document-server/backend/src/api/routes/documents.py

"""
문서 관리 API 라우터
"""
from fastapi import APIRouter, Depends, HTTPException, status, UploadFile, File, Form
from sqlalchemy.ext.asyncio import AsyncSession
from sqlalchemy import select, delete, and_, or_, update
from sqlalchemy.orm import selectinload
from typing import List, Optional
import os
import uuid
from uuid import UUID
import aiofiles
from pathlib import Path

from ...core.database import get_db
from ...core.config import settings
from ...models.user import User
from ...models.document import Document, Tag
from ...models.book import Book
from ..dependencies import get_current_active_user, get_current_admin_user
from pydantic import BaseModel
from datetime import datetime


class DocumentResponse(BaseModel):
    """문서 응답"""
    id: str
    title: str
    description: Optional[str]
    html_path: Optional[str]  # PDF만 업로드하는 경우 None 가능
    pdf_path: Optional[str]
    thumbnail_path: Optional[str]
    file_size: Optional[int]
    page_count: Optional[int]
    language: str
    is_public: bool
    is_processed: bool
    created_at: datetime
    updated_at: Optional[datetime]
    document_date: Optional[datetime]
    uploader_name: Optional[str]
    tags: List[str] = []
    
    # 서적 정보
    book_id: Optional[str] = None
    book_title: Optional[str] = None
    book_author: Optional[str] = None
    
    # 소분류 정보
    category_id: Optional[str] = None
    category_name: Optional[str] = None
    sort_order: int = 0
    
    # PDF 매칭 정보
    matched_pdf_id: Optional[str] = None

    class Config:
        from_attributes = True


class TagResponse(BaseModel):
    """태그 응답"""
    id: str
    name: str
    color: str
    description: Optional[str]
    document_count: int = 0

    class Config:
        from_attributes = True


class CreateTagRequest(BaseModel):
    """태그 생성 요청"""
    name: str
    color: str = "#3B82F6"
    description: Optional[str] = None


router = APIRouter()


@router.get("/", response_model=List[DocumentResponse])
async def list_documents(
    skip: int = 0,
    limit: int = 50,
    tag: Optional[str] = None,
    search: Optional[str] = None,
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """문서 목록 조회"""
    query = select(Document).options(
        selectinload(Document.uploader),
        selectinload(Document.tags),
        selectinload(Document.book),  # 서적 정보 추가
        selectinload(Document.category)  # 소분류 정보 추가
    )
    
    # 권한 필터링 (관리자가 아니면 공개 문서 + 자신이 업로드한 문서만)
    if not current_user.is_admin:
        query = query.where(
            or_(
                Document.is_public == True,
                Document.uploaded_by == current_user.id
            )
        )
    
    # 태그 필터링
    if tag:
        query = query.join(Document.tags).where(Tag.name == tag)
    
    # 검색 필터링
    if search:
        query = query.where(
            or_(
                Document.title.ilike(f"%{search}%"),
                Document.description.ilike(f"%{search}%")
            )
        )
    
    query = query.order_by(Document.created_at.desc()).offset(skip).limit(limit)
    
    result = await db.execute(query)
    documents = result.scalars().all()
    
    # 응답 데이터 변환
    response_data = []
    for doc in documents:
        doc_data = DocumentResponse(
            id=str(doc.id),
            title=doc.title,
            description=doc.description,
            html_path=doc.html_path,  # None 가능 (PDF만 업로드한 경우)
            pdf_path=doc.pdf_path,
            thumbnail_path=doc.thumbnail_path,
            file_size=doc.file_size,
            page_count=doc.page_count,
            language=doc.language,
            is_public=doc.is_public,
            is_processed=doc.is_processed,
            created_at=doc.created_at,
            updated_at=doc.updated_at,
            document_date=doc.document_date,
            uploader_name=doc.uploader.full_name or doc.uploader.email,
            tags=[tag.name for tag in doc.tags],
            # 서적 정보 추가
            book_id=str(doc.book.id) if doc.book else None,
            book_title=doc.book.title if doc.book else None,
            book_author=doc.book.author if doc.book else None,
            # 소분류 정보 추가
            category_id=str(doc.category.id) if doc.category else None,
            category_name=doc.category.name if doc.category else None,
            sort_order=doc.sort_order,
            # PDF 매칭 정보 추가
            matched_pdf_id=str(doc.matched_pdf_id) if doc.matched_pdf_id else None
        )
        response_data.append(doc_data)
    
    return response_data


@router.get("/hierarchy/structured", response_model=dict)
async def get_documents_by_hierarchy(
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """계층구조별 문서 조회 (서적 > 소분류 > 문서)"""
    # 모든 문서 조회 (기존 로직 재사용)
    query = select(Document).options(
        selectinload(Document.uploader),
        selectinload(Document.tags),
        selectinload(Document.book),
        selectinload(Document.category)
    )
    
    # 권한 필터링
    if not current_user.is_admin:
        query = query.where(
            or_(
                Document.is_public == True,
                Document.uploaded_by == current_user.id
            )
        )
    
    # 정렬: 서적별 > 소분류별 > 문서 순서별
    query = query.order_by(
        Document.book_id.nulls_last(),  # 서적 있는 것 먼저
        Document.category_id.nulls_last(),  # 소분류 있는 것 먼저
        Document.sort_order,
        Document.created_at.desc()
    )
    
    result = await db.execute(query)
    documents = result.scalars().all()
    
    # 계층구조로 그룹화
    hierarchy = {
        "books": {},  # 서적별 그룹
        "uncategorized": []  # 미분류 문서들
    }
    
    for doc in documents:
        doc_data = {
            "id": str(doc.id),
            "title": doc.title,
            "description": doc.description,
            "created_at": doc.created_at.isoformat(),
            "uploader_name": doc.uploader.full_name or doc.uploader.email,
            "tags": [tag.name for tag in doc.tags],
            "sort_order": doc.sort_order,
            "book_id": str(doc.book.id) if doc.book else None,
            "book_title": doc.book.title if doc.book else None,
            "category_id": str(doc.category.id) if doc.category else None,
            "category_name": doc.category.name if doc.category else None
        }
        
        if doc.book:
            # 서적이 있는 문서
            book_id = str(doc.book.id)
            if book_id not in hierarchy["books"]:
                hierarchy["books"][book_id] = {
                    "id": book_id,
                    "title": doc.book.title,
                    "author": doc.book.author,
                    "categories": {},
                    "uncategorized_documents": []
                }
            
            if doc.category:
                # 소분류가 있는 문서
                category_id = str(doc.category.id)
                if category_id not in hierarchy["books"][book_id]["categories"]:
                    hierarchy["books"][book_id]["categories"][category_id] = {
                        "id": category_id,
                        "name": doc.category.name,
                        "documents": []
                    }
                hierarchy["books"][book_id]["categories"][category_id]["documents"].append(doc_data)
            else:
                # 서적은 있지만 소분류가 없는 문서
                hierarchy["books"][book_id]["uncategorized_documents"].append(doc_data)
        else:
            # 서적이 없는 미분류 문서
            hierarchy["uncategorized"].append(doc_data)
    
    return hierarchy


@router.post("/", response_model=DocumentResponse)
async def upload_document(
    title: str = Form(...),
    description: Optional[str] = Form(None),
    document_date: Optional[str] = Form(None),
    language: Optional[str] = Form("ko"),
    is_public: bool = Form(False),
    tags: Optional[List[str]] = Form(None),  # 태그 리스트
    book_id: Optional[str] = Form(None),  # 서적 ID 추가
    html_file: UploadFile = File(...),
    pdf_file: Optional[UploadFile] = File(None),
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """문서 업로드"""
    # 파일 확장자 확인 (HTML 또는 PDF 허용)
    file_extension = html_file.filename.lower()
    is_pdf_file = file_extension.endswith('.pdf')
    is_html_file = file_extension.endswith(('.html', '.htm'))
    
    if not (is_html_file or is_pdf_file):
        raise HTTPException(
            status_code=status.HTTP_400_BAD_REQUEST,
            detail="Only HTML and PDF files are allowed"
        )
    
    if pdf_file and not pdf_file.filename.lower().endswith('.pdf'):
        raise HTTPException(
            status_code=status.HTTP_400_BAD_REQUEST,
            detail="Only PDF files are allowed for the original document"
        )
    
    # 고유 파일명 생성
    doc_id = str(uuid.uuid4())
    
    # 메인 파일 처리 (HTML 또는 PDF) - 폴더 분리
    if is_pdf_file:
        main_filename = f"{doc_id}.pdf"
        pdf_dir = os.path.join(settings.UPLOAD_DIR, "pdfs")
        os.makedirs(pdf_dir, exist_ok=True)  # PDF 폴더 생성
        main_path = os.path.join(pdf_dir, main_filename)
        html_path = None  # PDF만 업로드하는 경우 html_path는 None
        pdf_path = main_path  # PDF 파일인 경우 pdf_path에 저장
    else:
        main_filename = f"{doc_id}.html"
        html_dir = os.path.join(settings.UPLOAD_DIR, "documents")
        os.makedirs(html_dir, exist_ok=True)  # HTML 폴더 생성
        main_path = os.path.join(html_dir, main_filename)
        html_path = main_path
        pdf_path = None
    
    # 추가 PDF 파일 처리 (HTML 파일과 함께 업로드된 경우)
    additional_pdf_path = None
    if pdf_file:
        additional_pdf_filename = f"{doc_id}_additional.pdf"
        pdf_dir = os.path.join(settings.UPLOAD_DIR, "pdfs")
        os.makedirs(pdf_dir, exist_ok=True)  # PDF 폴더 생성
        additional_pdf_path = os.path.join(pdf_dir, additional_pdf_filename)
    
    try:
        # 메인 파일 저장 (HTML 또는 PDF)
        async with aiofiles.open(main_path, 'wb') as f:
            content = await html_file.read()
            await f.write(content)
        
        # 추가 PDF 파일 저장 (HTML과 함께 업로드된 경우)
        if pdf_file and additional_pdf_path:
            async with aiofiles.open(additional_pdf_path, 'wb') as f:
                additional_content = await pdf_file.read()
                await f.write(additional_content)
            # HTML 파일인 경우 추가 PDF를 pdf_path로 설정
            if is_html_file:
                pdf_path = additional_pdf_path
        
        # 서적 ID 검증 (있는 경우)
        validated_book_id = None
        if book_id:
            try:
                # UUID 형식 검증 및 서적 존재 확인
                from uuid import UUID
                book_uuid = UUID(book_id)
                book_result = await db.execute(select(Book).where(Book.id == book_uuid))
                book = book_result.scalar_one_or_none()
                if book:
                    validated_book_id = book_uuid
            except (ValueError, Exception):
                # 잘못된 UUID 형식이거나 서적이 없으면 무시
                pass

        # 문서 메타데이터 생성
        document = Document(
            id=doc_id,
            title=title,
            description=description,
            html_path=html_path,
            pdf_path=pdf_path,
            language=language,
            file_size=len(content),  # HTML 파일 크기
            uploaded_by=current_user.id,
            original_filename=html_file.filename,
            is_public=is_public,
            document_date=datetime.fromisoformat(document_date) if document_date else None,
            book_id=validated_book_id  # 서적 ID 추가
        )
        
        db.add(document)
        await db.flush()  # ID 생성을 위해
        
        # 태그 처리
        if tags:
            tag_names = [tag.strip() for tag in tags.split(',') if tag.strip()]
            for tag_name in tag_names:
                # 기존 태그 찾기 또는 생성
                result = await db.execute(select(Tag).where(Tag.name == tag_name))
                tag = result.scalar_one_or_none()
                
                if not tag:
                    tag = Tag(
                        name=tag_name,
                        created_by=current_user.id
                    )
                    db.add(tag)
                    await db.flush()
                
                document.tags.append(tag)
        
        await db.commit()
        
        # 문서 정보를 다시 로드 (태그 포함)
        result = await db.execute(
            select(Document)
            .options(selectinload(Document.tags))
            .where(Document.id == document.id)
        )
        document_with_tags = result.scalar_one()
        
        # 응답 데이터 생성
        return DocumentResponse(
            id=str(document_with_tags.id),
            title=document_with_tags.title,
            description=document_with_tags.description,
            html_path=document_with_tags.html_path,
            pdf_path=document_with_tags.pdf_path,
            thumbnail_path=document_with_tags.thumbnail_path,
            file_size=document_with_tags.file_size,
            page_count=document_with_tags.page_count,
            language=document_with_tags.language,
            is_public=document_with_tags.is_public,
            is_processed=document_with_tags.is_processed,
            created_at=document_with_tags.created_at,
            updated_at=document_with_tags.updated_at,
            document_date=document_with_tags.document_date,
            uploader_name=current_user.full_name or current_user.email,
            tags=[tag.name for tag in document_with_tags.tags],
            matched_pdf_id=str(document_with_tags.matched_pdf_id) if document_with_tags.matched_pdf_id else None
        )
    
    except Exception as e:
        # 파일 정리
        if os.path.exists(main_path):
            os.remove(main_path)
        if additional_pdf_path and os.path.exists(additional_pdf_path):
            os.remove(additional_pdf_path)
        
        raise HTTPException(
            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
            detail=f"Failed to upload document: {str(e)}"
        )


@router.get("/{document_id}", response_model=DocumentResponse)
async def get_document(
    document_id: str,
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """문서 상세 조회"""
    result = await db.execute(
        select(Document)
        .options(selectinload(Document.uploader), selectinload(Document.tags))
        .where(Document.id == document_id)
    )
    document = result.scalar_one_or_none()
    
    if not document:
        raise HTTPException(
            status_code=status.HTTP_404_NOT_FOUND,
            detail="Document not found"
        )
    
    # 권한 확인
    if not document.is_public and document.uploaded_by != current_user.id and not current_user.is_admin:
        raise HTTPException(
            status_code=status.HTTP_403_FORBIDDEN,
            detail="Not enough permissions"
        )
    
    return DocumentResponse(
        id=str(document.id),
        title=document.title,
        description=document.description,
        html_path=document.html_path,
        pdf_path=document.pdf_path,
        thumbnail_path=document.thumbnail_path,
        file_size=document.file_size,
        page_count=document.page_count,
        language=document.language,
        is_public=document.is_public,
        is_processed=document.is_processed,
        created_at=document.created_at,
        updated_at=document.updated_at,
        document_date=document.document_date,
        uploader_name=document.uploader.full_name or document.uploader.email,
        tags=[tag.name for tag in document.tags],
        matched_pdf_id=str(document.matched_pdf_id) if document.matched_pdf_id else None
    )


@router.get("/{document_id}/content")
async def get_document_content(
    document_id: str,
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """문서 HTML 콘텐츠 조회"""
    try:
        doc_uuid = UUID(document_id)
    except ValueError:
        raise HTTPException(status_code=400, detail="Invalid document ID format")
    
    # 문서 조회
    query = select(Document).where(Document.id == doc_uuid)
    result = await db.execute(query)
    document = result.scalar_one_or_none()
    
    if not document:
        raise HTTPException(status_code=404, detail="Document not found")
    
    # 권한 확인
    if not current_user.is_admin and not document.is_public and document.uploaded_by != current_user.id:
        raise HTTPException(status_code=403, detail="Access denied")
    
    # HTML 파일 읽기
    import os
    from fastapi.responses import HTMLResponse
    
    # html_path는 이미 전체 경로를 포함하고 있음
    file_path = document.html_path
    
    if not os.path.exists(file_path):
        raise HTTPException(status_code=404, detail="Document file not found")
    
    try:
        with open(file_path, 'r', encoding='utf-8') as f:
            content = f.read()
        return HTMLResponse(content=content)
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Error reading document: {str(e)}")


class UpdateDocumentRequest(BaseModel):
    """문서 업데이트 요청"""
    title: Optional[str] = None
    description: Optional[str] = None
    sort_order: Optional[int] = None
    matched_pdf_id: Optional[str] = None
    is_public: Optional[bool] = None
    tags: Optional[List[str]] = None


@router.put("/{document_id}", response_model=DocumentResponse)
async def update_document(
    document_id: str,
    update_data: UpdateDocumentRequest,
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """문서 정보 업데이트"""
    try:
        doc_uuid = UUID(document_id)
    except ValueError:
        raise HTTPException(status_code=400, detail="Invalid document ID format")
    
    # 문서 조회
    result = await db.execute(
        select(Document)
        .options(selectinload(Document.tags), selectinload(Document.uploader), selectinload(Document.book))
        .where(Document.id == doc_uuid)
    )
    document = result.scalar_one_or_none()
    
    if not document:
        raise HTTPException(
            status_code=status.HTTP_404_NOT_FOUND,
            detail="Document not found"
        )
    
    # 권한 확인 (관리자이거나 문서 소유자)
    if not current_user.is_admin and document.uploaded_by != current_user.id:
        raise HTTPException(
            status_code=status.HTTP_403_FORBIDDEN,
            detail="Not enough permissions to update this document"
        )
    
    # 업데이트할 필드들 적용
    update_fields = update_data.model_dump(exclude_unset=True)
    
    for field, value in update_fields.items():
        if field == "matched_pdf_id":
            # PDF 매칭 처리
            if value:
                try:
                    pdf_uuid = UUID(value)
                    # PDF 문서가 실제로 존재하는지 확인
                    pdf_result = await db.execute(select(Document).where(Document.id == pdf_uuid))
                    pdf_doc = pdf_result.scalar_one_or_none()
                    if pdf_doc:
                        setattr(document, field, pdf_uuid)
                except ValueError:
                    # 잘못된 UUID 형식이면 무시
                    pass
            else:
                # None으로 설정하여 매칭 해제
                setattr(document, field, None)
        elif field == "tags":
            # 태그 처리
            if value is not None:
                # 기존 태그 관계 제거
                document.tags.clear()
                
                # 새 태그 추가
                for tag_name in value:
                    tag_name = tag_name.strip()
                    if tag_name:
                        # 기존 태그 찾기 또는 생성
                        tag_result = await db.execute(select(Tag).where(Tag.name == tag_name))
                        tag = tag_result.scalar_one_or_none()
                        
                        if not tag:
                            tag = Tag(
                                name=tag_name,
                                created_by=current_user.id
                            )
                            db.add(tag)
                            await db.flush()
                        
                        document.tags.append(tag)
        else:
            # 일반 필드 업데이트
            setattr(document, field, value)
    
    # 업데이트 시간 갱신
    document.updated_at = datetime.utcnow()
    
    await db.commit()
    await db.refresh(document)
    
    # 응답 데이터 생성
    return DocumentResponse(
        id=str(document.id),
        title=document.title,
        description=document.description,
        html_path=document.html_path,
        pdf_path=document.pdf_path,
        thumbnail_path=document.thumbnail_path,
        file_size=document.file_size,
        page_count=document.page_count,
        language=document.language,
        is_public=document.is_public,
        is_processed=document.is_processed,
        created_at=document.created_at,
        updated_at=document.updated_at,
        document_date=document.document_date,
        uploader_name=document.uploader.full_name or document.uploader.email,
        tags=[tag.name for tag in document.tags],
        book_id=str(document.book.id) if document.book else None,
        book_title=document.book.title if document.book else None,
        book_author=document.book.author if document.book else None,
        sort_order=document.sort_order,
        matched_pdf_id=str(document.matched_pdf_id) if document.matched_pdf_id else None
    )


@router.get("/{document_id}/download")
async def download_document(
    document_id: str,
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """문서 파일 다운로드"""
    try:
        doc_uuid = UUID(document_id)
    except ValueError:
        raise HTTPException(status_code=400, detail="Invalid document ID format")
    
    # 문서 조회
    query = select(Document).where(Document.id == doc_uuid)
    result = await db.execute(query)
    document = result.scalar_one_or_none()
    
    if not document:
        raise HTTPException(status_code=404, detail="Document not found")
    
    # 권한 확인
    if not current_user.is_admin and not document.is_public and document.uploaded_by != current_user.id:
        raise HTTPException(status_code=403, detail="Access denied")
    
    # 다운로드할 파일 경로 결정 (PDF 우선, 없으면 HTML)
    file_path = document.pdf_path if document.pdf_path else document.html_path
    
    if not file_path or not os.path.exists(file_path):
        raise HTTPException(status_code=404, detail="Document file not found")
    
    # 파일 응답
    from fastapi.responses import FileResponse
    
    # 파일명 설정
    filename = document.original_filename
    if not filename:
        extension = '.pdf' if document.pdf_path else '.html'
        filename = f"{document.title}{extension}"
    
    return FileResponse(
        path=file_path,
        filename=filename,
        media_type='application/octet-stream'
    )


@router.get("/{document_id}/navigation")
async def get_document_navigation(
    document_id: str,
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """문서 네비게이션 정보 조회 (이전/다음 문서)"""
    # 현재 문서 조회
    result = await db.execute(select(Document).where(Document.id == document_id))
    current_doc = result.scalar_one_or_none()
    
    if not current_doc:
        raise HTTPException(
            status_code=status.HTTP_404_NOT_FOUND,
            detail="Document not found"
        )
    
    # 권한 확인
    if not current_doc.is_public and current_doc.uploaded_by != current_user.id and not current_user.is_admin:
        raise HTTPException(
            status_code=status.HTTP_403_FORBIDDEN,
            detail="Access denied"
        )
    
    navigation_info = {
        "current": {
            "id": str(current_doc.id),
            "title": current_doc.title,
            "sort_order": current_doc.sort_order
        },
        "previous": None,
        "next": None,
        "book_info": None
    }
    
    # 서적에 속한 문서인 경우 이전/다음 문서 조회
    if current_doc.book_id:
        # 같은 서적의 HTML 문서들만 조회 (PDF 제외)
        book_docs_result = await db.execute(
            select(Document)
            .where(
                and_(
                    Document.book_id == current_doc.book_id,
                    Document.html_path.isnot(None),  # HTML 문서만
                    or_(Document.is_public == True, Document.uploaded_by == current_user.id, current_user.is_admin == True)
                )
            )
            .order_by(Document.sort_order.asc().nulls_last(), Document.created_at.asc())
        )
        book_docs = book_docs_result.scalars().all()
        
        # 현재 문서의 인덱스 찾기
        current_index = None
        for i, doc in enumerate(book_docs):
            if doc.id == current_doc.id:
                current_index = i
                break
        
        if current_index is not None:
            # 이전 문서
            if current_index > 0:
                prev_doc = book_docs[current_index - 1]
                navigation_info["previous"] = {
                    "id": str(prev_doc.id),
                    "title": prev_doc.title,
                    "sort_order": prev_doc.sort_order
                }
            
            # 다음 문서
            if current_index < len(book_docs) - 1:
                next_doc = book_docs[current_index + 1]
                navigation_info["next"] = {
                    "id": str(next_doc.id),
                    "title": next_doc.title,
                    "sort_order": next_doc.sort_order
                }
        
        # 서적 정보 추가
        from ...models.book import Book
        book_result = await db.execute(select(Book).where(Book.id == current_doc.book_id))
        book = book_result.scalar_one_or_none()
        if book:
            navigation_info["book_info"] = {
                "id": str(book.id),
                "title": book.title,
                "author": book.author
            }
    
    return navigation_info


@router.delete("/{document_id}")
async def delete_document(
    document_id: str,
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """문서 삭제"""
    result = await db.execute(select(Document).where(Document.id == document_id))
    document = result.scalar_one_or_none()
    
    if not document:
        raise HTTPException(
            status_code=status.HTTP_404_NOT_FOUND,
            detail="Document not found"
        )
    
    # 권한 확인 (관리자만)
    if not current_user.is_admin:
        raise HTTPException(
            status_code=status.HTTP_403_FORBIDDEN,
            detail="Only administrators can delete documents"
        )
    
    # 파일 삭제
    if document.html_path and os.path.exists(document.html_path):
        os.remove(document.html_path)
    if document.pdf_path and os.path.exists(document.pdf_path):
        os.remove(document.pdf_path)
    if document.thumbnail_path and os.path.exists(document.thumbnail_path):
        os.remove(document.thumbnail_path)
    
    # 관련 데이터 안전하게 삭제 (외래키 제약 조건 해결)
    from ...models.highlight import Highlight
    from ...models.note import Note
    from ...models.bookmark import Bookmark
    
    try:
        print(f"DEBUG: Starting deletion of document {document_id}")
        
        # 0. PDF 참조 해제 (외래키 제약조건 해결)
        # 이 문서를 matched_pdf_id로 참조하는 모든 문서의 참조를 NULL로 설정
        await db.execute(
            update(Document)
            .where(Document.matched_pdf_id == document_id)
            .values(matched_pdf_id=None)
        )
        print(f"DEBUG: Cleared matched_pdf_id references to document {document_id}")
        
        # 1. 먼저 해당 문서의 모든 하이라이트 ID 조회
        highlight_ids_result = await db.execute(select(Highlight.id).where(Highlight.document_id == document_id))
        highlight_ids = [row[0] for row in highlight_ids_result.fetchall()]
        print(f"DEBUG: Found {len(highlight_ids)} highlights to delete")
        
        # 2. 하이라이트에 연결된 모든 메모 삭제
        total_notes_deleted = 0
        for highlight_id in highlight_ids:
            note_result = await db.execute(delete(Note).where(Note.highlight_id == highlight_id))
            total_notes_deleted += note_result.rowcount
        print(f"DEBUG: Deleted {total_notes_deleted} notes by highlight_id")
        
        # 3. document_id로 직접 연결된 메모도 삭제 (혹시 있다면)
        direct_note_result = await db.execute(delete(Note).where(Note.document_id == document_id))
        print(f"DEBUG: Deleted {direct_note_result.rowcount} notes by document_id")
        
        # 4. 북마크 삭제
        bookmark_result = await db.execute(delete(Bookmark).where(Bookmark.document_id == document_id))
        print(f"DEBUG: Deleted {bookmark_result.rowcount} bookmarks")
        
        # 5. 하이라이트 삭제 (이제 메모가 모두 삭제되었으므로 안전)
        highlight_result = await db.execute(delete(Highlight).where(Highlight.document_id == document_id))
        print(f"DEBUG: Deleted {highlight_result.rowcount} highlights")
        
        # 6. 문서-태그 관계는 SQLAlchemy가 자동으로 처리
        
        # 7. 마지막으로 문서 삭제
        doc_result = await db.execute(delete(Document).where(Document.id == document_id))
        print(f"DEBUG: Deleted {doc_result.rowcount} documents")
        
        # 8. 커밋
        await db.commit()
        print(f"DEBUG: Successfully deleted document {document_id}")
        
    except Exception as e:
        print(f"ERROR: Failed to delete document {document_id}: {e}")
        await db.rollback()
        raise HTTPException(
            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
            detail=f"Failed to delete document: {str(e)}"
        )
    
    return {"message": "Document deleted successfully"}


@router.get("/tags/", response_model=List[TagResponse])
async def list_tags(
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """태그 목록 조회"""
    result = await db.execute(select(Tag).order_by(Tag.name))
    tags = result.scalars().all()
    
    # 각 태그의 문서 수 계산
    response_data = []
    for tag in tags:
        # 문서 수 계산 (권한 고려)
        doc_query = select(Document).join(Document.tags).where(Tag.id == tag.id)
        if not current_user.is_admin:
            doc_query = doc_query.where(
                or_(
                    Document.is_public == True,
                    Document.uploaded_by == current_user.id
                )
            )
        doc_result = await db.execute(doc_query)
        document_count = len(doc_result.scalars().all())
        
        tag_data = TagResponse(
            id=str(tag.id),
            name=tag.name,
            color=tag.color,
            description=tag.description,
            document_count=document_count
        )
        response_data.append(tag_data)
    
    return response_data


@router.post("/tags/", response_model=TagResponse)
async def create_tag(
    tag_data: CreateTagRequest,
    current_user: User = Depends(get_current_active_user),
    db: AsyncSession = Depends(get_db)
):
    """태그 생성"""
    # 중복 확인
    result = await db.execute(select(Tag).where(Tag.name == tag_data.name))
    existing_tag = result.scalar_one_or_none()
    
    if existing_tag:
        raise HTTPException(
            status_code=status.HTTP_400_BAD_REQUEST,
            detail="Tag already exists"
        )
    
    # 태그 생성
    tag = Tag(
        name=tag_data.name,
        color=tag_data.color,
        description=tag_data.description,
        created_by=current_user.id
    )
    
    db.add(tag)
    await db.commit()
    await db.refresh(tag)
    
    return TagResponse(
        id=str(tag.id),
        name=tag.name,
        color=tag.color,
        description=tag.description,
        document_count=0
    )