ChingTech OS：Line Bot 文件讀取功能（Word/Excel/PDF）

READABLE_FILE_EXTENSIONS = {
    # 純文字格式
    ".txt", ".md", ".json", ".csv", ".log",
    ".xml", ".yaml", ".yml",
    # Office 文件（透過 document_reader 解析）
    ".docx", ".xlsx", ".pptx",
    # PDF 文件（透過 document_reader 解析）
    ".pdf",
}

@dataclass
class DocumentContent:
    """文件解析結果"""
    text: str                      # 提取的純文字內容
    format: str                    # 原始格式 (docx, xlsx, pptx, pdf)
    page_count: Optional[int]      # 頁數（PDF）或工作表數（Excel）
    metadata: dict                 # 額外資訊（標題、作者等）
    truncated: bool                # 是否因大小限制而截斷
    error: Optional[str]           # 解析錯誤訊息（部分成功時）

def extract_text(file_path: str) -> DocumentContent:
    path = Path(file_path)

    # 1. 檢查檔案存在
    if not path.exists():
        raise FileNotFoundError(f"檔案不存在: {file_path}")

    ext = path.suffix.lower()

    # 2. 檢查舊版格式
    if ext in LEGACY_EXTENSIONS:
        raise UnsupportedFormatError(
            f"不支援舊版格式 {ext}，請轉存為新版格式"
        )

    # 3. 檢查支援的格式
    if ext not in SUPPORTED_EXTENSIONS:
        raise UnsupportedFormatError(f"不支援的檔案格式: {ext}")

    # 4. 檢查檔案大小
    file_size = path.stat().st_size
    max_size = MAX_FILE_SIZE.get(ext, 10 * 1024 * 1024)
    if file_size > max_size:
        raise FileTooLargeError(...)

    # 5. 根據格式分派解析器
    extractors = {
        ".docx": _extract_docx,
        ".xlsx": _extract_xlsx,
        ".pptx": _extract_pptx,
        ".pdf": _extract_pdf,
    }
    return extractors[ext](file_path)

def _extract_docx(file_path: str) -> DocumentContent:
    doc = Document(file_path)
    paragraphs = []

    # 提取段落
    for para in doc.paragraphs:
        if para.text.strip():
            paragraphs.append(para.text)

    # 提取表格
    for table in doc.tables:
        for row in table.rows:
            row_text = " | ".join(cell.text.strip() for cell in row.cells)
            if row_text.strip():
                paragraphs.append(row_text)

    text = "\n".join(paragraphs)

    # 提取元資料
    metadata = {}
    core_props = doc.core_properties
    if core_props.title:
        metadata["title"] = core_props.title
    if core_props.author:
        metadata["author"] = core_props.author

    return DocumentContent(text=text, format="docx", ...)

def _extract_xlsx(file_path: str) -> DocumentContent:
    wb = load_workbook(file_path, data_only=True, read_only=True)
    result = []

    for sheet_name in wb.sheetnames:
        sheet = wb[sheet_name]
        result.append(f"=== 工作表: {sheet_name} ===")

        for row in sheet.iter_rows(values_only=True):
            if any(cell is not None for cell in row):
                row_text = " | ".join(
                    str(cell) if cell is not None else ""
                    for cell in row
                )
                result.append(row_text)

    wb.close()
    return DocumentContent(text="\n".join(result), format="xlsx", ...)

def _extract_pptx(file_path: str) -> DocumentContent:
    prs = Presentation(file_path)
    result = []

    for i, slide in enumerate(prs.slides, 1):
        result.append(f"=== 投影片 {i} ===")

        for shape in slide.shapes:
            if hasattr(shape, "text") and shape.text.strip():
                result.append(shape.text)

    return DocumentContent(text="\n".join(result), format="pptx", ...)

def _extract_pdf(file_path: str) -> DocumentContent:
    with fitz.open(file_path) as doc:
        if doc.needs_pass:
            raise PasswordProtectedError("此文件有密碼保護，無法讀取")

        result = []
        has_text = False

        for page in doc:
            text = page.get_text()
            if text.strip():
                result.append(text)
                has_text = True

        # 純圖片 PDF 的特殊處理
        if not has_text:
            return DocumentContent(
                text="此 PDF 為掃描圖片，沒有可提取的文字。",
                format="pdf",
                metadata={"is_scanned": True},
                error="純圖片 PDF，無文字層"
            )

        return DocumentContent(text="\n".join(result), format="pdf", ...)

class DocumentReadError(ServiceError):
    """文件讀取錯誤（基礎類別）"""

class PasswordProtectedError(DocumentReadError):
    """文件有密碼保護"""

class CorruptedFileError(DocumentReadError):
    """文件損壞"""

class UnsupportedFormatError(DocumentReadError):
    """不支援的格式"""

class FileTooLargeError(DocumentReadError):
    """檔案過大"""

try:
    doc = Document(file_path)
except Exception as e:
    error_msg = str(e).lower()
    if "password" in error_msg or "encrypted" in error_msg:
        raise PasswordProtectedError("此文件有密碼保護，無法讀取")
    raise CorruptedFileError(f"無法解析 Word 文件: {e}")

def convert_pdf_to_images(
    file_path: str,
    output_dir: str,
    pages: str = "all",
    dpi: int = 150,
    output_format: str = "png",
    max_pages: int = 20
) -> PdfConversionResult:
    with fitz.open(file_path) as doc:
        if doc.needs_pass:
            raise PasswordProtectedError("此 PDF 有密碼保護")

        total_pages = len(doc)
        page_indices = _parse_pages_param(pages, total_pages)

        # pages="0" 時只回傳頁數資訊，不實際轉換
        if not page_indices:
            return PdfConversionResult(
                success=True,
                total_pages=total_pages,
                converted_pages=0,
                images=[],
                message=f"此 PDF 共有 {total_pages} 頁"
            )

        # 執行轉換
        zoom = dpi / 72
        mat = fitz.Matrix(zoom, zoom)
        images = []

        for idx in page_indices:
            page = doc[idx]
            pix = page.get_pixmap(matrix=mat)
            img_path = output_path / f"page-{idx + 1}.{output_format}"
            pix.save(str(img_path))
            images.append(str(img_path))

        return PdfConversionResult(
            success=True,
            total_pages=total_pages,
            converted_pages=len(images),
            images=images,
            message=f"已將全部 {total_pages} 頁轉換為圖片"
        )

# 解析文件
result = document_reader.extract_text(tmp_path)
text_content = result.text

# 寫入純文字暫存檔
with open(temp_path, "w", encoding="utf-8") as f:
    f.write(text_content)

# 同時保留 PDF 原始檔和文字版
# 回傳格式："PDF:/tmp/bot-files/xxx.pdf|TXT:/tmp/bot-files/xxx.txt"
if is_pdf:
    with open(pdf_temp_path, "wb") as f:
        f.write(content)
    return f"PDF:{pdf_temp_path}|TXT:{temp_path}"

# linebot_ai.py 中的工具設定
nanobanana_tools = [
    "mcp__nanobanana__generate_image",
    "mcp__nanobanana__edit_image",
]

"""圖片生成 Fallback 機制

整合兩層圖片生成服務：
1. nanobanana MCP（內建 Gemini Pro -> Flash 自動 fallback）
2. Hugging Face FLUX（最後備用，30 秒超時）
"""

# nanobanana 輸出的路徑需要轉換為相對路徑
# /tmp/ching-tech-os-cli/nanobanana-output/xxx.jpg
#   -> nanobanana-output/xxx.jpg
if "nanobanana-output/" in file_path:
    relative_path = "nanobanana-output/" + file_path.split("nanobanana-output/")[-1]

def should_trigger_ai(
    message_content: str,
    is_group: bool,
    is_reply_to_bot: bool = False,
) -> bool:
    if not is_group:
        # 個人對話：所有訊息都觸發
        return True

    # 群組對話：回覆機器人訊息時觸發
    if is_reply_to_bot:
        return True

    # 群組對話：被 @mention 時觸發
    content_lower = message_content.lower()
    for name in settings.line_bot_trigger_names:
        if f"@{name.lower()}" in content_lower:
            return True

    return False

def create_text_message_with_mention(
    text: str,
    mention_user_id: str | None = None,
) -> TextMessage | TextMessageV2:
    if mention_user_id:
        # 使用 TextMessageV2 + mention
        return TextMessageV2(
            text=MENTION_PLACEHOLDER + text,  # "{user} " + 回覆文字
            substitution={
                MENTION_KEY: MentionSubstitutionObject(
                    mentionee=UserMentionTarget(userId=mention_user_id)
                )
            },
        )
    else:
        return TextMessage(text=text)