manual 트리거 정책 그대로.
    """
    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    qs = await session.get(StudyQuizSession, session_id)
    if qs is None or qs.user_id != user.id or qs.study_topic_id != topic_id:
        raise HTTPException(status_code=404, detail="quiz_session 을 찾을 수 없습니다")

    from services.study.session_analysis_enqueue import request_session_analysis_regenerate
    res = await request_session_analysis_regenerate(
        session, user_id=user.id, study_quiz_session_id=session_id,
    )
    await session.commit()
    return RegenerateSummaryResponse(enqueued=res["enqueued"], reason=res.get("reason"))


# ─── PR-12-A: 반복 출제 / 유사 유형 배치 카운트 ───


class RelatedTypesBulkRequest(BaseModel):
    question_ids: list[int] = Field(..., min_length=1, max_length=200)


class RelatedTypesBulkItem(BaseModel):
    repeat_related_count: int
    repeat_round_count: int
    repeat_grade: str | None = None
    similar_related_count: int
    similar_round_count: int


class RelatedTypesBulkResponse(BaseModel):
    # 입력 question_ids 전체 보존 — 권한 없음/임베딩 미준비/회차 미지정도 (0,0,0,0).
    items: dict[int, RelatedTypesBulkItem]


@router.post("/{topic_id}/related-types-bulk", response_model=RelatedTypesBulkResponse)
async def related_types_bulk(
    topic_id: int,
    body: RelatedTypesBulkRequest,
    user: Annotated[User, Depends(get_current_user)],
    session: Annotated[AsyncSession, Depends(get_session)],
):
    """카드별 배지(round_count) 표시용 배치 카운트.

    비교 대상 = 같은 topic 안 모든 ready 문제 (입력 qid 끼리만 비교 X).
    응답 dict 는 입력 qid 전체 보존 — 누락 X.
    """
    from services.study.related_types import classify_related_bulk

    topic = await session.get(StudyTopic, topic_id)
    _verify_topic_ownership(topic, user)

    counts = await classify_related_bulk(
        session,
        user_id=user.id,
        study_topic_id=topic_id,
        question_ids=body.question_ids,
    )
    return RelatedTypesBulkResponse(
        items={
            qid: RelatedTypesBulkItem(
                repeat_related_count=c.repeat_related_count,
                repeat_round_count=c.repeat_round_count,
                repeat_grade=c.repeat_grade,
                similar_related_count=c.similar_related_count,
                similar_round_count=c.similar_round_count,
            )
            for qid, c in counts.items()
        }
    )