diff --git a/docs/plugins/actions/export-to-word.md b/docs/plugins/actions/export-to-word.md
index 7410440..db4f271 100644
--- a/docs/plugins/actions/export-to-word.md
+++ b/docs/plugins/actions/export-to-word.md
@@ -1,7 +1,7 @@
# Export to Word
Action
-v0.2.0
+v0.4.0
Export conversation to Word (.docx) with **syntax highlighting**, **native math equations**, **Mermaid diagrams**, **citations**, and **enhanced table formatting**.
@@ -34,11 +34,34 @@ You can configure the following settings via the **Valves** button in the plugin
| Valve | Description | Default |
| :--- | :--- | :--- |
| `TITLE_SOURCE` | Source for document title/filename. Options: `chat_title`, `ai_generated`, `markdown_title` | `chat_title` |
-| `MERMAID_JS_URL` | URL for the Mermaid.js library (for diagram rendering). | `https://cdn.jsdelivr.net/npm/mermaid@10.9.1/dist/mermaid.min.js` |
-| `MERMAID_PNG_SCALE` | Scale factor for Mermaid PNG generation (Resolution). Higher = clearer but larger file size. | `3.0` |
-| `MERMAID_DISPLAY_SCALE` | Scale factor for Mermaid visual size in Word. >1.0 to enlarge, <1.0 to shrink. | `1.5` |
-| `MERMAID_OPTIMIZE_LAYOUT` | Automatically convert LR (Left-Right) flowcharts to TD (Top-Down) for better fit. | `True` |
+| `MAX_EMBED_IMAGE_MB` | Maximum image size to embed into DOCX (MB). | `20` |
+| `UI_LANGUAGE` | User interface language. Options: `en` (English), `zh` (Chinese). | `en` |
+| `FONT_LATIN` | Font name for Latin characters. | `Times New Roman` |
+| `FONT_ASIAN` | Font name for Asian characters. | `SimSun` |
+| `FONT_CODE` | Font name for code blocks. | `Consolas` |
+| `TABLE_HEADER_COLOR` | Table header background color (Hex without #). | `F2F2F2` |
+| `TABLE_ZEBRA_COLOR` | Table alternating row background color (Hex without #). | `FBFBFB` |
+| `MERMAID_JS_URL` | URL for the Mermaid.js library. | `https://cdn.jsdelivr.net/npm/mermaid@11.12.2/dist/mermaid.min.js` |
+| `MERMAID_JSZIP_URL` | URL for the JSZip library (required for DOCX manipulation). | `https://cdnjs.cloudflare.com/ajax/libs/jszip/3.10.1/jszip.min.js` |
+| `MERMAID_PNG_SCALE` | Scale factor for Mermaid PNG generation (Resolution). | `3.0` |
+| `MERMAID_DISPLAY_SCALE` | Scale factor for Mermaid visual size in Word. | `1.0` |
+| `MERMAID_OPTIMIZE_LAYOUT` | Automatically convert LR (Left-Right) flowcharts to TD (Top-Down). | `False` |
+| `MERMAID_BACKGROUND` | Background color for Mermaid diagrams (e.g., `white`, `transparent`). | `transparent` |
| `MERMAID_CAPTIONS_ENABLE` | Enable/disable figure captions for Mermaid diagrams. | `True` |
+| `MERMAID_CAPTION_STYLE` | Paragraph style name for Mermaid captions. | `Caption` |
+| `MERMAID_CAPTION_PREFIX` | Caption prefix label (e.g., 'Figure'). Empty = auto-detect based on language. | `""` |
+| `MATH_ENABLE` | Enable LaTeX math block conversion. | `True` |
+| `MATH_INLINE_DOLLAR_ENABLE` | Enable inline `$ ... $` math conversion. | `True` |
+
+### User-Level Configuration (UserValves)
+
+Users can override the following settings in their personal settings:
+- `TITLE_SOURCE`
+- `UI_LANGUAGE`
+- `FONT_LATIN`, `FONT_ASIAN`, `FONT_CODE`
+- `TABLE_HEADER_COLOR`, `TABLE_ZEBRA_COLOR`
+- `MERMAID_...` (Selected Mermaid settings)
+- `MATH_...` (Math settings)
---
diff --git a/docs/plugins/actions/export-to-word.zh.md b/docs/plugins/actions/export-to-word.zh.md
index 75f6dce..ee574b1 100644
--- a/docs/plugins/actions/export-to-word.zh.md
+++ b/docs/plugins/actions/export-to-word.zh.md
@@ -1,7 +1,7 @@
# Export to Word（导出为 Word）
Action
-v0.2.0
+v0.4.0
将当前对话导出为完美格式的 Word 文档，支持**代码语法高亮**、**原生数学公式**、**Mermaid 图表**、**引用资料**以及**增强表格**渲染。
@@ -34,11 +34,34 @@ Export to Word 插件会把聊天消息从 Markdown 转成精致的 Word 文档
| Valve | 说明 | 默认值 |
| :--- | :--- | :--- |
| `TITLE_SOURCE` | 文档标题/文件名的来源。选项：`chat_title` (对话标题), `ai_generated` (AI 生成), `markdown_title` (Markdown 标题) | `chat_title` |
-| `MERMAID_JS_URL` | Mermaid.js 库的 URL（用于图表渲染）。 | `https://cdn.jsdelivr.net/npm/mermaid@10.9.1/dist/mermaid.min.js` |
-| `MERMAID_PNG_SCALE` | Mermaid PNG 生成缩放比例（分辨率）。越高越清晰但文件越大。 | `3.0` |
-| `MERMAID_DISPLAY_SCALE` | Mermaid 在 Word 中的显示比例（视觉大小）。>1.0 放大, <1.0 缩小。 | `1.5` |
-| `MERMAID_OPTIMIZE_LAYOUT` | 优化 Mermaid 布局: 自动将 LR (左右) 转换为 TD (上下) 以适应页面。 | `True` |
+| `MAX_EMBED_IMAGE_MB` | 嵌入图片的最大大小 (MB)。 | `20` |
+| `UI_LANGUAGE` | 界面语言。选项：`en` (英语), `zh` (中文)。 | `zh` |
+| `FONT_LATIN` | 英文字体名称。 | `Calibri` |
+| `FONT_ASIAN` | 中文字体名称。 | `SimSun` |
+| `FONT_CODE` | 代码字体名称。 | `Consolas` |
+| `TABLE_HEADER_COLOR` | 表头背景色（十六进制，不带#）。 | `F2F2F2` |
+| `TABLE_ZEBRA_COLOR` | 表格隔行背景色（十六进制，不带#）。 | `FBFBFB` |
+| `MERMAID_JS_URL` | Mermaid.js 库的 URL。 | `https://cdn.jsdelivr.net/npm/mermaid@11.12.2/dist/mermaid.min.js` |
+| `MERMAID_JSZIP_URL` | JSZip 库的 URL（用于 DOCX 操作）。 | `https://cdnjs.cloudflare.com/ajax/libs/jszip/3.10.1/jszip.min.js` |
+| `MERMAID_PNG_SCALE` | Mermaid PNG 生成缩放比例（分辨率）。 | `3.0` |
+| `MERMAID_DISPLAY_SCALE` | Mermaid 在 Word 中的显示比例（视觉大小）。 | `1.0` |
+| `MERMAID_OPTIMIZE_LAYOUT` | 优化 Mermaid 布局: 自动将 LR (左右) 转换为 TD (上下)。 | `False` |
+| `MERMAID_BACKGROUND` | Mermaid 图表背景色（如 `white`, `transparent`）。 | `transparent` |
| `MERMAID_CAPTIONS_ENABLE` | 启用/禁用 Mermaid 图表的图注。 | `True` |
+| `MERMAID_CAPTION_STYLE` | Mermaid 图注的段落样式名称。 | `Caption` |
+| `MERMAID_CAPTION_PREFIX` | 图注前缀（如 '图'）。留空则根据语言自动检测。 | `""` |
+| `MATH_ENABLE` | 启用 LaTeX 数学公式块转换。 | `True` |
+| `MATH_INLINE_DOLLAR_ENABLE` | 启用行内 `$ ... $` 数学公式转换。 | `True` |
+
+### 用户级配置 (UserValves)
+
+用户可以在个人设置中覆盖以下配置：
+- `TITLE_SOURCE`
+- `UI_LANGUAGE`
+- `FONT_LATIN`, `FONT_ASIAN`, `FONT_CODE`
+- `TABLE_HEADER_COLOR`, `TABLE_ZEBRA_COLOR`
+- `MERMAID_...` (部分 Mermaid 设置)
+- `MATH_...` (数学公式设置)
---
diff --git a/docs/plugins/actions/index.md b/docs/plugins/actions/index.md
index f2d5d0b..ac2060e 100644
--- a/docs/plugins/actions/index.md
+++ b/docs/plugins/actions/index.md
@@ -63,7 +63,7 @@ Actions are interactive plugins that:
Export the current conversation to a formatted Word doc with **syntax highlighting**, **native math equations**, **Mermaid diagrams**, **citations**, and **enhanced table formatting**.
- **Version:** 0.2.0
+ **Version:** 0.4.0
[:octicons-arrow-right-24: Documentation](export-to-word.md)
diff --git a/docs/plugins/actions/index.zh.md b/docs/plugins/actions/index.zh.md
index f3e43db..24dee87 100644
--- a/docs/plugins/actions/index.zh.md
+++ b/docs/plugins/actions/index.zh.md
@@ -63,7 +63,7 @@ Actions 是交互式插件，能够：
将当前对话导出为完美格式的 Word 文档，支持**代码语法高亮**、**原生数学公式**、**Mermaid 图表**、**引用资料**以及**增强表格**渲染。
- **版本：** 0.2.0
+ **版本：** 0.4.0
[:octicons-arrow-right-24: 查看文档](export-to-word.md)
diff --git a/plugins/actions/export_to_docx/README.md b/plugins/actions/export_to_docx/README.md
index 10c57a3..eaed66a 100644
--- a/plugins/actions/export_to_docx/README.md
+++ b/plugins/actions/export_to_docx/README.md
@@ -24,11 +24,24 @@ You can configure the following settings via the **Valves** button in the plugin
- `chat_title`: Use the conversation title (default).
- `ai_generated`: Use AI to generate a short title based on the content.
- `markdown_title`: Extract the first h1/h2 heading from the Markdown content.
-- **MERMAID_JS_URL**: URL for the Mermaid.js library (for diagram rendering).
+- **MAX_EMBED_IMAGE_MB**: Maximum image size to embed into DOCX (MB). Default: `20`.
+- **UI_LANGUAGE**: User interface language, supports `en` (English) and `zh` (Chinese). Default: `en`.
+- **FONT_LATIN**: Font name for Latin characters. Default: `Times New Roman`.
+- **FONT_ASIAN**: Font name for Asian characters. Default: `SimSun`.
+- **FONT_CODE**: Font name for code blocks. Default: `Consolas`.
+- **TABLE_HEADER_COLOR**: Table header background color (Hex without #). Default: `F2F2F2`.
+- **TABLE_ZEBRA_COLOR**: Table alternating row background color (Hex without #). Default: `FBFBFB`.
+- **MERMAID_JS_URL**: URL for the Mermaid.js library.
+- **MERMAID_JSZIP_URL**: URL for the JSZip library (required for DOCX manipulation).
- **MERMAID_PNG_SCALE**: Scale factor for Mermaid PNG generation (Resolution). Default: `3.0`.
-- **MERMAID_DISPLAY_SCALE**: Scale factor for Mermaid visual size in Word. Default: `1.5`.
-- **MERMAID_OPTIMIZE_LAYOUT**: Automatically convert LR (Left-Right) flowcharts to TD (Top-Down). Default: `True`.
-- **MERMAID_CAPTIONS_ENABLE**: Enable/disable figure captions for Mermaid diagrams.
+- **MERMAID_DISPLAY_SCALE**: Scale factor for Mermaid visual size in Word. Default: `1.0`.
+- **MERMAID_OPTIMIZE_LAYOUT**: Automatically convert LR (Left-Right) flowcharts to TD (Top-Down). Default: `False`.
+- **MERMAID_BACKGROUND**: Background color for Mermaid diagrams (e.g., `white`, `transparent`). Default: `transparent`.
+- **MERMAID_CAPTIONS_ENABLE**: Enable/disable figure captions for Mermaid diagrams. Default: `True`.
+- **MERMAID_CAPTION_STYLE**: Paragraph style name for Mermaid captions. Default: `Caption`.
+- **MERMAID_CAPTION_PREFIX**: Caption prefix label (e.g., 'Figure'). Empty = auto-detect based on language.
+- **MATH_ENABLE**: Enable LaTeX math block conversion (`\[...\]` and `$$...$$`). Default: `True`.
+- **MATH_INLINE_DOLLAR_ENABLE**: Enable inline `$ ... $` math conversion. Default: `True`.
## Supported Markdown Syntax
@@ -75,6 +88,20 @@ All dependencies are declared in the plugin docstring.
## Changelog
+### v0.4.0
+
+- **Multi-language Support**: Added UI language switching (English/Chinese) with localized messages.
+- **Font & Style Configuration**: Customizable fonts for Latin/Asian text and code, plus table colors.
+- **Mermaid Enhancements**:
+ - Hybrid client-side rendering (SVG+PNG) for better clarity and compatibility.
+ - Configurable background color, fixing issues in dark mode.
+ - Added error boundaries to prevent export failures on render errors.
+- **Performance**: Real-time progress updates for large document exports.
+- **Bug Fixes**:
+ - Fixed parsing errors in Markdown tables containing code blocks or links.
+ - Fixed parsing issues with underscores (`_`), asterisks (`*`), and tildes (`~`) used as long separators.
+ - Enhanced error handling for image embedding.
+
### v0.3.0
- **Mermaid Diagrams**: Native support for rendering Mermaid diagrams as images in Word.
diff --git a/plugins/actions/export_to_docx/README_CN.md b/plugins/actions/export_to_docx/README_CN.md
index b5026d5..07a63b3 100644
--- a/plugins/actions/export_to_docx/README_CN.md
+++ b/plugins/actions/export_to_docx/README_CN.md
@@ -24,11 +24,24 @@
- `chat_title`：使用对话标题（默认）。
- `ai_generated`：使用 AI 根据内容生成简短标题。
- `markdown_title`：从 Markdown 内容中提取第一个一级或二级标题。
-- **MERMAID_JS_URL**：Mermaid.js 库的 URL（用于图表渲染）。
+- **MAX_EMBED_IMAGE_MB**：嵌入图片的最大大小 (MB)。默认：`20`。
+- **UI_LANGUAGE**：界面语言，支持 `en` (英语) 和 `zh` (中文)。默认：`zh`。
+- **FONT_LATIN**：英文字体名称。默认：`Calibri`。
+- **FONT_ASIAN**：中文字体名称。默认：`SimSun`。
+- **FONT_CODE**：代码字体名称。默认：`Consolas`。
+- **TABLE_HEADER_COLOR**：表头背景色（十六进制，不带#）。默认：`F2F2F2`。
+- **TABLE_ZEBRA_COLOR**：表格隔行背景色（十六进制，不带#）。默认：`FBFBFB`。
+- **MERMAID_JS_URL**：Mermaid.js 库的 URL。
+- **MERMAID_JSZIP_URL**：JSZip 库的 URL（用于 DOCX 操作）。
- **MERMAID_PNG_SCALE**：Mermaid PNG 生成缩放比例（分辨率）。默认：`3.0`。
-- **MERMAID_DISPLAY_SCALE**：Mermaid 在 Word 中的显示比例（视觉大小）。默认：`1.5`。
-- **MERMAID_OPTIMIZE_LAYOUT**：自动将 LR（左右）流程图转换为 TD（上下）。默认：`True`。
-- **MERMAID_CAPTIONS_ENABLE**：启用/禁用 Mermaid 图表的图注。
+- **MERMAID_DISPLAY_SCALE**：Mermaid 在 Word 中的显示比例（视觉大小）。默认：`1.0`。
+- **MERMAID_OPTIMIZE_LAYOUT**：自动将 LR（左右）流程图转换为 TD（上下）。默认：`False`。
+- **MERMAID_BACKGROUND**：Mermaid 图表背景色（如 `white`, `transparent`）。默认：`transparent`。
+- **MERMAID_CAPTIONS_ENABLE**：启用/禁用 Mermaid 图表的图注。默认：`True`。
+- **MERMAID_CAPTION_STYLE**：Mermaid 图注的段落样式名称。默认：`Caption`。
+- **MERMAID_CAPTION_PREFIX**：图注前缀（如 '图'）。留空则根据语言自动检测。
+- **MATH_ENABLE**：启用 LaTeX 数学公式块转换（`\[...\]` 和 `$$...$$`）。默认：`True`。
+- **MATH_INLINE_DOLLAR_ENABLE**：启用行内 `$ ... $` 数学公式转换。默认：`True`。
## 支持的 Markdown 语法
@@ -75,6 +88,20 @@
## 更新日志
+### v0.4.0
+
+- **多语言支持**: 新增界面语言切换（中文/英文），提示信息更友好。
+- **字体与样式配置**: 支持自定义中英文字体、代码字体以及表格颜色。
+- **Mermaid 增强**:
+ - 客户端混合渲染（SVG+PNG），提高清晰度与兼容性。
+ - 支持背景色配置，修复深色模式下的显示问题。
+ - 增加错误边界，渲染失败时显示提示而非中断导出。
+- **性能优化**: 导出大型文档时提供实时进度反馈。
+- **Bug 修复**:
+ - 修复 Markdown 表格中包含代码块或链接时的解析错误。
+ - 修复下划线（`_`）、星号（`*`）、波浪号（`~`）作为长分隔符时的解析问题。
+ - 增强图片嵌入的错误处理。
+
### v0.3.0
- **Mermaid 图表**: 原生支持将 Mermaid 图表渲染为 Word 中的图片。
diff --git a/plugins/actions/export_to_docx/export_to_word.py b/plugins/actions/export_to_docx/export_to_word.py
index 409d179..53ad57e 100644
--- a/plugins/actions/export_to_docx/export_to_word.py
+++ b/plugins/actions/export_to_docx/export_to_word.py
@@ -1,47 +1,51 @@
"""
-title: Export to Word
+title: Export to Word (Enhanced)
author: Fu-Jie
author_url: https://github.com/Fu-Jie
funding_url: https://github.com/Fu-Jie/awesome-openwebui
-version: 0.3.0
+version: 0.4.0
icon_url: data:image/svg+xml;base64,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
-requirements: python-docx==1.1.2, latex2mathml, mathml2omml
-description: Export conversation to Word (.docx) with syntax highlighting, native math equations (LaTeX), Mermaid diagrams, citations, and enhanced table formatting.
+requirements: python-docx, Pygments, latex2mathml, mathml2omml
+description: Export current conversation from Markdown to Word (.docx) with Mermaid diagrams rendered client-side (Mermaid.js, SVG+PNG), LaTeX math, real hyperlinks, improved tables, syntax highlighting, and blockquote support.
+notes: Enhanced version based on rbb-dev's fork (https://github.com/rbb-dev/awesome-openwebui). Added i18n support, configurable fonts/colors, and parallel PNG rendering optimization.
"""
-import os
+from __future__ import annotations
+
import re
import base64
import datetime
+import time
import io
import asyncio
import logging
-from typing import (
- Optional,
- Callable,
- Awaitable,
- Any,
- List,
- Tuple,
- Union,
- Dict,
- Literal,
- cast,
-)
+import hashlib
+import struct
+import zlib
+import binascii
+from pathlib import Path
+from dataclasses import dataclass
+from typing import Optional, Callable, Awaitable, Any, List, Tuple, Dict, cast
+from urllib.parse import quote
from docx import Document
from docx.shared import Pt, Inches, RGBColor, Cm
from docx.enum.text import WD_ALIGN_PARAGRAPH, WD_LINE_SPACING
from docx.enum.table import WD_TABLE_ALIGNMENT
from docx.enum.style import WD_STYLE_TYPE
from docx.opc.constants import RELATIONSHIP_TYPE as RT
-from docx.oxml.ns import qn
-from docx.oxml import OxmlElement, parse_xml
+from docx.oxml import parse_xml
+from docx.oxml.ns import qn, nsmap
+from docx.oxml import OxmlElement
from open_webui.models.chats import Chats
from open_webui.models.users import Users
from open_webui.utils.chat import generate_chat_completion
from pydantic import BaseModel, Field
-from dataclasses import dataclass
+# Files are used to embed internal /api/v1/files//content images.
+try:
+ from open_webui.models.files import Files # type: ignore
+except Exception: # pragma: no cover - depends on host Open WebUI runtime
+ Files = None
# Pygments for syntax highlighting
try:
@@ -61,6 +65,31 @@ try:
except Exception:
LATEX_MATH_AVAILABLE = False
+
+logging.basicConfig(
+ level=logging.INFO,
+ format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
+)
+logger = logging.getLogger(__name__)
+
+_AUTO_URL_RE = re.compile(r"(?:https?://|www\.)[^\s<>()]+")
+_DATA_IMAGE_URL_RE = re.compile(
+ r"^data:(?Pimage/[a-z0-9.+-]+)\s*;\s*base64\s*,\s*(?P.*)$",
+ re.IGNORECASE | re.DOTALL,
+)
+_OWUI_API_FILE_ID_RE = re.compile(
+ r"/api/v1/files/(?P[A-Za-z0-9-]+)(?:/content)?(?:[/?#]|$)",
+ re.IGNORECASE,
+)
+_CURRENCY_NUMBER_RE = re.compile(r"^\d[\d,]*(?:\.\d+)?$")
+
+_TRANSPARENT_1PX_PNG = base64.b64decode(
+ "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVQImWNgYGBgAAAABQABDQottAAAAABJRU5ErkJggg=="
+)
+
+_ASVG_NS = "http://schemas.microsoft.com/office/drawing/2016/SVG/main"
+nsmap.setdefault("asvg", _ASVG_NS)
+
_REASONING_DETAILS_RE = re.compile(
r"]*\btype\s*=\s*(?:\"reasoning\"|'reasoning'|reasoning)[^>]*>.*?",
re.IGNORECASE | re.DOTALL,
@@ -71,13 +100,6 @@ _ANALYSIS_RE = re.compile(
)
-logging.basicConfig(
- level=logging.INFO,
- format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
-)
-logger = logging.getLogger(__name__)
-
-
@dataclass(frozen=True)
class _CitationRef:
idx: int
@@ -87,60 +109,203 @@ class _CitationRef:
source_id: str
-@dataclass
-class _MermaidFenceBlock:
- info_raw: str
- language: str
- attrs: List[str]
- source: str
-
-
class Action:
+ # Internationalization message dictionaries
+ _I18N_MESSAGES: Dict[str, Dict[str, str]] = {
+ "en": {
+ "converting": "Converting to Word document...",
+ "exported": "Word document exported",
+ "success": "Successfully exported to {filename}",
+ "error_no_content": "No content found to export!",
+ "error_export": "Error exporting Word document: {error}",
+ "export_failed": "Export failed: {error}",
+ "figure_prefix": "Figure",
+ "references": "References",
+ },
+ "zh": {
+ "converting": "正在转换为 Word 文档...",
+ "exported": "Word 文档导出完成",
+ "success": "成功导出至 {filename}",
+ "error_no_content": "没有找到可导出的内容！",
+ "error_export": "导出 Word 文档时出错: {error}",
+ "export_failed": "导出失败: {error}",
+ "figure_prefix": "图",
+ "references": "参考文献",
+ },
+ }
+
class Valves(BaseModel):
TITLE_SOURCE: str = Field(
default="chat_title",
description="Title Source: 'chat_title' (Chat Title), 'ai_generated' (AI Generated), 'markdown_title' (Markdown Title)",
)
+
+ MAX_EMBED_IMAGE_MB: int = Field(
+ default=20,
+ description="Maximum image size to embed into DOCX (MB). Applies to data URLs and /api/v1/files//content images.",
+ )
+
+ # Font configuration
+ FONT_LATIN: str = Field(
+ default="Times New Roman",
+ description="Font for Latin characters (e.g., 'Times New Roman', 'Calibri', 'Arial')",
+ )
+ FONT_ASIAN: str = Field(
+ default="SimSun",
+ description="Font for Asian characters (e.g., 'SimSun', 'Microsoft YaHei', 'PingFang SC')",
+ )
+ FONT_CODE: str = Field(
+ default="Consolas",
+ description="Font for code blocks and inline code (e.g., 'Consolas', 'Courier New', 'Monaco')",
+ )
+
+ # Table styling
+ TABLE_HEADER_COLOR: str = Field(
+ default="F2F2F2",
+ description="Table header background color (hex, without #)",
+ )
+ TABLE_ZEBRA_COLOR: str = Field(
+ default="FBFBFB",
+ description="Table zebra stripe background color for alternate rows (hex, without #)",
+ )
+
MERMAID_JS_URL: str = Field(
- default="https://cdn.jsdelivr.net/npm/mermaid@10.9.1/dist/mermaid.min.js",
+ default="https://cdn.jsdelivr.net/npm/mermaid@11.12.2/dist/mermaid.min.js",
description="Mermaid JS CDN URL",
)
MERMAID_JSZIP_URL: str = Field(
default="https://cdnjs.cloudflare.com/ajax/libs/jszip/3.10.1/jszip.min.js",
- description="JSZip CDN URL (for DOCX manipulation)",
- )
- MERMAID_OPTIMIZE_LAYOUT: bool = Field(
- default=True,
- description="Optimize Mermaid Layout: Automatically convert LR (Left-Right) to TD (Top-Down) for better fit.",
+ description="JSZip CDN URL (DOCX manipulation)",
)
MERMAID_PNG_SCALE: float = Field(
default=3.0,
- description="Mermaid PNG Scale (Resolution): Higher = clearer but larger file size. Default: 3.0",
+ description="PNG render resolution multiplier (higher = clearer, larger file)",
)
MERMAID_DISPLAY_SCALE: float = Field(
- default=1.5,
- description="Mermaid Display Scale (Visual Size): >1.0 to enlarge, <1.0 to shrink. Default: 1.5",
+ default=1.0,
+ description="Diagram width relative to available page width (<=1 recommended)",
)
+ MERMAID_OPTIMIZE_LAYOUT: bool = Field(
+ default=False,
+ description="Optimize Mermaid layout: convert LR to TD for graph/flowchart",
+ )
+ MERMAID_BACKGROUND: str = Field(
+ default="",
+ description="Mermaid background color. Empty = transparent (recommended for Word dark mode). Used only for optional PNG fill.",
+ )
+
MERMAID_CAPTIONS_ENABLE: bool = Field(
default=True,
- description="Enable Mermaid Captions",
+ description="Add figure captions under Mermaid images/charts",
)
MERMAID_CAPTION_STYLE: str = Field(
default="Caption",
- description="Mermaid Caption Style Name",
+ description="Paragraph style name for Mermaid captions (uses 'Caption' if available, otherwise creates a safe custom style)",
)
MERMAID_CAPTION_PREFIX: str = Field(
- default="Figure",
- description="Mermaid Caption Prefix",
+ default="",
+ description="Caption prefix label (e.g., 'Figure' or '图'). Empty = auto-detect based on user language.",
+ )
+
+ MATH_ENABLE: bool = Field(
+ default=True,
+ description="Enable LaTeX math block conversion (\\[...\\] and $$...$$) into Word equations",
+ )
+ MATH_INLINE_DOLLAR_ENABLE: bool = Field(
+ default=True,
+ description="Enable inline $...$ math conversion into Word equations (conservative parsing to reduce false positives)",
+ )
+
+ # Language configuration
+ UI_LANGUAGE: str = Field(
+ default="en",
+ description="UI language for export messages. Options: 'en' (English), 'zh' (Chinese)",
+ )
+
+ class UserValves(BaseModel):
+ TITLE_SOURCE: str = Field(
+ default="chat_title",
+ description="Title Source: 'chat_title' (Chat Title), 'ai_generated' (AI Generated), 'markdown_title' (Markdown Title)",
+ )
+ UI_LANGUAGE: str = Field(
+ default="en",
+ description="UI language for export messages. Options: 'en' (English), 'zh' (Chinese)",
+ )
+ FONT_LATIN: str = Field(
+ default="Times New Roman",
+ description="Font for Latin characters (e.g., 'Times New Roman', 'Calibri', 'Arial')",
+ )
+ FONT_ASIAN: str = Field(
+ default="SimSun",
+ description="Font for Asian characters (e.g., 'SimSun', 'Microsoft YaHei', 'PingFang SC')",
+ )
+ FONT_CODE: str = Field(
+ default="Consolas",
+ description="Font for code blocks and inline code (e.g., 'Consolas', 'Courier New', 'Monaco')",
+ )
+ TABLE_HEADER_COLOR: str = Field(
+ default="F2F2F2",
+ description="Table header background color (hex, without #)",
+ )
+ TABLE_ZEBRA_COLOR: str = Field(
+ default="FBFBFB",
+ description="Table zebra stripe background color for alternate rows (hex, without #)",
+ )
+ MERMAID_PNG_SCALE: float = Field(
+ default=3.0,
+ description="PNG render resolution multiplier (higher = clearer, larger file)",
+ )
+ MERMAID_DISPLAY_SCALE: float = Field(
+ default=1.0,
+ description="Diagram width relative to available page width (<=1 recommended)",
+ )
+ MERMAID_OPTIMIZE_LAYOUT: bool = Field(
+ default=False,
+ description="Optimize Mermaid layout: convert LR to TD for graph/flowchart",
+ )
+ MERMAID_BACKGROUND: str = Field(
+ default="",
+ description="Mermaid background color. Empty = transparent (recommended for Word dark mode). Used only for optional PNG fill.",
+ )
+ MERMAID_CAPTIONS_ENABLE: bool = Field(
+ default=True,
+ description="Add figure captions under Mermaid images/charts",
+ )
+ MATH_ENABLE: bool = Field(
+ default=True,
+ description="Enable LaTeX math block conversion (\\\\[...\\\\] and $$...$$) into Word equations",
+ )
+ MATH_INLINE_DOLLAR_ENABLE: bool = Field(
+ default=True,
+ description="Enable inline $...$ math conversion into Word equations (conservative parsing to reduce false positives)",
)
def __init__(self):
self.valves = self.Valves()
- self._mermaid_figure_counter = 0
- self._caption_style_name = ""
+ self._mermaid_figure_counter: int = 0
+ self._mermaid_placeholder_counter: int = 0
+ self._caption_style_name: Optional[str] = None
self._citation_anchor_by_index: Dict[int, str] = {}
self._citation_refs: List[_CitationRef] = []
self._bookmark_id_counter: int = 1
+ self._active_doc: Optional[Document] = None
+ self._user_lang: str = "en" # Will be set per-request
+
+ def _get_lang_key(self, user_language: str) -> str:
+ """Convert user language code to i18n key (e.g., 'zh-CN' -> 'zh', 'en-US' -> 'en')."""
+ lang = (user_language or "en").lower().split("-")[0]
+ return lang if lang in self._I18N_MESSAGES else "en"
+
+ def _get_msg(self, key: str, **kwargs) -> str:
+ """Get internationalized message by key with optional formatting."""
+ messages = self._I18N_MESSAGES.get(self._user_lang, self._I18N_MESSAGES["en"])
+ msg = messages.get(key, self._I18N_MESSAGES["en"].get(key, key))
+ if kwargs:
+ try:
+ return msg.format(**kwargs)
+ except KeyError:
+ return msg
+ return msg
async def _send_notification(self, emitter: Callable, type: str, content: str):
await emitter(
@@ -158,15 +323,10 @@ class Action:
):
logger.info(f"action:{__name__}")
- # Reset counters for new request
- self._mermaid_figure_counter = 0
- self._bookmark_id_counter = 1
-
# Parse user info
+ user_name = "User"
+ user_id = "unknown_user"
if isinstance(__user__, (list, tuple)):
- user_language = (
- __user__[0].get("language", "en-US") if __user__ else "en-US"
- )
user_name = __user__[0].get("name", "User") if __user__[0] else "User"
user_id = (
__user__[0]["id"]
@@ -174,10 +334,21 @@ class Action:
else "unknown_user"
)
elif isinstance(__user__, dict):
- user_language = __user__.get("language", "en-US")
user_name = __user__.get("name", "User")
user_id = __user__.get("id", "unknown_user")
+ # Apply UserValves if present
+ if __user__ and "valves" in __user__:
+ # Update self.valves with user-specific values
+ # Note: This assumes per-request instantiation or that we are okay with modifying the singleton.
+ # Given the plugin architecture, we'll update it for this execution.
+ for key, value in __user__["valves"].model_dump().items():
+ if hasattr(self.valves, key):
+ setattr(self.valves, key, value)
+
+ # Get user language from Valves configuration
+ self._user_lang = self._get_lang_key(self.valves.UI_LANGUAGE)
+
if __event_emitter__:
last_assistant_message = body["messages"][-1]
@@ -185,7 +356,7 @@ class Action:
{
"type": "status",
"data": {
- "description": "Converting to Word document...",
+ "description": self._get_msg("converting"),
"done": False,
},
}
@@ -193,10 +364,12 @@ class Action:
try:
message_content = last_assistant_message["content"]
+ if isinstance(message_content, str):
+ message_content = self._strip_reasoning_blocks(message_content)
if not message_content or not message_content.strip():
await self._send_notification(
- __event_emitter__, "error", "No content found to export!"
+ __event_emitter__, "error", self._get_msg("error_no_content")
)
return
@@ -210,22 +383,22 @@ class Action:
chat_title = await self.fetch_chat_title(chat_id, user_id)
if (
- self.valves.TITLE_SOURCE == "chat_title"
- or not self.valves.TITLE_SOURCE
+ self.valves.TITLE_SOURCE.strip() == "chat_title"
+ or not self.valves.TITLE_SOURCE.strip()
):
title = chat_title
- elif self.valves.TITLE_SOURCE == "markdown_title":
+ elif self.valves.TITLE_SOURCE.strip() == "markdown_title":
title = self.extract_title(message_content)
- elif self.valves.TITLE_SOURCE == "ai_generated":
+ elif self.valves.TITLE_SOURCE.strip() == "ai_generated":
title = await self.generate_title_using_ai(
body, message_content, user_id, __request__
)
# Fallback logic
if not title:
- if self.valves.TITLE_SOURCE != "chat_title" and chat_title:
+ if self.valves.TITLE_SOURCE.strip() != "chat_title" and chat_title:
title = chat_title
- elif self.valves.TITLE_SOURCE != "markdown_title":
+ elif self.valves.TITLE_SOURCE.strip() != "markdown_title":
extracted = self.extract_title(message_content)
if extracted:
title = extracted
@@ -233,10 +406,15 @@ class Action:
current_datetime = datetime.datetime.now()
formatted_date = current_datetime.strftime("%Y%m%d")
- if title:
- filename = f"{self.clean_filename(title)}.docx"
+ cleaned_title = self.clean_filename(title) if title else ""
+ if cleaned_title:
+ filename = f"{cleaned_title}.docx"
else:
- filename = f"{user_name}_{formatted_date}.docx"
+ clean_user = self.clean_filename(user_name)
+ filename = f"{clean_user}_{formatted_date}.docx"
+
+ # Escape filename for JS string
+ js_filename = filename.replace("\\", "\\\\").replace('"', '\\"')
top_heading = ""
if chat_title:
@@ -246,17 +424,15 @@ class Action:
# Create Word document; if no h1 exists, inject chat title as h1
has_h1 = bool(re.search(r"^#\s+.+$", message_content, re.MULTILINE))
-
- # Extract sources if available (for citations)
sources = (
last_assistant_message.get("sources") or body.get("sources") or []
)
-
- doc = self.markdown_to_docx(
+ doc = await self.markdown_to_docx(
message_content,
top_heading=top_heading,
has_h1=has_h1,
sources=sources,
+ event_emitter=__event_emitter__,
)
# Save to memory
@@ -266,13 +442,6 @@ class Action:
file_content = doc_buffer.read()
base64_blob = base64.b64encode(file_content).decode("utf-8")
- # Escape message_content for JavaScript template literal
- escaped_content = (
- message_content.replace("\\", "\\\\") # Escape backslashes first
- .replace("`", "\\`") # Escape backticks
- .replace("${", "\\${") # Escape template literal expressions
- )
-
# Trigger file download
if __event_call__:
await __event_call__(
@@ -281,183 +450,19 @@ class Action:
"data": {
"code": f"""
(async function() {{
- try {{
- // Parse document.xml to find placeholders and extract optimized code
- // We do this FIRST to get the actual code to render (which might have been optimized in Python)
-
- // Load JSZip
- if (!window.JSZip) {{
- await new Promise((resolve, reject) => {{
- const script = document.createElement("script");
- script.src = "{self.valves.MERMAID_JSZIP_URL}";
- script.onload = resolve;
- script.onerror = reject;
- document.head.appendChild(script);
- }});
- }}
+ const base64Data = "{base64_blob}";
+ const filename = "{js_filename}";
+ const mermaidUrl = "{self.valves.MERMAID_JS_URL}";
+ const jszipUrl = "{self.valves.MERMAID_JSZIP_URL}";
+ const pngScale = {float(self.valves.MERMAID_PNG_SCALE)};
+ const displayScale = {float(self.valves.MERMAID_DISPLAY_SCALE)};
+ const bgRaw = "{(self.valves.MERMAID_BACKGROUND or '').strip()}";
+ const bg = (bgRaw || "").trim();
+ const bgFill = (bg && bg.toLowerCase() !== "transparent") ? bg : "";
+ const themeBackground = bgFill || "transparent";
- const base64Data = "{base64_blob}";
- const binaryData = atob(base64Data);
- const arrayBuffer = new Uint8Array(binaryData.length);
- for (let i = 0; i < binaryData.length; i++) {{
- arrayBuffer[i] = binaryData.charCodeAt(i);
- }}
-
- const zip = new JSZip();
- await zip.loadAsync(arrayBuffer);
-
- // Parse document.xml
- const docXml = await zip.file("word/document.xml").async("string");
- const parser = new DOMParser();
- const xmlDoc = parser.parseFromString(docXml, "application/xml");
-
- const drawings = xmlDoc.getElementsByTagName("w:drawing");
- const placeholderInfo = [];
-
- for (let i = 0; i < drawings.length; i++) {{
- const drawing = drawings[i];
- const docPr = drawing.getElementsByTagName("wp:docPr")[0];
- if (docPr) {{
- const descr = docPr.getAttribute("descr");
- if (descr && descr.startsWith("MERMAID_SRC:")) {{
- const encodedCode = descr.substring("MERMAID_SRC:".length);
- const code = decodeURIComponent(encodedCode);
-
- // Find the blip and extent to replace
- const parent = drawing.parentNode; // w:r usually, or w:drawing parent
- // We need to find a:blip and wp:extent within this drawing
- const blip = drawing.getElementsByTagName("a:blip")[0];
- const extent = drawing.getElementsByTagName("wp:extent")[0];
-
- if (blip && extent) {{
- const rId = blip.getAttribute("r:embed");
- placeholderInfo.push({{ rId, extent, code }});
- }}
- }}
- }}
- }}
-
- if (placeholderInfo.length === 0) {{
- console.log("No Mermaid placeholders found in DOCX.");
- // Just download the file as is
- const blob = new Blob([arrayBuffer], {{type: "application/vnd.openxmlformats-officedocument.wordprocessingml.document"}});
- const url = URL.createObjectURL(blob);
- const a = document.createElement("a");
- a.style.display = "none";
- a.href = url;
- a.download = "{filename}";
- document.body.appendChild(a);
- a.click();
- URL.revokeObjectURL(url);
- document.body.removeChild(a);
- return;
- }}
-
- console.log(`Found ${{placeholderInfo.length}} Mermaid placeholders.`);
-
- // Load Mermaid
- if (!window.mermaid) {{
- await new Promise((resolve, reject) => {{
- const script = document.createElement("script");
- script.src = "{self.valves.MERMAID_JS_URL}";
- script.onload = resolve;
- script.onerror = reject;
- document.head.appendChild(script);
- }});
- }}
-
- mermaid.initialize({{
- startOnLoad: false,
- theme: 'default',
- }});
-
- // Read rels XML once
- const relsXml = await zip.file("word/_rels/document.xml.rels").async("string");
- const relsDoc = parser.parseFromString(relsXml, "application/xml");
- const relationships = relsDoc.getElementsByTagName("Relationship");
- const rIdToPath = {{}};
-
- for (let i = 0; i < relationships.length; i++) {{
- const rel = relationships[i];
- rIdToPath[rel.getAttribute("Id")] = rel.getAttribute("Target");
- }}
-
- // Render and replace
- console.log(`Processing ${{placeholderInfo.length}} diagrams...`);
-
- for (let i = 0; i < placeholderInfo.length; i++) {{
- const {{ rId, extent, code }} = placeholderInfo[i];
- const imagePath = "word/" + rIdToPath[rId];
-
- console.log(`Block ${{i + 1}}/${{placeholderInfo.length}}: Rendering and replacing at ${{imagePath}}`);
-
- // Render SVG
- const id = "mermaid-export-" + i;
- const {{ svg }} = await mermaid.render(id, code);
-
- // Convert SVG to PNG
- const canvas = document.createElement("canvas");
- const ctx = canvas.getContext("2d");
- const img = new Image();
-
- // Get SVG dimensions
- const svgMatch = svg.match(/viewBox="[^"]*\s+[^"]*\s+([^"\s]+)\s+([^"\s]+)"/);
- let width = 800;
- let height = 600;
- if (svgMatch) {{
- width = parseFloat(svgMatch[1]);
- height = parseFloat(svgMatch[2]);
- }}
-
- // Scale up for better quality
- const scale = {self.valves.MERMAID_PNG_SCALE};
- canvas.width = width * scale;
- canvas.height = height * scale;
-
- await new Promise((resolve, reject) => {{
- img.onload = resolve;
- img.onerror = reject;
- img.src = "data:image/svg+xml;base64," + btoa(unescape(encodeURIComponent(svg)));
- }});
-
- ctx.scale(scale, scale);
- ctx.drawImage(img, 0, 0, width, height);
-
- const pngDataUrl = canvas.toDataURL("image/png");
- const pngBase64 = pngDataUrl.split(",")[1];
-
- // Replace image in ZIP
- zip.file(imagePath, pngBase64, {{base64: true}});
-
- // Update dimensions in document.xml (EMUs)
- // 1 inch = 914400 EMUs, 1 pixel ≈ 9525 EMUs at 96 DPI
- // Max width: ~6 inches (page width minus margins)
- const maxWidthEmu = 5486400; // 6 inches
- const displayScale = {self.valves.MERMAID_DISPLAY_SCALE};
- let emuWidth = Math.round(width * 9525 * displayScale);
- let emuHeight = Math.round(height * 9525 * displayScale);
-
- // Scale down if too wide
- if (emuWidth > maxWidthEmu) {{
- const scaleFactor = maxWidthEmu / emuWidth;
- emuWidth = maxWidthEmu;
- emuHeight = Math.round(emuHeight * scaleFactor);
- }}
-
- extent.setAttribute("cx", emuWidth);
- extent.setAttribute("cy", emuHeight);
- }}
-
- // Serialize updated XML
- const serializer = new XMLSerializer();
- const newDocXml = serializer.serializeToString(xmlDoc);
- zip.file("word/document.xml", newDocXml);
-
- // Generate final blob
- const finalBlob = await zip.generateAsync({{type: "blob"}});
- const filename = "{filename}";
-
- const url = URL.createObjectURL(finalBlob);
+ function downloadBlob(blob, filename) {{
+ const url = URL.createObjectURL(blob);
const a = document.createElement("a");
a.style.display = "none";
a.href = url;
@@ -466,9 +471,362 @@ class Action:
a.click();
URL.revokeObjectURL(url);
document.body.removeChild(a);
+ }}
+
+ async function loadScript(url, globalName) {{
+ if (globalName && window[globalName]) return;
+ await new Promise((resolve, reject) => {{
+ const script = document.createElement("script");
+ script.src = url;
+ script.onload = resolve;
+ script.onerror = reject;
+ document.head.appendChild(script);
+ }});
+ }}
+
+ function decodeBase64ToUint8Array(b64) {{
+ const binary = atob(b64);
+ const bytes = new Uint8Array(binary.length);
+ for (let i = 0; i < binary.length; i++) bytes[i] = binary.charCodeAt(i);
+ return bytes;
+ }}
+
+ function parseViewBox(vb) {{
+ if (!vb) return null;
+ const parts = vb.trim().split(/\\s+/).map(Number);
+ if (parts.length !== 4 || parts.some((n) => !isFinite(n))) return null;
+ return {{ minX: parts[0], minY: parts[1], width: parts[2], height: parts[3] }};
+ }}
+
+ function normalizeSvgForWord(svgText) {{
+ const parser = new DOMParser();
+ const doc = parser.parseFromString(svgText, "image/svg+xml");
+ const svgEl = doc.documentElement;
+ if (!svgEl || svgEl.tagName.toLowerCase() !== "svg") return svgText;
+
+ // Pad viewBox a little to reduce clipping in Word.
+ const vb0 = parseViewBox(svgEl.getAttribute("viewBox"));
+ if (vb0 && vb0.width > 0 && vb0.height > 0) {{
+ const minDim = Math.min(vb0.width, vb0.height);
+ let pad = Math.max(8.0, minDim * 0.02);
+ pad = Math.min(pad, 24.0);
+ const vb = {{
+ minX: vb0.minX - pad,
+ minY: vb0.minY - pad,
+ width: vb0.width + 2 * pad,
+ height: vb0.height + 2 * pad,
+ }};
+ svgEl.setAttribute("viewBox", `${{vb.minX}} ${{vb.minY}} ${{vb.width}} ${{vb.height}}`);
+ }}
+
+ const vb = parseViewBox(svgEl.getAttribute("viewBox"));
+ const widthAttr = (svgEl.getAttribute("width") || "").trim();
+ const heightAttr = (svgEl.getAttribute("height") || "").trim();
+ const widthPct = widthAttr.endsWith("%");
+ const heightPct = heightAttr.endsWith("%");
+ if (vb && vb.width > 0 && vb.height > 0 && (!widthAttr || !heightAttr || widthPct || heightPct)) {{
+ svgEl.setAttribute("width", `${{vb.width}}`);
+ svgEl.setAttribute("height", `${{vb.height}}`);
+ }}
+
+ svgEl.removeAttribute("style");
+ svgEl.setAttribute("preserveAspectRatio", "xMidYMid meet");
+ svgEl.setAttribute("overflow", "visible");
+
+ const removeNode = (n) => {{
+ try {{ n && n.parentNode && n.parentNode.removeChild(n); }} catch (_e) {{}}
+ }};
+
+ // Remove Mermaid/OWUI background rectangles to avoid \"white box\" rendering in Word dark mode.
+ svgEl
+ .querySelectorAll('rect[data-owui-bg=\"1\"], rect.background, rect[class~=\"background\"], rect#background')
+ .forEach(removeNode);
+ try {{
+ const isWhiteish = (fill) => {{
+ const f = (fill || "").trim().toLowerCase();
+ return (
+ f === "white" ||
+ f === "#fff" ||
+ f === "#ffffff" ||
+ f === "rgb(255,255,255)" ||
+ f === "rgb(255, 255, 255)"
+ );
+ }};
+ const nearly = (a, b) => Math.abs(a - b) <= 1e-3;
+ const rectMatches = (r, box) => {{
+ if (!box) return false;
+ const x = parseFloat(r.getAttribute("x") || "0");
+ const y = parseFloat(r.getAttribute("y") || "0");
+ const w = parseFloat(r.getAttribute("width") || "");
+ const h = parseFloat(r.getAttribute("height") || "");
+ if (!isFinite(x) || !isFinite(y) || !isFinite(w) || !isFinite(h)) return false;
+ return (
+ nearly(x, box.minX) &&
+ nearly(y, box.minY) &&
+ nearly(w, box.width) &&
+ nearly(h, box.height)
+ );
+ }};
+ const vbNow = parseViewBox(svgEl.getAttribute("viewBox"));
+ svgEl.querySelectorAll("rect[fill]").forEach((r) => {{
+ const fill = r.getAttribute("fill");
+ if (!isWhiteish(fill)) return;
+ if (rectMatches(r, vb0) || rectMatches(r, vbNow)) removeNode(r);
+ }});
+ }} catch (_e) {{}}
+ try {{
+ const vbCanvas = parseViewBox(svgEl.getAttribute(\"viewBox\")) || vb0 || vb;
+ if (vbCanvas) {{
+ const existing = svgEl.querySelector('rect[data-owui-canvas=\"1\"]');
+ const rect = existing || doc.createElementNS(\"http://www.w3.org/2000/svg\", \"rect\");
+ rect.setAttribute(\"data-owui-canvas\", \"1\");
+ rect.setAttribute(\"x\", `${{vbCanvas.minX}}`);
+ rect.setAttribute(\"y\", `${{vbCanvas.minY}}`);
+ rect.setAttribute(\"width\", `${{vbCanvas.width}}`);
+ rect.setAttribute(\"height\", `${{vbCanvas.height}}`);
+ rect.setAttribute(\"fill\", \"#FFFFFF\");
+ // Word quirk: without a full-canvas rect with *non-zero* opacity, Word will often
+ // only offer \"Convert to Shape\" when clicking on an actual stroke/fill (not empty space).
+ // We keep this rect nearly transparent and non-interactive.
+ rect.setAttribute(\"fill-opacity\", \"0.001\");
+ rect.setAttribute(\"stroke\", \"none\");
+ rect.setAttribute(\"stroke-opacity\", \"0\");
+ rect.setAttribute(\"pointer-events\", \"none\");
+ if (!existing) {{
+ const first = svgEl.firstChild;
+ svgEl.insertBefore(rect, first);
+ }}
+ }}
+ }} catch (_e) {{}}
+
+ return new XMLSerializer().serializeToString(svgEl);
+ }}
+
+ function getMaxWidthEmu(xmlDoc) {{
+ try {{
+ const sects = xmlDoc.getElementsByTagName("w:sectPr");
+ const sect = sects && sects.length ? sects[sects.length - 1] : null;
+ if (!sect) return 5486400; // 6 in
+ const pgSz = sect.getElementsByTagName("w:pgSz")[0];
+ const pgMar = sect.getElementsByTagName("w:pgMar")[0];
+ if (!pgSz || !pgMar) return 5486400;
+ const pageW = parseInt(pgSz.getAttribute("w:w") || "", 10);
+ const left = parseInt(pgMar.getAttribute("w:left") || "", 10);
+ const right = parseInt(pgMar.getAttribute("w:right") || "", 10);
+ if (!isFinite(pageW) || !isFinite(left) || !isFinite(right)) return 5486400;
+ const twips = Math.max(1, pageW - left - right);
+ return Math.round(twips * 635); // 1 twip = 635 EMU
+ }} catch (_e) {{
+ return 5486400;
+ }}
+ }}
+
+ function getChildByTag(parent, tag) {{
+ const nodes = parent.getElementsByTagName(tag);
+ return nodes && nodes.length ? nodes[0] : null;
+ }}
+
+ try {{
+ await loadScript(jszipUrl, "JSZip");
+ await loadScript(mermaidUrl, "mermaid");
+
+ // Mermaid init: disable htmlLabels to keep SVG Word-friendly; PNG fallback still included.
+ try {{
+ window.mermaid.initialize({{
+ startOnLoad: false,
+ theme: "default",
+ themeVariables: {{
+ background: themeBackground,
+ fontFamily: "Calibri, Segoe UI, Arial, sans-serif",
+ fontSize: "10pt",
+ }},
+ themeCSS: ".slice {{ font-size: 10pt !important; }}\\n.legend text {{ font-size: 10pt !important; }}\\n.pieTitleText {{ font-size: 10pt !important; }}",
+ fontFamily: "Calibri, Segoe UI, Arial, sans-serif",
+ securityLevel: "strict",
+ flowchart: {{ htmlLabels: false }},
+ }});
+ }} catch (_e) {{
+ // Ignore and proceed with defaults.
+ }}
+
+ const bytes = decodeBase64ToUint8Array(base64Data);
+ const zip = new window.JSZip();
+ await zip.loadAsync(bytes);
+
+ const docXml = await zip.file("word/document.xml").async("string");
+ const relsXml = await zip.file("word/_rels/document.xml.rels").async("string");
+ const parser = new DOMParser();
+ const xmlDoc = parser.parseFromString(docXml, "application/xml");
+ const relsDoc = parser.parseFromString(relsXml, "application/xml");
+
+ // Build rId -> target path mapping
+ const rels = relsDoc.getElementsByTagName("Relationship");
+ const rIdToTarget = {{}};
+ for (let i = 0; i < rels.length; i++) {{
+ const rel = rels[i];
+ const id = rel.getAttribute("Id");
+ const target = rel.getAttribute("Target");
+ if (id && target) rIdToTarget[id] = target;
+ }}
+
+ const maxWidthEmu = getMaxWidthEmu(xmlDoc);
+ const maxWidthEmuScaled = Math.max(1, Math.round(maxWidthEmu * Math.min(1.0, Math.max(0.1, displayScale || 1.0))));
+
+ const drawings = xmlDoc.getElementsByTagName("w:drawing");
+ const placeholders = [];
+
+ for (let i = 0; i < drawings.length; i++) {{
+ const drawing = drawings[i];
+ const docPr = getChildByTag(drawing, "wp:docPr");
+ if (!docPr) continue;
+ const descr = docPr.getAttribute("descr") || "";
+ if (!descr.startsWith("MERMAID_SRC:")) continue;
+ const encoded = descr.substring("MERMAID_SRC:".length);
+ const code = decodeURIComponent(encoded);
+
+ const blip = getChildByTag(drawing, "a:blip");
+ const ridPng = blip ? blip.getAttribute("r:embed") : null;
+ const svgBlip = getChildByTag(drawing, "asvg:svgBlip");
+ const ridSvg = svgBlip ? svgBlip.getAttribute("r:embed") : null;
+
+ const container = getChildByTag(drawing, "wp:inline") || getChildByTag(drawing, "wp:anchor");
+ const extent = container ? getChildByTag(container, "wp:extent") : null;
+
+ const xfrm = getChildByTag(drawing, "a:xfrm");
+ const xfrmExt = xfrm ? getChildByTag(xfrm, "a:ext") : null;
+
+ placeholders.push({{ code, ridPng, ridSvg, extent, xfrmExt, svgBlip }});
+ }}
+
+ if (!placeholders.length) {{
+ const blob = new Blob([bytes], {{ type: "application/vnd.openxmlformats-officedocument.wordprocessingml.document" }});
+ downloadBlob(blob, filename);
+ return;
+ }}
+
+ // Phase 1: Render all Mermaid diagrams sequentially (mermaid needs DOM)
+ const renderResults = [];
+ for (let i = 0; i < placeholders.length; i++) {{
+ const item = placeholders[i];
+ try {{
+ const id = "owui-mermaid-" + i;
+ const rendered = await window.mermaid.render(id, item.code);
+ let svgText = rendered && rendered.svg ? rendered.svg : rendered;
+ if (!svgText || typeof svgText !== "string") throw new Error("Mermaid returned empty SVG");
+
+ svgText = normalizeSvgForWord(svgText);
+ const hasForeignObject = / 0 && vb.height > 0 ? (vb.width / vb.height) : (4/3);
+
+ const widthEmu = maxWidthEmuScaled;
+ const heightEmu = Math.max(1, Math.round(widthEmu / ratio));
+
+ renderResults.push({{ item, svgText, widthEmu, heightEmu, success: true }});
+ }} catch (err) {{
+ console.error("Mermaid render failed for block", i, err);
+ renderResults.push({{ item, svgText: null, widthEmu: 0, heightEmu: 0, success: false }});
+ }}
+ }}
+
+ // Phase 2: Convert SVG to PNG in parallel for performance
+ async function svgToPng(svgText, targetWidthPx, targetHeightPx) {{
+ const canvas = document.createElement("canvas");
+ const ctx = canvas.getContext("2d");
+ const scale = Math.max(1.0, pngScale || 1.0);
+ canvas.width = Math.round(targetWidthPx * scale);
+ canvas.height = Math.round(targetHeightPx * scale);
+ ctx.setTransform(1, 0, 0, 1, 0, 0);
+ if (bgFill) {{
+ ctx.fillStyle = bgFill;
+ ctx.fillRect(0, 0, canvas.width, canvas.height);
+ }}
+ ctx.scale(scale, scale);
+
+ const img = new Image();
+ await new Promise((resolve, reject) => {{
+ img.onload = resolve;
+ img.onerror = reject;
+ img.src = "data:image/svg+xml;base64," + btoa(unescape(encodeURIComponent(svgText)));
+ }});
+
+ ctx.drawImage(img, 0, 0, targetWidthPx, targetHeightPx);
+ const pngDataUrl = canvas.toDataURL("image/png");
+ return pngDataUrl.split(",")[1];
+ }}
+
+ // Create PNG conversion promises for parallel execution
+ const pngPromises = renderResults.map(async (result, i) => {{
+ if (!result.success || !result.svgText) return null;
+ const {{ item, widthEmu, heightEmu }} = result;
+ if (!item.ridPng || !rIdToTarget[item.ridPng]) return null;
+
+ const targetWidthPx = Math.max(1, Math.round(widthEmu / 9525));
+ const targetHeightPx = Math.max(1, Math.round(heightEmu / 9525));
+
+ try {{
+ const pngBase64 = await svgToPng(result.svgText, targetWidthPx, targetHeightPx);
+ return {{ index: i, pngBase64, path: "word/" + rIdToTarget[item.ridPng] }};
+ }} catch (err) {{
+ console.error("PNG conversion failed for block", i, err);
+ return null;
+ }}
+ }});
+
+ // Wait for all PNG conversions to complete
+ const pngResults = await Promise.all(pngPromises);
+
+ // Phase 3: Update ZIP with all results
+ for (let i = 0; i < renderResults.length; i++) {{
+ const result = renderResults[i];
+ if (!result.success) continue;
+
+ const {{ item, svgText, widthEmu, heightEmu }} = result;
+
+ // Update extent in XML
+ if (item.extent) {{
+ item.extent.setAttribute("cx", `${{widthEmu}}`);
+ item.extent.setAttribute("cy", `${{heightEmu}}`);
+ }}
+ if (item.xfrmExt) {{
+ item.xfrmExt.setAttribute("cx", `${{widthEmu}}`);
+ item.xfrmExt.setAttribute("cy", `${{heightEmu}}`);
+ }}
+
+ // Write SVG part
+ if (item.ridSvg && rIdToTarget[item.ridSvg]) {{
+ zip.file("word/" + rIdToTarget[item.ridSvg], svgText);
+ }}
+ }}
+
+ // Write PNG files from parallel results
+ for (const pngResult of pngResults) {{
+ if (pngResult && pngResult.pngBase64) {{
+ zip.file(pngResult.path, pngResult.pngBase64, {{ base64: true }});
+ }}
+ }}
+
+ const newDocXml = new XMLSerializer().serializeToString(xmlDoc);
+ zip.file("word/document.xml", newDocXml);
+
+ const finalBlob = await zip.generateAsync({{
+ type: "blob",
+ compression: "DEFLATE",
+ compressionOptions: {{ level: 6 }},
+ }});
+ downloadBlob(finalBlob, filename);
}} catch (error) {{
- console.error('Error triggering download:', error);
- alert('Export failed: ' + error.message);
+ console.error("Export pipeline failed:", error);
+ const bytes = decodeBase64ToUint8Array(base64Data);
+ const blob = new Blob([bytes], {{ type: "application/vnd.openxmlformats-officedocument.wordprocessingml.document" }});
+ downloadBlob(blob, filename);
}}
}})();
"""
@@ -479,12 +837,17 @@ class Action:
await __event_emitter__(
{
"type": "status",
- "data": {"description": "Word document exported", "done": True},
+ "data": {
+ "description": self._get_msg("exported"),
+ "done": True,
+ },
}
)
await self._send_notification(
- __event_emitter__, "success", f"Successfully exported to {filename}"
+ __event_emitter__,
+ "success",
+ self._get_msg("success", filename=filename),
)
return {"message": "Download triggered"}
@@ -495,7 +858,7 @@ class Action:
{
"type": "status",
"data": {
- "description": f"Export failed: {str(e)}",
+ "description": self._get_msg("export_failed", error=str(e)),
"done": True,
},
}
@@ -503,7 +866,7 @@ class Action:
await self._send_notification(
__event_emitter__,
"error",
- f"Error exporting Word document: {str(e)}",
+ self._get_msg("error_export", error=str(e)),
)
async def generate_title_using_ai(
@@ -593,7 +956,9 @@ class Action:
def _load_chat():
if user_id:
- return Chats.get_chat_by_id_and_user_id(id=chat_id, user_id=user_id)
+ chat = Chats.get_chat_by_id_and_user_id(id=chat_id, user_id=user_id)
+ if chat:
+ return chat
return Chats.get_chat_by_id(chat_id)
try:
@@ -609,9 +974,489 @@ class Action:
title = data.get("title") or getattr(chat, "title", "")
return title.strip() if isinstance(title, str) else ""
+ def clean_filename(self, name: str) -> str:
+ """Clean illegal characters from filename and strip emoji."""
+ if not isinstance(name, str):
+ return ""
+
+ def _is_emoji_codepoint(codepoint: int) -> bool:
+ # Common emoji ranges + flag regional indicators.
+ return (
+ 0x1F000 <= codepoint <= 0x1FAFF
+ or 0x1F1E6 <= codepoint <= 0x1F1FF
+ or 0x2600 <= codepoint <= 0x26FF
+ or 0x2700 <= codepoint <= 0x27BF
+ or 0x2300 <= codepoint <= 0x23FF
+ or 0x2B00 <= codepoint <= 0x2BFF
+ )
+
+ def _is_emoji_modifier(codepoint: int) -> bool:
+ # VS15/VS16, ZWJ, keycap, skin tones, and tag characters used in some emoji sequences.
+ return (
+ codepoint in (0x200D, 0xFE0E, 0xFE0F, 0x20E3)
+ or 0x1F3FB <= codepoint <= 0x1F3FF
+ or 0xE0020 <= codepoint <= 0xE007F
+ )
+
+ without_emoji = "".join(
+ ch
+ for ch in name
+ if not (_is_emoji_codepoint(ord(ch)) or _is_emoji_modifier(ord(ch)))
+ )
+ cleaned = re.sub(r'[\\/*?:"<>|]', "", without_emoji)
+ cleaned = re.sub(r"\s+", " ", cleaned).strip().strip(".")
+ return cleaned[:50].strip()
+
+ def _max_embed_image_bytes(self) -> int:
+ mb = getattr(self.valves, "MAX_EMBED_IMAGE_MB", 20)
+ try:
+ mb_i = int(mb)
+ except Exception:
+ mb_i = 20
+ mb_i = max(1, mb_i)
+ return mb_i * 1024 * 1024
+
+ def _extract_owui_api_file_id(self, url: str) -> Optional[str]:
+ if not isinstance(url, str) or not url:
+ return None
+ m = _OWUI_API_FILE_ID_RE.search(url)
+ if not m:
+ return None
+ fid = (m.group("id") or "").strip()
+ return fid or None
+
+ def _read_file_bytes_limited(self, path: Path, max_bytes: int) -> Optional[bytes]:
+ try:
+ if not path.exists():
+ return None
+ try:
+ size = path.stat().st_size
+ if size > max_bytes:
+ return None
+ except Exception:
+ pass
+ with path.open("rb") as f:
+ data = f.read(max_bytes + 1)
+ if len(data) > max_bytes:
+ return None
+ return data
+ except Exception:
+ return None
+
+ def _decode_base64_limited(self, b64: str, max_bytes: int) -> Optional[bytes]:
+ if not isinstance(b64, str):
+ return None
+ s = re.sub(r"\s+", "", b64.strip())
+ if not s:
+ return None
+
+ # Rough pre-check: base64 expands by ~4/3. Avoid decoding clearly oversized payloads.
+ est = (len(s) * 3) // 4
+ if est > max_bytes:
+ return None
+
+ pad = (-len(s)) % 4
+ if pad:
+ s = s + ("=" * pad)
+ try:
+ out = base64.b64decode(s, validate=False)
+ except (binascii.Error, ValueError):
+ return None
+ if len(out) > max_bytes:
+ return None
+ return out
+
+ def _image_bytes_from_data_url(self, url: str, max_bytes: int) -> Optional[bytes]:
+ if not isinstance(url, str):
+ return None
+ m = _DATA_IMAGE_URL_RE.match(url.strip())
+ if not m:
+ return None
+ b64 = m.group("b64") or ""
+ return self._decode_base64_limited(b64, max_bytes)
+
+ def _image_bytes_from_owui_file_id(
+ self, file_id: str, max_bytes: int
+ ) -> Optional[bytes]:
+ if not file_id or Files is None:
+ return None
+ try:
+ file_obj = Files.get_file_by_id(file_id)
+ except Exception:
+ return None
+ if not file_obj:
+ return None
+
+ # Common patterns across Open WebUI versions / storage backends.
+ data_field = getattr(file_obj, "data", None)
+ if isinstance(data_field, dict):
+ blob_value = data_field.get("bytes")
+ if isinstance(blob_value, (bytes, bytearray)):
+ raw = bytes(blob_value)
+ return raw if len(raw) <= max_bytes else None
+ for key in ("b64", "base64", "data"):
+ inline = data_field.get(key)
+ if isinstance(inline, str) and inline.strip():
+ return self._decode_base64_limited(inline, max_bytes)
+
+ for attr in ("path", "file_path", "absolute_path"):
+ candidate = getattr(file_obj, attr, None)
+ if isinstance(candidate, str) and candidate.strip():
+ raw = self._read_file_bytes_limited(Path(candidate), max_bytes)
+ if raw is not None:
+ return raw
+
+ for attr in ("content", "blob", "data"):
+ raw = getattr(file_obj, attr, None)
+ if isinstance(raw, (bytes, bytearray)):
+ b = bytes(raw)
+ return b if len(b) <= max_bytes else None
+
+ return None
+
+ def _add_image_placeholder(self, paragraph, alt: str, reason: str):
+ label = (alt or "").strip() or "image"
+ msg = f"[{label} not embedded: {reason}]"
+ self._add_text_run(paragraph, msg, bold=False, italic=False, strike=False)
+
+ def _try_embed_image(
+ self, paragraph, image_bytes: bytes
+ ) -> Tuple[bool, Optional[str]]:
+ if not image_bytes:
+ return False, "empty image bytes"
+ try:
+ run = paragraph.add_run()
+ width = None
+ if self._active_doc is not None:
+ try:
+ width = self._available_block_width(self._active_doc)
+ except Exception:
+ width = None
+ run.add_picture(cast(Any, io.BytesIO(image_bytes)), width=width)
+ return True, None
+ except Exception as e:
+ return False, str(e)
+
+ def _embed_markdown_image(self, paragraph, alt: str, url: str):
+ max_bytes = self._max_embed_image_bytes()
+ u = (url or "").strip()
+ if not u:
+ self._add_image_placeholder(paragraph, alt, "missing URL")
+ return
+
+ image_bytes: Optional[bytes] = None
+ if u.lower().startswith("data:"):
+ image_bytes = self._image_bytes_from_data_url(u, max_bytes)
+ if image_bytes is None:
+ self._add_image_placeholder(
+ paragraph,
+ alt,
+ f"invalid data URL or exceeds {self.valves.MAX_EMBED_IMAGE_MB}MB",
+ )
+ return
+ else:
+ file_id = self._extract_owui_api_file_id(u)
+ if not file_id:
+ # External images are not fetched; treat as non-embeddable.
+ self._add_image_placeholder(paragraph, alt, "external URL")
+ return
+ image_bytes = self._image_bytes_from_owui_file_id(file_id, max_bytes)
+ if image_bytes is None:
+ self._add_image_placeholder(
+ paragraph, alt, f"file unavailable ({file_id})"
+ )
+ return
+
+ success, error_msg = self._try_embed_image(paragraph, image_bytes)
+ if not success:
+ self._add_image_placeholder(
+ paragraph, alt, f"unsupported image type: {error_msg}"
+ )
+
+ async def markdown_to_docx(
+ self,
+ markdown_text: str,
+ top_heading: str = "",
+ has_h1: bool = False,
+ sources: Optional[List[dict]] = None,
+ event_emitter: Optional[Callable] = None,
+ ) -> Document:
+ """
+ Convert Markdown text to Word document
+ Supports: headings, paragraphs, bold, italic, code blocks, lists, tables, links
+ Additionally: Mermaid fenced blocks (```mermaid) rendered client-side via Mermaid.js (SVG+PNG),
+ LaTeX math to Word equations, and OpenWebUI citations to References.
+ """
+ doc = Document()
+ self._active_doc = doc
+ try:
+ self._mermaid_figure_counter = 0
+ self._mermaid_placeholder_counter = 0
+ self._caption_style_name = None
+ self._citation_anchor_by_index = {}
+ self._citation_refs = self._build_citation_refs(sources or [])
+ self._bookmark_id_counter = 1
+ for ref in self._citation_refs:
+ self._citation_anchor_by_index[ref.idx] = ref.anchor
+
+ # Set default fonts
+ self.set_document_default_font(doc)
+
+ # If there is no h1 in content, prepend chat title as h1 when provided
+ if top_heading and not has_h1:
+ self.add_heading(doc, top_heading, 1)
+
+ lines = markdown_text.split("\n")
+ i = 0
+ in_code_block = False
+ code_block_content = []
+ code_block_info_raw = ""
+ code_block_lang = ""
+ code_block_attrs: List[str] = []
+ in_math_block = False
+ math_block_delim = ""
+ math_block_lines: List[str] = []
+ in_list = False
+ list_items = []
+ list_type = None # 'ordered' or 'unordered'
+
+ total_lines = len(lines)
+ last_update_time = time.time()
+
+ while i < len(lines):
+ # Update status every 2 seconds
+ if event_emitter and time.time() - last_update_time > 2.0:
+ progress = int((i / total_lines) * 100)
+ await event_emitter(
+ {
+ "type": "status",
+ "data": {
+ "description": f"{self._get_msg('converting')} ({progress}%)",
+ "done": False,
+ },
+ }
+ )
+ last_update_time = time.time()
+
+ line = lines[i]
+
+ # Handle display math blocks (\[...\] or $$...$$)
+ if not in_code_block and self.valves.MATH_ENABLE:
+ single_line = self._extract_single_line_math(line)
+ if single_line is not None:
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+ self._add_display_equation(doc, single_line)
+ i += 1
+ continue
+
+ if not in_math_block:
+ stripped = line.strip()
+ if stripped in (r"\[", "$$"):
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+ in_math_block = True
+ math_block_delim = stripped
+ math_block_lines = []
+ i += 1
+ continue
+ else:
+ stripped = line.strip()
+ close = r"\]" if math_block_delim == r"\[" else "$$"
+ if stripped == close:
+ in_math_block = False
+ latex = "\n".join(math_block_lines).strip()
+ self._add_display_equation(doc, latex)
+ math_block_delim = ""
+ math_block_lines = []
+ i += 1
+ continue
+ math_block_lines.append(line)
+ i += 1
+ continue
+
+ # Handle code blocks
+ if line.strip().startswith("```"):
+ if not in_code_block:
+ # Process pending list first
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ in_code_block = True
+ code_block_info_raw = line.strip()[3:].strip()
+ code_block_lang, code_block_attrs = self._parse_fence_info(
+ code_block_info_raw
+ )
+ code_block_content = []
+ else:
+ # End code block
+ in_code_block = False
+ code_text = "\n".join(code_block_content)
+ if code_block_lang.lower() == "mermaid":
+ self._insert_mermaid_placeholder(doc, code_text)
+ else:
+ self.add_code_block(doc, code_text, code_block_lang)
+ code_block_content = []
+ code_block_info_raw = ""
+ code_block_lang = ""
+ code_block_attrs = []
+ i += 1
+ continue
+
+ if in_code_block:
+ code_block_content.append(line)
+ i += 1
+ continue
+
+ # Handle tables
+ if line.strip().startswith("|") and line.strip().endswith("|"):
+ # Process pending list first
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ table_lines = []
+ while i < len(lines) and lines[i].strip().startswith("|"):
+ table_lines.append(lines[i])
+ i += 1
+ self.add_table(doc, table_lines)
+ continue
+
+ # Handle headings
+ header_match = re.match(r"^(#{1,6})\s+(.+)$", line.strip())
+ if header_match:
+ # Process pending list first
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ level = len(header_match.group(1))
+ text = header_match.group(2)
+ self.add_heading(doc, text, level)
+ i += 1
+ continue
+
+ # Handle unordered lists
+ unordered_match = re.match(r"^(\s*)[-*+]\s+(.+)$", line)
+ if unordered_match:
+ if not in_list or list_type != "unordered":
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = True
+ list_type = "unordered"
+ indent = len(unordered_match.group(1)) // 2
+ list_items.append((indent, unordered_match.group(2)))
+ i += 1
+ continue
+
+ # Handle ordered lists
+ ordered_match = re.match(r"^(\s*)\d+[.)]\s+(.+)$", line)
+ if ordered_match:
+ if not in_list or list_type != "ordered":
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = True
+ list_type = "ordered"
+ indent = len(ordered_match.group(1)) // 2
+ list_items.append((indent, ordered_match.group(2)))
+ i += 1
+ continue
+
+ # Handle blockquotes
+ if line.strip().startswith(">"):
+ # Process pending list first
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ # Collect consecutive quote lines
+ blockquote_lines = []
+ while i < len(lines) and lines[i].strip().startswith(">"):
+ # Remove leading > and optional space
+ quote_line = re.sub(r"^>\s?", "", lines[i])
+ blockquote_lines.append(quote_line)
+ i += 1
+ self.add_blockquote(doc, "\n".join(blockquote_lines))
+ continue
+
+ # Handle horizontal rules
+ if re.match(r"^[-*_]{3,}$", line.strip()):
+ # Process pending list first
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ self.add_horizontal_rule(doc)
+ i += 1
+ continue
+
+ # Handle empty lines
+ if not line.strip():
+ # End list
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+ i += 1
+ continue
+
+ # Handle normal paragraphs
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ self.add_paragraph(doc, line)
+ i += 1
+
+ # Process remaining list
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+
+ # If math block wasn't closed, render it as plain text for robustness.
+ if in_math_block and math_block_lines:
+ self.add_paragraph(doc, r"\[")
+ for l in math_block_lines:
+ self.add_paragraph(doc, l)
+ self.add_paragraph(doc, r"\]")
+
+ if self._citation_refs:
+ self._add_references_section(doc)
+
+ return doc
+ finally:
+ self._active_doc = None
+
+ def _extract_single_line_math(self, line: str) -> Optional[str]:
+ s = line.strip()
+ # \[ ... \]
+ m = re.match(r"^\\\[(.*)\\\]$", s)
+ if m:
+ return m.group(1).strip()
+ # $$ ... $$
+ m = re.match(r"^\$\$(.*)\$\$$", s)
+ if m:
+ return m.group(1).strip()
+ return None
+
def _strip_reasoning_blocks(self, text: str) -> str:
"""
Strip model reasoning blocks from assistant Markdown before export.
+
+ OpenWebUI can include reasoning as interleaved ...
+ (and sometimes / blocks). These should never be exported into DOCX.
"""
if not text:
return text
@@ -655,49 +1500,7 @@ class Action:
xml = f'{omml}'
return parse_xml(xml)
- def _add_inline_equation(
- self,
- paragraph,
- latex: str,
- bold: bool = False,
- italic: bool = False,
- strike: bool = False,
- ):
- latex = (latex or "").strip()
- if not latex:
- return
-
- if not LATEX_MATH_AVAILABLE:
- self._add_text_run(
- paragraph, f"\\({latex}\\)", bold=bold, italic=italic, strike=strike
- )
- return
-
- try:
- mathml = latex_to_mathml(latex)
- omml = mathml2omml.convert(mathml)
- o_math = self._omml_oMath_element(omml)
- run = paragraph.add_run()
- run.bold = bold
- run.italic = italic
- run.font.strike = strike
- cast(Any, run)._r.append(o_math)
- except Exception as exc:
- logger.warning(f"Inline math conversion failed; keeping literal: {exc}")
- self._add_text_run(
- paragraph, f"\\({latex}\\)", bold=bold, italic=italic, strike=strike
- )
-
- def _omml_oMath_element(self, omml: str):
- # Ensure the OMML element declares the math namespace so parse_xml works.
- m_ns = "http://schemas.openxmlformats.org/officeDocument/2006/math"
- s = (omml or "").strip()
- if s.startswith("") and s.endswith(""):
- inner = s[len("") : -len("")]
- s = f'{inner}'
- elif s.startswith("", 1)[0]:
- s = s.replace(" List[_CitationRef]:
citation_idx_map: Dict[str, int] = {}
@@ -805,7 +1608,7 @@ class Action:
cast(Any, paragraph)._p.append(hyperlink)
def _add_references_section(self, doc: Document):
- self.add_heading(doc, "References", 2)
+ self.add_heading(doc, self._get_msg("references"), 2)
for ref in self._citation_refs:
para = doc.add_paragraph(style="List Number")
@@ -818,6 +1621,760 @@ class Action:
para, ref.title, bold=False, italic=False, strike=False
)
+ def _parse_fence_info(self, info_raw: str) -> Tuple[str, List[str]]:
+ parts = [p for p in (info_raw or "").split() if p.strip()]
+ if not parts:
+ return "", []
+ return parts[0], parts[1:]
+
+ def _normalize_mermaid_text(self, source: str) -> str:
+ text = (source or "").replace("\r\n", "\n").replace("\r", "\n").strip()
+ return text + "\n"
+
+ def _prepare_mermaid_for_js(self, source: str) -> str:
+ """
+ Prepare Mermaid source for client-side rendering:
+ - strip title directives (caption already carries it),
+ """
+ text = self._strip_mermaid_title_for_render(source)
+ return text
+
+ def _png_with_text_chunk(self, png_bytes: bytes, keyword: str, value: str) -> bytes:
+ """
+ Ensure placeholder PNGs stay distinct in the DOCX package:
+ python-docx may deduplicate identical image bytes into one media part.
+ We insert a small tEXt chunk so each placeholder is unique, without changing
+ dimensions or requiring external imaging libraries.
+ """
+ if not png_bytes.startswith(b"\x89PNG\r\n\x1a\n"):
+ return png_bytes
+
+ keyword_b = (keyword or "owui").encode("latin-1", errors="ignore")[:79]
+ keyword_b = keyword_b.replace(b"\x00", b"") or b"owui"
+ value_b = (value or "").encode("latin-1", errors="ignore")
+ data = keyword_b + b"\x00" + value_b
+ chunk_type = b"tEXt"
+ crc = zlib.crc32(chunk_type + data) & 0xFFFFFFFF
+ chunk = (
+ struct.pack("!I", len(data)) + chunk_type + data + struct.pack("!I", crc)
+ )
+
+ out = bytearray()
+ out.extend(png_bytes[:8])
+ offset = 8
+ inserted = False
+ while offset + 8 <= len(png_bytes):
+ length = struct.unpack("!I", png_bytes[offset : offset + 4])[0]
+ ctype = png_bytes[offset + 4 : offset + 8]
+ chunk_total = 12 + length
+ if offset + chunk_total > len(png_bytes):
+ break
+ if ctype == b"IEND" and not inserted:
+ out.extend(chunk)
+ inserted = True
+ out.extend(png_bytes[offset : offset + chunk_total])
+ offset += chunk_total
+ if ctype == b"IEND":
+ break
+ if not inserted:
+ return png_bytes
+ return bytes(out)
+
+ def _make_mermaid_placeholder_png(self, seed: str) -> bytes:
+ return self._png_with_text_chunk(_TRANSPARENT_1PX_PNG, "owui", seed)
+
+ def _dummy_mermaid_svg_bytes(self) -> bytes:
+ return (
+ ''
+ ).encode("utf-8")
+
+ def _insert_mermaid_placeholder(self, doc: Document, mermaid_source: str):
+ caption_title: Optional[str] = (
+ self._extract_mermaid_title(mermaid_source)
+ if self.valves.MERMAID_CAPTIONS_ENABLE
+ else None
+ )
+
+ source_for_render = mermaid_source
+ if self.valves.MERMAID_OPTIMIZE_LAYOUT:
+ source_for_render = re.sub(
+ r"^(graph|flowchart)\s+LR\b",
+ r"\1 TD",
+ source_for_render,
+ flags=re.MULTILINE | re.IGNORECASE,
+ )
+ source_for_render = self._prepare_mermaid_for_js(source_for_render)
+
+ self._mermaid_placeholder_counter += 1
+ seed = hashlib.sha256(
+ f"{self._mermaid_placeholder_counter}\n{source_for_render}".encode(
+ "utf-8", errors="replace"
+ )
+ ).hexdigest()[:16]
+ png_bytes = self._make_mermaid_placeholder_png(seed)
+
+ try:
+ shape = doc.add_picture(cast(Any, io.BytesIO(png_bytes)))
+ except Exception as e:
+ logger.warning(f"Failed to add Mermaid placeholder image: {e}")
+ self.add_paragraph(doc, f"[Mermaid placeholder failed: {e}]")
+ return
+ try:
+ doc.paragraphs[-1].alignment = WD_ALIGN_PARAGRAPH.CENTER
+ except Exception:
+ pass
+
+ # Attach a dummy SVG part so we can later overwrite it client-side (SVG+PNG).
+ self._attach_svg_blip(doc, shape, self._dummy_mermaid_svg_bytes())
+
+ try:
+ encoded = quote(source_for_render)
+ inline = shape._inline
+ docPr = inline.docPr
+ docPr.set("descr", f"MERMAID_SRC:{encoded}")
+ docPr.set("title", "Mermaid Diagram Placeholder")
+ except Exception as exc:
+ logger.warning(f"Failed to annotate Mermaid placeholder: {exc}")
+
+ self._add_mermaid_caption(doc, caption_title)
+
+ def _extract_mermaid_title(self, source: str) -> Optional[str]:
+ lines = self._normalize_mermaid_text(source).split("\n")
+ header_found = False
+ for raw in lines:
+ line = raw.strip()
+ if not line:
+ continue
+ if line.startswith("%%{") and line.endswith("}%%"):
+ continue
+ if line.startswith("%%"):
+ continue
+ # diagram header line
+ if not header_found:
+ header_found = True
+ # Mermaid beta/diagram headers can embed a title on the header line, e.g.:
+ # - radar-beta title Foo
+ # - xychart-beta title: "Foo"
+ mt = re.match(
+ r"^(?P\S.*?)(?:\s+title\s*:?\s+)(?P.+)$",
+ line,
+ re.IGNORECASE,
+ )
+ if mt:
+ title = (mt.group("title") or "").strip().strip('"').strip("'")
+ if title:
+ return title
+ continue
+
+ # title "Foo" / title Foo
+ m = re.match(r'^title\s*:?\s+"(.+)"\s*$', line, re.IGNORECASE)
+ if m:
+ return m.group(1).strip()
+ m = re.match(r"^title\s*:?\s+(.+)$", line, re.IGNORECASE)
+ if m:
+ return m.group(1).strip().strip('"').strip("'")
+ return None
+
+ def _strip_mermaid_title_for_render(self, source: str) -> str:
+ """
+ Removes Mermaid title directives from the source before rendering.
+ Captions already carry the title.
+ """
+ lines = self._normalize_mermaid_text(source).split("\n")
+ out: List[str] = []
+ header_found = False
+ title_stripped = False
+ meaningful_after_header = False
+
+ for raw in lines:
+ line = raw.rstrip("\n")
+ stripped = line.strip()
+
+ if not stripped:
+ out.append(line)
+ continue
+
+ if stripped.startswith("%%{") and stripped.endswith("}%%"):
+ out.append(line)
+ continue
+ if stripped.startswith("%%"):
+ out.append(line)
+ continue
+
+ if not header_found:
+ header_found = True
+ # Some Mermaid diagram headers can embed a title on the header line, e.g.:
+ # - radar-beta title Foo
+ # - xychart-beta title: "Foo"
+ mt = re.match(
+ r"^(?P<header>\S.*?)(?:\s+title\s*:?\s+)(?P<title>.+)$",
+ stripped,
+ re.IGNORECASE,
+ )
+ if mt:
+ cleaned = (mt.group("header") or "").strip()
+ out.append(cleaned if cleaned else stripped)
+ title_stripped = True
+ continue
+ out.append(line)
+ continue
+
+ if not title_stripped and not meaningful_after_header:
+ # Strip a standalone title directive line early in the diagram.
+ if re.match(r'^title\s*:?\s+(".+"|.+)$', stripped, re.IGNORECASE):
+ title_stripped = True
+ continue
+
+ # Consider this a meaningful content line after header.
+ meaningful_after_header = True
+ out.append(line)
+
+ return "\n".join(out).strip() + "\n"
+
+ def _ensure_caption_style(self, doc: Document) -> str:
+ if self._caption_style_name is not None:
+ return self._caption_style_name
+
+ style_name = (self.valves.MERMAID_CAPTION_STYLE or "").strip()
+ if style_name == "":
+ # Empty means: do not apply a caption style.
+ self._caption_style_name = ""
+ return ""
+
+ # Prefer existing style if present.
+ try:
+ _ = doc.styles[style_name]
+ self._caption_style_name = style_name
+ return style_name
+ except KeyError:
+ pass
+
+ # If user requested "Caption" but it's missing, create a safe custom style name.
+ if style_name.lower() == "caption":
+ style_name = "OWUI Caption"
+
+ try:
+ _ = doc.styles[style_name]
+ self._caption_style_name = style_name
+ return style_name
+ except KeyError:
+ pass
+
+ try:
+ style = doc.styles.add_style(style_name, WD_STYLE_TYPE.PARAGRAPH)
+ style.font.name = "Calibri"
+ style.font.size = Pt(9)
+ style.font.color.rgb = RGBColor(80, 80, 80)
+ style.paragraph_format.space_before = Pt(2)
+ style.paragraph_format.space_after = Pt(8)
+ self._caption_style_name = style_name
+ return style_name
+ except Exception:
+ self._caption_style_name = "Normal"
+ return "Normal"
+
+ def _add_mermaid_caption(self, doc: Document, title: Optional[str]):
+ if not self.valves.MERMAID_CAPTIONS_ENABLE:
+ return
+
+ # Use configured prefix, or auto-detect from user language
+ prefix = (self.valves.MERMAID_CAPTION_PREFIX or "").strip()
+ if prefix == "":
+ prefix = self._get_msg("figure_prefix")
+
+ if prefix == "" and not title:
+ return
+
+ self._mermaid_figure_counter += 1
+ if prefix == "":
+ caption = title or ""
+ else:
+ base = f"{prefix} {self._mermaid_figure_counter}"
+ caption = f"{base}: {title}" if title else base
+ if caption == "":
+ return
+
+ para = doc.add_paragraph()
+ style_name = self._ensure_caption_style(doc)
+ if style_name:
+ para.style = style_name
+ para.alignment = WD_ALIGN_PARAGRAPH.CENTER
+ self.add_formatted_text(para, caption)
+
+ def _available_block_width(self, doc: Document):
+ section = doc.sections[0]
+ return section.page_width - section.left_margin - section.right_margin
+
+ def _attach_svg_blip(self, doc: Document, inline_shape: Any, svg_bytes: bytes):
+ if not svg_bytes:
+ return
+
+ try:
+ pkg = doc.part.package
+ partname = pkg.next_partname("/word/media/image%d.svg")
+ from docx.opc.part import Part
+
+ svg_part = Part(partname, "image/svg+xml", svg_bytes)
+ rid_svg = doc.part.relate_to(svg_part, RT.IMAGE)
+
+ inline = inline_shape._inline
+ blips = inline.xpath(".//a:blip")
+ if not blips:
+ return
+ blip = blips[0]
+
+ existing = blip.xpath(".//asvg:svgBlip")
+ if existing:
+ existing[0].set(qn("r:embed"), rid_svg)
+ return
+
+ extLst = OxmlElement("a:extLst")
+ ext = OxmlElement("a:ext")
+ ext.set("uri", "{96DAC541-7B7A-43D3-8B79-37D633B846F1}")
+
+ svgBlip = OxmlElement("asvg:svgBlip")
+ svgBlip.set(qn("r:embed"), rid_svg)
+ ext.append(svgBlip)
+ extLst.append(ext)
+ blip.append(extLst)
+ except Exception as exc:
+ logger.warning(f"Failed to attach SVG blip; keeping PNG fallback: {exc}")
+
+ # (Mermaid warning paragraphs removed)
+
+ def set_document_default_font(self, doc: Document):
+ """Set document default font using configured fonts."""
+ style = doc.styles["Normal"]
+ font = style.font
+ font.name = self.valves.FONT_LATIN
+ font.size = Pt(11)
+ # Set Asian font
+ style._element.rPr.rFonts.set(qn("w:eastAsia"), self.valves.FONT_ASIAN)
+
+ # Set paragraph format
+ paragraph_format = style.paragraph_format
+ paragraph_format.line_spacing_rule = WD_LINE_SPACING.ONE_POINT_FIVE
+ paragraph_format.space_after = Pt(6)
+
+ def add_heading(self, doc: Document, text: str, level: int):
+ """Add heading"""
+ # Word heading levels start from 0, Markdown from 1
+ heading_level = min(level, 9) # Word supports up to Heading 9
+ heading = doc.add_heading(level=heading_level)
+
+ # Parse and add formatted text
+ self.add_formatted_text(heading, text)
+
+ def add_paragraph(self, doc: Document, text: str):
+ """Add paragraph with inline formatting support"""
+ paragraph = doc.add_paragraph()
+ self.add_formatted_text(paragraph, text)
+
+ def add_formatted_text(self, paragraph, text: str):
+ """
+ Parse Markdown inline formatting and add to paragraph.
+ Supports: bold, italic, inline code, links, strikethrough, auto-link URLs,
+ and inline LaTeX math \\(...\\) when MATH_ENABLE is on.
+ """
+ self._add_inline_segments(
+ paragraph, text or "", bold=False, italic=False, strike=False
+ )
+
+ def _add_text_run(self, paragraph, s: str, bold: bool, italic: bool, strike: bool):
+ if not s:
+ return
+ run = paragraph.add_run(s)
+ if bold:
+ run.bold = True
+ if italic:
+ run.italic = True
+ if strike:
+ run.font.strike = True
+
+ def _add_inline_code(self, paragraph, s: str):
+ if s == "":
+ return
+
+ def _add_code_run(chunk: str):
+ if not chunk:
+ return
+ run = paragraph.add_run(chunk)
+ run.font.name = self.valves.FONT_CODE
+ run._element.rPr.rFonts.set(qn("w:eastAsia"), self.valves.FONT_CODE)
+ run.font.size = Pt(10)
+ shading = OxmlElement("w:shd")
+ shading.set(qn("w:fill"), "E8E8E8")
+ run._element.rPr.append(shading)
+
+ i = 0
+ for m in _AUTO_URL_RE.finditer(s):
+ start, end = m.span()
+ if start > i:
+ _add_code_run(s[i:start])
+
+ raw = m.group(0)
+ trimmed = raw
+ while trimmed and trimmed[-1] in ".,;:!?)]}":
+ trimmed = trimmed[:-1]
+ suffix = raw[len(trimmed) :]
+
+ normalized = self._normalize_url(trimmed)
+ if normalized:
+ self._add_hyperlink_code(
+ paragraph, display_text=trimmed, url=normalized
+ )
+ else:
+ _add_code_run(raw)
+
+ if suffix:
+ _add_code_run(suffix)
+
+ i = end
+
+ if i < len(s):
+ _add_code_run(s[i:])
+
+ def _add_hyperlink_code(self, paragraph, display_text: str, url: str):
+ u = self._normalize_url(url)
+ if not u:
+ self._add_inline_code(paragraph, display_text)
+ return
+
+ part = getattr(paragraph, "part", None)
+ if part is None or not hasattr(part, "relate_to"):
+ self._add_inline_code(paragraph, display_text)
+ return
+
+ r_id = part.relate_to(u, RT.HYPERLINK, is_external=True)
+
+ hyperlink = OxmlElement("w:hyperlink")
+ hyperlink.set(qn("r:id"), r_id)
+
+ new_run = OxmlElement("w:r")
+ rPr = OxmlElement("w:rPr")
+
+ rFonts = OxmlElement("w:rFonts")
+ rFonts.set(qn("w:ascii"), self.valves.FONT_CODE)
+ rFonts.set(qn("w:hAnsi"), self.valves.FONT_CODE)
+ rFonts.set(qn("w:eastAsia"), self.valves.FONT_CODE)
+ rPr.append(rFonts)
+
+ sz = OxmlElement("w:sz")
+ sz.set(qn("w:val"), "20") # 10pt
+ rPr.append(sz)
+ sz_cs = OxmlElement("w:szCs")
+ sz_cs.set(qn("w:val"), "20")
+ rPr.append(sz_cs)
+
+ shading = OxmlElement("w:shd")
+ shading.set(qn("w:fill"), "E8E8E8")
+ rPr.append(shading)
+
+ new_run.append(rPr)
+
+ t = OxmlElement("w:t")
+ t.text = display_text
+ new_run.append(t)
+
+ hyperlink.append(new_run)
+ cast(Any, paragraph)._p.append(hyperlink)
+
+ def _add_inline_segments(
+ self, paragraph, text: str, bold: bool, italic: bool, strike: bool
+ ):
+ i = 0
+ n = len(text)
+
+ def next_special(start: int) -> int:
+ candidates = []
+ for ch in ("`", "!", "[", "*", "_", "~", "$", "\\"):
+ idx = text.find(ch, start)
+ if idx != -1:
+ candidates.append(idx)
+ idx = text.find(r"\(", start)
+ if idx != -1:
+ candidates.append(idx)
+ idx = text.find("http://", start)
+ if idx != -1:
+ candidates.append(idx)
+ idx = text.find("https://", start)
+ if idx != -1:
+ candidates.append(idx)
+ idx = text.find("www.", start)
+ if idx != -1:
+ candidates.append(idx)
+ return min(candidates) if candidates else n
+
+ while i < n:
+ # Markdown image: 
+ if text.startswith("![", i):
+ close = text.find("]", i + 2)
+ if close != -1 and close + 1 < n and text[close + 1] == "(":
+ close_paren = text.find(")", close + 2)
+ if close_paren != -1:
+ alt = text[i + 2 : close]
+ url = text[close + 2 : close_paren].strip()
+ # Allow angle-bracket wrapped URLs: 
+ if url.startswith("<") and url.endswith(">") and len(url) >= 2:
+ url = url[1:-1].strip()
+ self._embed_markdown_image(paragraph, alt=alt, url=url)
+ i = close_paren + 1
+ continue
+
+ if text[i] == "`":
+ j = text.find("`", i + 1)
+ if j != -1:
+ self._add_inline_code(paragraph, text[i + 1 : j])
+ i = j + 1
+ continue
+
+ if text.startswith(r"\(", i):
+ j = text.find(r"\)", i + 2)
+ if j != -1:
+ self._add_inline_equation(
+ paragraph,
+ text[i + 2 : j],
+ bold=bold,
+ italic=italic,
+ strike=strike,
+ )
+ i = j + 2
+ continue
+
+ # Handle backslash escapes
+ if text[i] == "\\":
+ if i + 1 < n:
+ ch = text[i + 1]
+ # Standard Markdown escapes + $ for math
+ if ch in "\\`*_{}[]()#+-.!|$":
+ self._add_text_run(paragraph, ch, bold, italic, strike)
+ i += 2
+ continue
+ # Keep other backslashes literal
+ self._add_text_run(paragraph, "\\", bold, italic, strike)
+ i += 1
+ continue
+
+ # Handle long run of underscores (fill-in-the-blank)
+ if text[i] == "_":
+ run_len = 0
+ while i + run_len < n and text[i + run_len] == "_":
+ run_len += 1
+ if run_len >= 4:
+ self._add_text_run(
+ paragraph, text[i : i + run_len], bold, italic, strike
+ )
+ i += run_len
+ continue
+
+ # Handle long run of asterisks (separator/mask)
+ if text[i] == "*":
+ run_len = 0
+ while i + run_len < n and text[i + run_len] == "*":
+ run_len += 1
+ if run_len >= 4:
+ self._add_text_run(
+ paragraph, text[i : i + run_len], bold, italic, strike
+ )
+ i += run_len
+ continue
+
+ # Handle long run of tildes (separator)
+ if text[i] == "~":
+ run_len = 0
+ while i + run_len < n and text[i + run_len] == "~":
+ run_len += 1
+ if run_len >= 4:
+ self._add_text_run(
+ paragraph, text[i : i + run_len], bold, italic, strike
+ )
+ i += run_len
+ continue
+
+ # Inline $...$ math (conservative parsing)
+ if (
+ text[i] == "$"
+ and self.valves.MATH_ENABLE
+ and self.valves.MATH_INLINE_DOLLAR_ENABLE
+ ):
+ # Avoid treating $$ as inline math here (block math uses $$ on its own line).
+ if text.startswith("$$", i):
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+
+ # Markdown-ish heuristics to reduce false positives:
+ # - Do not allow whitespace right after opening or right before closing
+ # - Avoid cases like "USD$5" where opening is attached to an alnum
+ if i + 1 >= n or text[i + 1].isspace():
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+ if i > 0 and text[i - 1].isalnum():
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+
+ j = i + 1
+ while True:
+ j = text.find("$", j)
+ if j == -1:
+ break
+ # Skip escaped dollars inside: "\$"
+ if j > 0 and text[j - 1] == "\\":
+ j += 1
+ continue
+ break
+
+ if j != -1:
+ inner = text[i + 1 : j]
+ if (
+ inner
+ and "\n" not in inner
+ and not inner[0].isspace()
+ and not inner[-1].isspace()
+ ):
+ # Treat "$5" as currency more often than math.
+ if _CURRENCY_NUMBER_RE.match(inner) and (
+ i == 0 or text[i - 1].isspace()
+ ):
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+ # Disallow digit immediately following the closing $ (common in prices like "$5.00" already handled above).
+ if j + 1 < n and text[j + 1].isdigit():
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+ self._add_inline_equation(
+ paragraph, inner, bold=bold, italic=italic, strike=strike
+ )
+ i = j + 1
+ continue
+
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+
+ if text.startswith("~~", i):
+ j = text.find("~~", i + 2)
+ if j != -1:
+ self._add_inline_segments(
+ paragraph,
+ text[i + 2 : j],
+ bold=bold,
+ italic=italic,
+ strike=True,
+ )
+ i = j + 2
+ continue
+
+ if text.startswith("**", i):
+ j = text.find("**", i + 2)
+ if j != -1:
+ self._add_inline_segments(
+ paragraph,
+ text[i + 2 : j],
+ bold=True,
+ italic=italic,
+ strike=strike,
+ )
+ i = j + 2
+ continue
+
+ if text.startswith("__", i):
+ j = text.find("__", i + 2)
+ if j != -1:
+ self._add_inline_segments(
+ paragraph,
+ text[i + 2 : j],
+ bold=True,
+ italic=italic,
+ strike=strike,
+ )
+ i = j + 2
+ continue
+
+ if text[i] == "*" and (i + 1 >= n or text[i + 1] != "*"):
+ j = text.find("*", i + 1)
+ if j != -1:
+ self._add_inline_segments(
+ paragraph,
+ text[i + 1 : j],
+ bold=bold,
+ italic=True,
+ strike=strike,
+ )
+ i = j + 1
+ continue
+
+ if text[i] == "_" and (i + 1 >= n or text[i + 1] != "_"):
+ j = text.find("_", i + 1)
+ if j != -1:
+ self._add_inline_segments(
+ paragraph,
+ text[i + 1 : j],
+ bold=bold,
+ italic=True,
+ strike=strike,
+ )
+ i = j + 1
+ continue
+
+ if text[i] == "[":
+ close = text.find("]", i + 1)
+ if close != -1 and close + 1 < n and text[close + 1] == "(":
+ close_paren = text.find(")", close + 2)
+ if close_paren != -1:
+ label = text[i + 1 : close]
+ url = text[close + 2 : close_paren]
+ self._add_hyperlink(paragraph, label, url)
+ i = close_paren + 1
+ continue
+ # Citation marker like [12] -> internal link to References.
+ if close != -1:
+ inner = text[i + 1 : close].strip()
+ if inner.isdigit():
+ idx = int(inner)
+ anchor = self._citation_anchor_by_index.get(idx)
+ if anchor:
+ self._add_internal_hyperlink(paragraph, f"[{idx}]", anchor)
+ i = close + 1
+ continue
+
+ m = _AUTO_URL_RE.match(text, i)
+ if m:
+ raw = m.group(0)
+ trimmed = raw
+ while trimmed and trimmed[-1] in ".,;:!?)]}":
+ trimmed = trimmed[:-1]
+ suffix = raw[len(trimmed) :]
+
+ normalized = self._normalize_url(trimmed)
+ if normalized:
+ # Display the original (trimmed) text; use normalized URL as the target.
+ self._add_hyperlink(
+ paragraph, trimmed, normalized, display_text=trimmed
+ )
+ else:
+ self._add_text_run(paragraph, raw, bold, italic, strike)
+ i += len(raw)
+ continue
+
+ if suffix:
+ self._add_text_run(paragraph, suffix, bold, italic, strike)
+ i += len(raw)
+ continue
+
+ j = next_special(i)
+ if j == i:
+ # Unmatched special character; treat literally to avoid infinite loops.
+ self._add_text_run(paragraph, text[i], bold, italic, strike)
+ i += 1
+ else:
+ self._add_text_run(paragraph, text[i:j], bold, italic, strike)
+ i = j
+
def _normalize_url(self, url: str) -> str:
u = (url or "").strip()
if u.lower().startswith("www."):
@@ -872,566 +2429,49 @@ class Action:
hyperlink.append(new_run)
cast(Any, paragraph)._p.append(hyperlink)
- def _add_text_run(self, paragraph, s: str, bold: bool, italic: bool, strike: bool):
- if not s:
- return
- run = paragraph.add_run(s)
- if bold:
- run.bold = True
- if italic:
- run.italic = True
- if strike:
- run.font.strike = True
-
- # Set Chinese font (copying from existing add_paragraph logic)
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimSun")
-
- def _add_inline_code(self, paragraph, s: str):
- if s == "":
- return
-
- # Simple inline code without URL parsing for now, or copy full logic if needed.
- # For now, just basic styling to match existing.
- run = paragraph.add_run(s)
- run.font.name = "Consolas"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimHei")
- run.font.size = Pt(10)
- shading = OxmlElement("w:shd")
- shading.set(qn("w:fill"), "E8E8E8")
- run._element.rPr.append(shading)
-
- def _add_hyperlink_code(self, paragraph, display_text: str, url: str):
- u = self._normalize_url(url)
- if not u:
- self._add_inline_code(paragraph, display_text)
- return
-
- part = getattr(paragraph, "part", None)
- if part is None or not hasattr(part, "relate_to"):
- self._add_inline_code(paragraph, display_text)
- return
-
- r_id = part.relate_to(u, RT.HYPERLINK, is_external=True)
-
- hyperlink = OxmlElement("w:hyperlink")
- hyperlink.set(qn("r:id"), r_id)
-
- new_run = OxmlElement("w:r")
- rPr = OxmlElement("w:rPr")
-
- rFonts = OxmlElement("w:rFonts")
- rFonts.set(qn("w:ascii"), "Consolas")
- rFonts.set(qn("w:hAnsi"), "Consolas")
- rFonts.set(qn("w:eastAsia"), "SimHei")
- rPr.append(rFonts)
-
- sz = OxmlElement("w:sz")
- sz.set(qn("w:val"), "20") # 10pt
- rPr.append(sz)
-
- shading = OxmlElement("w:shd")
- shading.set(qn("w:fill"), "E8E8E8")
- rPr.append(shading)
-
- new_run.append(rPr)
-
- t = OxmlElement("w:t")
- t.text = display_text
- new_run.append(t)
-
- hyperlink.append(new_run)
- cast(Any, paragraph)._p.append(hyperlink)
-
- def clean_filename(self, name: str) -> str:
- """Clean illegal characters from filename"""
- return re.sub(r'[\\/*?:"<>|]', "", name).strip()[:50]
-
- def markdown_to_docx(
+ def _add_inline_equation(
self,
- markdown_text: str,
- top_heading: str = "",
- has_h1: bool = False,
- sources: Optional[List[dict]] = None,
- ) -> Document:
- """
- Convert Markdown text to Word document
- Supports: headings, paragraphs, bold, italic, code blocks, lists, tables, links,
- native math, citations, and stripped reasoning.
- """
- doc = Document()
+ paragraph,
+ latex: str,
+ bold: bool = False,
+ italic: bool = False,
+ strike: bool = False,
+ ):
+ latex = (latex or "").strip()
+ if not latex:
+ return
- # Set default fonts
- self.set_document_default_font(doc)
+ if not self.valves.MATH_ENABLE or not LATEX_MATH_AVAILABLE:
+ self._add_text_run(
+ paragraph, f"\\({latex}\\)", bold=bold, italic=italic, strike=strike
+ )
+ return
- # Build citation references
- self._citation_refs = self._build_citation_refs(sources)
+ try:
+ mathml = latex_to_mathml(latex)
+ omml = mathml2omml.convert(mathml)
+ o_math = self._omml_oMath_element(omml)
+ run = paragraph.add_run()
+ run.bold = bold
+ run.italic = italic
+ run.font.strike = strike
+ cast(Any, run)._r.append(o_math)
+ except Exception as exc:
+ logger.warning(f"Inline math conversion failed; keeping literal: {exc}")
+ self._add_text_run(
+ paragraph, f"\\({latex}\\)", bold=bold, italic=italic, strike=strike
+ )
- # Strip reasoning blocks
- markdown_text = self._strip_reasoning_blocks(markdown_text)
-
- # If there is no h1 in content, prepend chat title as h1 when provided
- if top_heading and not has_h1:
- self.add_heading(doc, top_heading, 1)
-
- lines = markdown_text.split("\n")
- i = 0
- in_code_block = False
- code_block_content = []
- code_block_lang = ""
- in_list = False
- list_items = []
- list_type = None # 'ordered' or 'unordered'
-
- while i < len(lines):
- line = lines[i]
-
- # Handle code blocks
- if line.strip().startswith("```"):
- if not in_code_block:
- # Process pending list first
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- in_code_block = True
- code_block_info_raw = line.strip()[3:].strip()
- code_block_lang, code_block_attrs = self._parse_fence_info(
- code_block_info_raw
- )
- code_block_content = []
- else:
- # End code block
- in_code_block = False
- code_text = "\n".join(code_block_content)
-
- # Check for Mermaid or Flowchart
- mermaid_langs = {
- "mermaid",
- "flowchart",
- "sequence",
- "gantt",
- "class",
- "state",
- "pie",
- "er",
- "journey",
- "gitgraph",
- "mindmap",
- }
-
- if code_block_lang.lower() in mermaid_langs:
- # Create Mermaid Block Object
- block = _MermaidFenceBlock(
- info_raw=code_block_info_raw,
- language=code_block_lang,
- attrs=code_block_attrs,
- source=code_text,
- )
- # Handle Mermaid diagram
- if code_block_lang == "mermaid":
- # Optimize layout if enabled
- if self.valves.MERMAID_OPTIMIZE_LAYOUT:
- # Replace LR with TD for graph and flowchart
- code_text = re.sub(
- r"^(graph|flowchart)\s+LR\b",
- r"\1 TD",
- code_text,
- flags=re.MULTILINE | re.IGNORECASE,
- )
-
- self._insert_mermaid_placeholder(doc, code_text)
- else:
- # Insert Placeholder using the block object
- self._insert_mermaid_placeholder(doc, block)
- else:
- self.add_code_block(doc, code_text, code_block_lang)
-
- code_block_content = []
- code_block_lang = ""
- i += 1
- continue
-
- if in_code_block:
- code_block_content.append(line)
- i += 1
- continue
-
- # Handle Math Blocks: $$...$$ or \[...\]
- # Simple detection: if line starts with $$ or \[, treat as math block start
- stripped_line = line.strip()
- if stripped_line.startswith("$$") or stripped_line.startswith("\\["):
- # Process pending list first
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- # Check if it's a single-line block like $$ E=mc^2 $$
- if (
- stripped_line.startswith("$$")
- and stripped_line.endswith("$$")
- and len(stripped_line) > 2
- ) or (
- stripped_line.startswith("\\[")
- and stripped_line.endswith("\\]")
- and len(stripped_line) > 2
- ):
- # Extract content
- if stripped_line.startswith("$$"):
- math_content = stripped_line[2:-2]
- else:
- math_content = stripped_line[2:-2]
- self._add_display_equation(doc, math_content)
- i += 1
- continue
-
- # Multi-line math block
- math_lines = []
- # Remove opening marker
- if stripped_line.startswith("$$"):
- current_line_content = stripped_line[2:]
- end_marker = "$$"
- else:
- current_line_content = stripped_line[2:]
- end_marker = "\\]"
-
- if current_line_content.strip():
- math_lines.append(current_line_content)
-
- i += 1
- block_closed = False
- while i < len(lines):
- next_line = lines[i]
- if next_line.strip().endswith(end_marker):
- # Found end
- content_before_end = next_line.strip()[: -len(end_marker)]
- if content_before_end.strip():
- math_lines.append(content_before_end)
- block_closed = True
- i += 1
- break
- math_lines.append(next_line)
- i += 1
-
- self._add_display_equation(doc, "\n".join(math_lines))
- continue
-
- # Handle tables
- if line.strip().startswith("|") and line.strip().endswith("|"):
- # Process pending list first
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- table_lines = []
- while i < len(lines) and lines[i].strip().startswith("|"):
- table_lines.append(lines[i])
- i += 1
- self.add_table(doc, table_lines)
- continue
-
- # Handle headings
- header_match = re.match(r"^(#{1,6})\s+(.+)$", line.strip())
- if header_match:
- # Process pending list first
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- level = len(header_match.group(1))
- text = header_match.group(2)
- self.add_heading(doc, text, level)
- i += 1
- continue
-
- # Handle unordered lists
- unordered_match = re.match(r"^(\s*)[-*+]\s+(.+)$", line)
- if unordered_match:
- if not in_list or list_type != "unordered":
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = True
- list_type = "unordered"
- indent = len(unordered_match.group(1)) // 2
- list_items.append((indent, unordered_match.group(2)))
- i += 1
- continue
-
- # Handle ordered lists
- ordered_match = re.match(r"^(\s*)\d+[.)]\s+(.+)$", line)
- if ordered_match:
- if not in_list or list_type != "ordered":
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = True
- list_type = "ordered"
- indent = len(ordered_match.group(1)) // 2
- list_items.append((indent, ordered_match.group(2)))
- i += 1
- continue
-
- # Handle blockquotes
- if line.strip().startswith(">"):
- # Process pending list first
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- # Collect consecutive quote lines
- blockquote_lines = []
- while i < len(lines) and lines[i].strip().startswith(">"):
- # Remove leading > and optional space
- quote_line = re.sub(r"^>\s?", "", lines[i])
- blockquote_lines.append(quote_line)
- i += 1
- self.add_blockquote(doc, "\n".join(blockquote_lines))
- continue
-
- # Handle horizontal rules
- if re.match(r"^[-*_]{3,}$", line.strip()):
- # Process pending list first
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- self.add_horizontal_rule(doc)
- i += 1
- continue
-
- # Handle empty lines
- if not line.strip():
- # End list
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
- i += 1
- continue
-
- # Handle normal paragraphs
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- self.add_paragraph(doc, line)
- i += 1
-
- # Process remaining list
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
-
- # Add References Section if citations exist
- if self._citation_refs:
- self._add_references_section(doc)
-
- return doc
-
- def set_document_default_font(self, doc: Document):
- """Set document default fonts for both Chinese and English"""
- # Set Normal style
- style = doc.styles["Normal"]
- font = style.font
- font.name = "Times New Roman" # English font
- font.size = Pt(11)
-
- # Set Chinese font
- style._element.rPr.rFonts.set(qn("w:eastAsia"), "SimSun")
-
- # Set paragraph format
- paragraph_format = style.paragraph_format
- paragraph_format.line_spacing_rule = WD_LINE_SPACING.ONE_POINT_FIVE
- paragraph_format.space_after = Pt(6)
-
- def add_heading(self, doc: Document, text: str, level: int):
- """Add heading"""
- # Word heading levels start from 0, Markdown from 1
- heading_level = min(level, 9) # Word supports up to Heading 9
- heading = doc.add_heading(level=heading_level)
-
- # Parse and add formatted text
- self.add_formatted_text(heading, text)
-
- # Set Chinese font
- for run in heading.runs:
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimHei")
- run.font.color.rgb = RGBColor(0, 0, 0)
-
- def add_paragraph(self, doc: Document, text: str):
- """Add paragraph with inline formatting support"""
- paragraph = doc.add_paragraph()
- self.add_formatted_text(paragraph, text)
-
- # Set Chinese font
- for run in paragraph.runs:
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimSun")
-
- def add_formatted_text(self, paragraph, text: str):
- """
- Parse Markdown inline formatting and add to paragraph
- Supports: bold, italic, inline code, links, strikethrough, inline math, citations
- """
- # Define formatting patterns
- patterns = [
- # Inline Math \( ... \)
- (r"\\\((.+?)\\\)", {"math": True}),
- # Inline Math $...$ (single dollar signs, non-greedy)
- (r"(?<!\$)\$(?!\$)([^$]+?)\$(?!\$)", {"math": True}),
- # Citations [1], [2], etc.
- (r"\[(\d+)\]", {"citation": True}),
- # Bold italic ***text*** or ___text___
- (r"\*\*\*(.+?)\*\*\*|___(.+?)___", {"bold": True, "italic": True}),
- # Bold **text** or __text__
- (r"\*\*(.+?)\*\*|__(.+?)__", {"bold": True}),
- # Italic *text* or _text_
- (
- r"(?<!\*)\*(?!\*)(.+?)(?<!\*)\*(?!\*)|(?<!_)_(?!_)(.+?)(?<!_)_(?!_)",
- {"italic": True},
- ),
- # Inline code `code`
- (r"`([^`]+)`", {"code": True}),
- # Link [text](url)
- (r"\[([^\]]+)\]\(([^)]+)\)", {"link": True}),
- # Strikethrough ~~text~~
- (r"~~(.+?)~~", {"strike": True}),
- ]
-
- # Collect all matches
- all_matches = []
-
- for pattern, style in patterns:
- for match in re.finditer(pattern, text):
- # Get matched text content
- groups = match.groups()
- matched_text = next((g for g in groups if g is not None), "")
-
- # Special handling for citations to ensure they map to valid refs
- if style.get("citation"):
- try:
- idx = int(matched_text)
- # Only treat as citation if we have a corresponding reference
- # Check if idx exists in our refs list (1-based index)
- if not any(r.idx == idx for r in self._citation_refs):
- continue
- except ValueError:
- continue
-
- all_matches.append(
- {
- "start": match.start(),
- "end": match.end(),
- "text": matched_text,
- "style": style,
- "full_match": match.group(0),
- "url": (
- groups[1] if style.get("link") and len(groups) > 1 else None
- ),
- }
- )
-
- # Sort by position
- all_matches.sort(key=lambda x: x["start"])
-
- # Remove overlapping matches
- filtered_matches = []
- last_end = 0
- for m in all_matches:
- if m["start"] >= last_end:
- filtered_matches.append(m)
- last_end = m["end"]
-
- # Build final text
- pos = 0
- for match in filtered_matches:
- # Add plain text before match
- if match["start"] > pos:
- plain_text = text[pos : match["start"]]
- if plain_text:
- self._add_text_run(paragraph, plain_text, False, False, False)
-
- # Add formatted text
- style = match["style"]
- run_text = match["text"]
-
- if style.get("math"):
- self._add_inline_equation(paragraph, run_text)
- elif style.get("citation"):
- idx = int(run_text)
- # Find the anchor for this index
- ref = next((r for r in self._citation_refs if r.idx == idx), None)
- if ref:
- self._add_internal_hyperlink(paragraph, f"[{idx}]", ref.anchor)
- else:
- self._add_text_run(paragraph, f"[{idx}]", False, False, False)
- elif style.get("link"):
- # Link handling
- self._add_hyperlink(paragraph, run_text, match["url"])
- elif style.get("code"):
- # Inline code
- self._add_inline_code(paragraph, run_text)
- else:
- # For bold/italic/strike, check if the text contains inline math
- # Pattern for inline math: \(...\) or $...$
- math_pattern = r"(\\\((.+?)\\\)|\$([^$]+?)\$)"
- math_matches = list(re.finditer(math_pattern, run_text))
-
- if math_matches:
- # Process text with inline math
- text_pos = 0
- for math_match in math_matches:
- # Add text before math
- if math_match.start() > text_pos:
- before_text = run_text[text_pos : math_match.start()]
- self._add_text_run(
- paragraph,
- before_text,
- bold=style.get("bold", False),
- italic=style.get("italic", False),
- strike=style.get("strike", False),
- )
- # Add inline equation with formatting
- latex_content = math_match.group(2) or math_match.group(3)
- self._add_inline_equation(
- paragraph,
- latex_content,
- bold=style.get("bold", False),
- italic=style.get("italic", False),
- strike=style.get("strike", False),
- )
- text_pos = math_match.end()
- # Add remaining text after last math
- if text_pos < len(run_text):
- self._add_text_run(
- paragraph,
- run_text[text_pos:],
- bold=style.get("bold", False),
- italic=style.get("italic", False),
- strike=style.get("strike", False),
- )
- else:
- self._add_text_run(
- paragraph,
- run_text,
- bold=style.get("bold", False),
- italic=style.get("italic", False),
- strike=style.get("strike", False),
- )
-
- pos = match["end"]
-
- # Add remaining plain text
- if pos < len(text):
- self._add_text_run(paragraph, text[pos:], False, False, False)
+ def _omml_oMath_element(self, omml: str):
+ # Ensure the OMML element declares the math namespace so parse_xml works.
+ m_ns = "http://schemas.openxmlformats.org/officeDocument/2006/math"
+ s = (omml or "").strip()
+ if s.startswith("<m:oMath>") and s.endswith("</m:oMath>"):
+ inner = s[len("<m:oMath>") : -len("</m:oMath>")]
+ s = f'<m:oMath xmlns:m="{m_ns}">{inner}</m:oMath>'
+ elif s.startswith("<m:oMath") and "xmlns:m=" not in s.split(">", 1)[0]:
+ s = s.replace("<m:oMath", f'<m:oMath xmlns:m="{m_ns}"', 1)
+ return parse_xml(s)
def add_code_block(self, doc: Document, code: str, language: str = ""):
"""Add code block with syntax highlighting"""
@@ -1473,7 +2513,7 @@ class Action:
lang_para.paragraph_format.space_after = Pt(0)
lang_para.paragraph_format.left_indent = Cm(0.5)
lang_run = lang_para.add_run(language.upper())
- lang_run.font.name = "Consolas"
+ lang_run.font.name = self.valves.FONT_CODE
lang_run.font.size = Pt(8)
lang_run.font.color.rgb = RGBColor(100, 100, 100)
lang_run.font.bold = True
@@ -1502,8 +2542,8 @@ class Action:
if not token_value:
continue
run = paragraph.add_run(token_value)
- run.font.name = "Consolas"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimHei")
+ run.font.name = self.valves.FONT_CODE
+ run._element.rPr.rFonts.set(qn("w:eastAsia"), self.valves.FONT_CODE)
run.font.size = Pt(10)
# Apply color
@@ -1517,98 +2557,23 @@ class Action:
else:
# No syntax highlighting, plain text display
run = paragraph.add_run(code)
- run.font.name = "Consolas"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimHei")
+ run.font.name = self.valves.FONT_CODE
+ run._element.rPr.rFonts.set(qn("w:eastAsia"), self.valves.FONT_CODE)
run.font.size = Pt(10)
- def _insert_mermaid_placeholder(
- self, doc: Document, block: Union[_MermaidFenceBlock, str]
- ):
- self._mermaid_figure_counter += 1
- if isinstance(block, str):
- code = block
- else:
- code = block.source
-
- # Create unique transparent PNG for each placeholder
- # By varying image dimensions, we ensure python-docx doesn't reuse the same image file
- # Use figure_counter to create different sizes (1x1, 1x2, 1x3, ...)
- from PIL import Image
-
- # Create a transparent image with size 1 x counter (ensures each is unique)
- img = Image.new("RGBA", (1, self._mermaid_figure_counter), (0, 0, 0, 0))
- image_stream = io.BytesIO()
- img.save(image_stream, format="PNG")
- image_stream.seek(0)
-
- # Add paragraph with center alignment
- paragraph = doc.add_paragraph()
- paragraph.alignment = WD_ALIGN_PARAGRAPH.CENTER
- run = paragraph.add_run()
-
- # Add picture (default size, will be resized by JS)
- # We set a small size initially
- picture = run.add_picture(image_stream, width=Inches(1))
-
- # Set Alt Text (Description) to "MERMAID_SRC:<encoded_code>"
- # This is the magic link between Python and JS
- import urllib.parse
-
- encoded_code = urllib.parse.quote(code)
-
- # Access the underlying XML to set docPr descr
- # picture is an InlineShape, but run.add_picture returns an InlineShape proxy
- # We need to get the wp:docPr element
-
- inline = picture._inline
- docPr = inline.docPr
-
- # Use .set() to ensure attributes are written to XML
- docPr.set("descr", f"MERMAID_SRC:{encoded_code}")
- docPr.set("title", "Mermaid Diagram Placeholder")
-
- # Add Caption
- if self.valves.MERMAID_CAPTIONS_ENABLE:
- self._add_mermaid_caption(doc, self._mermaid_figure_counter)
-
- def _add_mermaid_caption(self, doc: Document, figure_number: int):
- if not self._caption_style_name:
- self._ensure_caption_style(doc)
- self._caption_style_name = self.valves.MERMAID_CAPTION_STYLE
-
- caption_text = f"{self.valves.MERMAID_CAPTION_PREFIX} {figure_number}"
- paragraph = doc.add_paragraph(caption_text, style=self._caption_style_name)
- paragraph.alignment = WD_ALIGN_PARAGRAPH.CENTER
- paragraph.paragraph_format.keep_with_next = False
-
- def _ensure_caption_style(self, doc: Document):
- style_name = self.valves.MERMAID_CAPTION_STYLE
- styles = doc.styles
- if style_name not in styles:
- style = styles.add_style(style_name, WD_STYLE_TYPE.PARAGRAPH)
- style.base_style = styles["Normal"]
- style.next_paragraph_style = styles["Normal"]
- font = style.font
- font.name = "Times New Roman"
- font.size = Pt(10)
- font.italic = True
- font.color.rgb = RGBColor(0x55, 0x55, 0x55) # Dark Grey
-
- def _parse_fence_info(self, info_raw: str) -> Tuple[str, List[str]]:
- parts = info_raw.split()
- if not parts:
- return "", []
- lang = parts[0]
- attrs = parts[1:]
- return lang, attrs
-
def add_table(self, doc: Document, table_lines: List[str]):
- """Add Markdown table with smart sizing, alignment, and hyperlinks/math support in cells."""
+ """Add Markdown table with sane Word sizing/spacing, alignment, and hyperlinks/math support in cells."""
if len(table_lines) < 2:
return
- header_fill = "F2F2F2"
- zebra_fill = "FBFBFB"
+ def _validate_hex(c: str, default: str) -> str:
+ c = c.strip().lstrip("#")
+ if re.fullmatch(r"[0-9A-Fa-f]{6}", c):
+ return c
+ return default
+
+ header_fill = _validate_hex(self.valves.TABLE_HEADER_COLOR, "F2F2F2")
+ zebra_fill = _validate_hex(self.valves.TABLE_ZEBRA_COLOR, "FBFBFB")
def _split_row(line: str) -> List[str]:
# Keep empty cells, trim surrounding pipes.
@@ -1712,7 +2677,7 @@ class Action:
def _fill_cell(cell, text: str, align: WD_ALIGN_PARAGRAPH, bold: bool = False):
cell.text = ""
parts = [
- p for p in re.split(r"(?:<br\s*/?>|\\n)", text or "") if p is not None
+ p for p in re.split(r"(?:<br\s*/?>|\n)", text or "") if p is not None
]
if not parts:
parts = [""]
@@ -1722,8 +2687,6 @@ class Action:
self.add_formatted_text(para, part)
for run in para.runs:
run.font.size = Pt(9)
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimSun")
if bold:
run.bold = True
@@ -1753,10 +2716,6 @@ class Action:
aligns[ci] if ci < len(aligns) else WD_ALIGN_PARAGRAPH.LEFT,
)
- def _available_block_width(self, doc: Document):
- section = doc.sections[0]
- return section.page_width - section.left_margin - section.right_margin
-
def _set_table_cell_margins(
self, table, top: int, bottom: int, left: int, right: int
):
@@ -1800,11 +2759,6 @@ class Action:
# Add formatted text
self.add_formatted_text(paragraph, text)
- # Set font
- for run in paragraph.runs:
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimSun")
-
def add_horizontal_rule(self, doc: Document):
"""Add horizontal rule"""
paragraph = doc.add_paragraph()
@@ -1851,7 +2805,5 @@ class Action:
# Set font to italic gray
for run in paragraph.runs:
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "KaiTi")
run.font.color.rgb = RGBColor(85, 85, 85) # Dark gray text
run.italic = True
diff --git a/plugins/actions/export_to_docx/export_to_word_cn.py b/plugins/actions/export_to_docx/export_to_word_cn.py
index a994ebc..950a198 100644
--- a/plugins/actions/export_to_docx/export_to_word_cn.py
+++ b/plugins/actions/export_to_docx/export_to_word_cn.py
@@ -1,47 +1,51 @@
"""
-title: 导出为 Word
+title: 导出为 Word (增强版)
author: Fu-Jie
author_url: https://github.com/Fu-Jie
funding_url: https://github.com/Fu-Jie/awesome-openwebui
-version: 0.3.0
+version: 0.4.0
icon_url: data:image/svg+xml;base64,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
-requirements: python-docx==1.1.2, Pygments>=2.15.0, latex2mathml, mathml2omml
-description: 将对话导出为 Word (.docx)，支持代码高亮、原生数学公式 (LaTeX)、Mermaid 图表、引用参考和增强表格格式。
+requirements: python-docx, Pygments, latex2mathml, mathml2omml
+description: 将对话导出为 Word (.docx)，支持 Mermaid 图表 (客户端渲染 SVG+PNG)、LaTeX 数学公式、真实超链接、增强表格格式、代码高亮和引用块。
+notes: 基于 rbb-dev 增强版 (https://github.com/rbb-dev/awesome-openwebui) 进一步优化。新增多语言支持、可配置字体/颜色、并行 PNG 渲染优化。
"""
-import os
+from __future__ import annotations
+
import re
import base64
import datetime
+import time
import io
import asyncio
import logging
-from typing import (
- Optional,
- Callable,
- Awaitable,
- Any,
- List,
- Tuple,
- Union,
- Dict,
- Literal,
- cast,
-)
+import hashlib
+import struct
+import zlib
+import binascii
+from pathlib import Path
+from dataclasses import dataclass
+from typing import Optional, Callable, Awaitable, Any, List, Tuple, Dict, cast
+from urllib.parse import quote
from docx import Document
from docx.shared import Pt, Inches, RGBColor, Cm
from docx.enum.text import WD_ALIGN_PARAGRAPH, WD_LINE_SPACING
from docx.enum.table import WD_TABLE_ALIGNMENT
from docx.enum.style import WD_STYLE_TYPE
from docx.opc.constants import RELATIONSHIP_TYPE as RT
-from docx.oxml.ns import qn
-from docx.oxml import OxmlElement, parse_xml
+from docx.oxml import parse_xml
+from docx.oxml.ns import qn, nsmap
+from docx.oxml import OxmlElement
from open_webui.models.chats import Chats
from open_webui.models.users import Users
from open_webui.utils.chat import generate_chat_completion
from pydantic import BaseModel, Field
-from dataclasses import dataclass
+# Files are used to embed internal /api/v1/files/<id>/content images.
+try:
+ from open_webui.models.files import Files # type: ignore
+except Exception: # pragma: no cover - depends on host Open WebUI runtime
+ Files = None
# Pygments for syntax highlighting
try:
@@ -61,6 +65,31 @@ try:
except Exception:
LATEX_MATH_AVAILABLE = False
+
+logging.basicConfig(
+ level=logging.INFO,
+ format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
+)
+logger = logging.getLogger(__name__)
+
+_AUTO_URL_RE = re.compile(r"(?:https?://|www\.)[^\s<>()]+")
+_DATA_IMAGE_URL_RE = re.compile(
+ r"^data:(?P<mime>image/[a-z0-9.+-]+)\s*;\s*base64\s*,\s*(?P<b64>.*)$",
+ re.IGNORECASE | re.DOTALL,
+)
+_OWUI_API_FILE_ID_RE = re.compile(
+ r"/api/v1/files/(?P<id>[A-Za-z0-9-]+)(?:/content)?(?:[/?#]|$)",
+ re.IGNORECASE,
+)
+_CURRENCY_NUMBER_RE = re.compile(r"^\d[\d,]*(?:\.\d+)?$")
+
+_TRANSPARENT_1PX_PNG = base64.b64decode(
+ "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVQImWNgYGBgAAAABQABDQottAAAAABJRU5ErkJggg=="
+)
+
+_ASVG_NS = "http://schemas.microsoft.com/office/drawing/2016/SVG/main"
+nsmap.setdefault("asvg", _ASVG_NS)
+
_REASONING_DETAILS_RE = re.compile(
r"<details\b[^>]*\btype\s*=\s*(?:\"reasoning\"|'reasoning'|reasoning)[^>]*>.*?</details\s*>",
re.IGNORECASE | re.DOTALL,
@@ -71,13 +100,6 @@ _ANALYSIS_RE = re.compile(
)
-logging.basicConfig(
- level=logging.INFO,
- format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
-)
-logger = logging.getLogger(__name__)
-
-
@dataclass(frozen=True)
class _CitationRef:
idx: int
@@ -87,60 +109,203 @@ class _CitationRef:
source_id: str
-@dataclass
-class _MermaidFenceBlock:
- info_raw: str
- language: str
- attrs: List[str]
- source: str
-
-
class Action:
+ # Internationalization message dictionaries
+ _I18N_MESSAGES: Dict[str, Dict[str, str]] = {
+ "en": {
+ "converting": "Converting to Word document...",
+ "exported": "Word document exported",
+ "success": "Successfully exported to {filename}",
+ "error_no_content": "No content found to export!",
+ "error_export": "Error exporting Word document: {error}",
+ "export_failed": "Export failed: {error}",
+ "figure_prefix": "Figure",
+ "references": "References",
+ },
+ "zh": {
+ "converting": "正在转换为 Word 文档...",
+ "exported": "Word 文档导出完成",
+ "success": "成功导出至 {filename}",
+ "error_no_content": "没有找到可导出的内容！",
+ "error_export": "导出 Word 文档时出错: {error}",
+ "export_failed": "导出失败: {error}",
+ "figure_prefix": "图",
+ "references": "参考文献",
+ },
+ }
+
class Valves(BaseModel):
TITLE_SOURCE: str = Field(
default="chat_title",
- description="标题来源: 'chat_title' (对话标题), 'ai_generated' (AI 生成), 'markdown_title' (Markdown 标题)",
+ description="Title Source: 'chat_title' (Chat Title), 'ai_generated' (AI Generated), 'markdown_title' (Markdown Title)",
)
+
+ MAX_EMBED_IMAGE_MB: int = Field(
+ default=20,
+ description="Maximum image size to embed into DOCX (MB). Applies to data URLs and /api/v1/files/<id>/content images.",
+ )
+
+ # Font configuration
+ FONT_LATIN: str = Field(
+ default="Calibri",
+ description="Font for Latin characters (e.g., 'Times New Roman', '', 'Arial')",
+ )
+ FONT_ASIAN: str = Field(
+ default="SimSun",
+ description="Font for Asian characters (e.g., 'SimSun', 'Microsoft YaHei', 'PingFang SC')",
+ )
+ FONT_CODE: str = Field(
+ default="Consolas",
+ description="Font for code blocks and inline code (e.g., 'Consolas', 'Courier New', 'Monaco')",
+ )
+
+ # Table styling
+ TABLE_HEADER_COLOR: str = Field(
+ default="F2F2F2",
+ description="Table header background color (hex, without #)",
+ )
+ TABLE_ZEBRA_COLOR: str = Field(
+ default="FBFBFB",
+ description="Table zebra stripe background color for alternate rows (hex, without #)",
+ )
+
MERMAID_JS_URL: str = Field(
- default="https://cdn.jsdelivr.net/npm/mermaid@10.9.1/dist/mermaid.min.js",
+ default="https://cdn.jsdelivr.net/npm/mermaid@11.12.2/dist/mermaid.min.js",
description="Mermaid JS CDN URL",
)
MERMAID_JSZIP_URL: str = Field(
default="https://cdnjs.cloudflare.com/ajax/libs/jszip/3.10.1/jszip.min.js",
- description="JSZip CDN URL (用于 DOCX 操作)",
- )
- MERMAID_OPTIMIZE_LAYOUT: bool = Field(
- default=True,
- description="优化 Mermaid 布局: 自动将 LR (左右) 转换为 TD (上下) 以适应页面。",
+ description="JSZip CDN URL (DOCX manipulation)",
)
MERMAID_PNG_SCALE: float = Field(
default=3.0,
- description="Mermaid PNG 缩放比例 (分辨率): 越高越清晰但文件越大。默认: 3.0",
+ description="PNG render resolution multiplier (higher = clearer, larger file)",
)
MERMAID_DISPLAY_SCALE: float = Field(
- default=1.5,
- description="Mermaid 显示比例 (视觉大小): >1.0 放大, <1.0 缩小。默认: 1.5",
+ default=1.0,
+ description="Diagram width relative to available page width (<=1 recommended)",
)
+ MERMAID_OPTIMIZE_LAYOUT: bool = Field(
+ default=False,
+ description="Optimize Mermaid layout: convert LR to TD for graph/flowchart",
+ )
+ MERMAID_BACKGROUND: str = Field(
+ default="",
+ description="Mermaid background color. Empty = transparent (recommended for Word dark mode). Used only for optional PNG fill.",
+ )
+
MERMAID_CAPTIONS_ENABLE: bool = Field(
default=True,
- description="启用 Mermaid 图表标题",
+ description="Add figure captions under Mermaid images/charts",
)
MERMAID_CAPTION_STYLE: str = Field(
default="Caption",
- description="Mermaid 标题样式名称",
+ description="Paragraph style name for Mermaid captions (uses 'Caption' if available, otherwise creates a safe custom style)",
)
MERMAID_CAPTION_PREFIX: str = Field(
- default="图",
- description="Mermaid 标题前缀",
+ default="",
+ description="Caption prefix label (e.g., 'Figure' or '图'). Empty = auto-detect based on user language.",
+ )
+
+ MATH_ENABLE: bool = Field(
+ default=True,
+ description="Enable LaTeX math block conversion (\\[...\\] and $$...$$) into Word equations",
+ )
+ MATH_INLINE_DOLLAR_ENABLE: bool = Field(
+ default=True,
+ description="Enable inline $...$ math conversion into Word equations (conservative parsing to reduce false positives)",
+ )
+
+ # Language configuration
+ UI_LANGUAGE: str = Field(
+ default="zh",
+ description="UI language for export messages. Options: 'en' (English), 'zh' (Chinese)",
+ )
+
+ class UserValves(BaseModel):
+ TITLE_SOURCE: str = Field(
+ default="chat_title",
+ description="Title Source: 'chat_title' (Chat Title), 'ai_generated' (AI Generated), 'markdown_title' (Markdown Title)",
+ )
+ UI_LANGUAGE: str = Field(
+ default="zh",
+ description="UI language for export messages. Options: 'en' (English), 'zh' (Chinese)",
+ )
+ FONT_LATIN: str = Field(
+ default="Calibri",
+ description="Font for Latin characters (e.g., 'Times New Roman', '', 'Arial')",
+ )
+ FONT_ASIAN: str = Field(
+ default="SimSun",
+ description="Font for Asian characters (e.g., 'SimSun', 'Microsoft YaHei', 'PingFang SC')",
+ )
+ FONT_CODE: str = Field(
+ default="Consolas",
+ description="Font for code blocks and inline code (e.g., 'Consolas', 'Courier New', 'Monaco')",
+ )
+ TABLE_HEADER_COLOR: str = Field(
+ default="F2F2F2",
+ description="Table header background color (hex, without #)",
+ )
+ TABLE_ZEBRA_COLOR: str = Field(
+ default="FBFBFB",
+ description="Table zebra stripe background color for alternate rows (hex, without #)",
+ )
+ MERMAID_PNG_SCALE: float = Field(
+ default=3.0,
+ description="PNG render resolution multiplier (higher = clearer, larger file)",
+ )
+ MERMAID_DISPLAY_SCALE: float = Field(
+ default=1.0,
+ description="Diagram width relative to available page width (<=1 recommended)",
+ )
+ MERMAID_OPTIMIZE_LAYOUT: bool = Field(
+ default=False,
+ description="Optimize Mermaid layout: convert LR to TD for graph/flowchart",
+ )
+ MERMAID_BACKGROUND: str = Field(
+ default="",
+ description="Mermaid background color. Empty = transparent (recommended for Word dark mode). Used only for optional PNG fill.",
+ )
+ MERMAID_CAPTIONS_ENABLE: bool = Field(
+ default=True,
+ description="Add figure captions under Mermaid images/charts",
+ )
+ MATH_ENABLE: bool = Field(
+ default=True,
+ description="Enable LaTeX math block conversion (\\\\[...\\\\] and $$...$$) into Word equations",
+ )
+ MATH_INLINE_DOLLAR_ENABLE: bool = Field(
+ default=True,
+ description="Enable inline $...$ math conversion into Word equations (conservative parsing to reduce false positives)",
)
def __init__(self):
self.valves = self.Valves()
- self._mermaid_figure_counter = 0
- self._caption_style_name = ""
+ self._mermaid_figure_counter: int = 0
+ self._mermaid_placeholder_counter: int = 0
+ self._caption_style_name: Optional[str] = None
self._citation_anchor_by_index: Dict[int, str] = {}
self._citation_refs: List[_CitationRef] = []
self._bookmark_id_counter: int = 1
+ self._active_doc: Optional[Document] = None
+ self._user_lang: str = "en" # Will be set per-request
+
+ def _get_lang_key(self, user_language: str) -> str:
+ """Convert user language code to i18n key (e.g., 'zh-CN' -> 'zh', 'en-US' -> 'en')."""
+ lang = (user_language or "en").lower().split("-")[0]
+ return lang if lang in self._I18N_MESSAGES else "en"
+
+ def _get_msg(self, key: str, **kwargs) -> str:
+ """Get internationalized message by key with optional formatting."""
+ messages = self._I18N_MESSAGES.get(self._user_lang, self._I18N_MESSAGES["en"])
+ msg = messages.get(key, self._I18N_MESSAGES["en"].get(key, key))
+ if kwargs:
+ try:
+ return msg.format(**kwargs)
+ except KeyError:
+ return msg
+ return msg
async def _send_notification(self, emitter: Callable, type: str, content: str):
await emitter(
@@ -158,83 +323,96 @@ class Action:
):
logger.info(f"action:{__name__}")
- # Reset counters for new request
- self._mermaid_figure_counter = 0
- self._bookmark_id_counter = 1
-
- # 解析用户信息
+ # Parse user info
+ user_name = "User"
+ user_id = "unknown_user"
if isinstance(__user__, (list, tuple)):
- user_language = (
- __user__[0].get("language", "zh-CN") if __user__ else "zh-CN"
- )
- user_name = __user__[0].get("name", "用户") if __user__[0] else "用户"
+ user_name = __user__[0].get("name", "User") if __user__[0] else "User"
user_id = (
__user__[0]["id"]
if __user__ and "id" in __user__[0]
else "unknown_user"
)
elif isinstance(__user__, dict):
- user_language = __user__.get("language", "zh-CN")
- user_name = __user__.get("name", "用户")
+ user_name = __user__.get("name", "User")
user_id = __user__.get("id", "unknown_user")
+ # Apply UserValves if present
+ if __user__ and "valves" in __user__:
+ # Update self.valves with user-specific values
+ for key, value in __user__["valves"].model_dump().items():
+ if hasattr(self.valves, key):
+ setattr(self.valves, key, value)
+
+ # Get user language from Valves configuration
+ self._user_lang = self._get_lang_key(self.valves.UI_LANGUAGE)
+
if __event_emitter__:
last_assistant_message = body["messages"][-1]
await __event_emitter__(
{
"type": "status",
- "data": {"description": "正在转换为 Word 文档...", "done": False},
+ "data": {
+ "description": self._get_msg("converting"),
+ "done": False,
+ },
}
)
try:
message_content = last_assistant_message["content"]
+ if isinstance(message_content, str):
+ message_content = self._strip_reasoning_blocks(message_content)
if not message_content or not message_content.strip():
await self._send_notification(
- __event_emitter__, "error", "没有找到可导出的内容！"
+ __event_emitter__, "error", self._get_msg("error_no_content")
)
return
- # 生成文件名
+ # Generate filename
title = ""
chat_id = self.extract_chat_id(body, __metadata__)
- # 直接通过 chat_id 获取标题，因为 body 中通常不包含标题
+ # Fetch chat_title directly via chat_id as it's usually missing in body
chat_title = ""
if chat_id:
chat_title = await self.fetch_chat_title(chat_id, user_id)
- # 根据配置决定文件名使用的标题
if (
- self.valves.TITLE_SOURCE == "chat_title"
- or not self.valves.TITLE_SOURCE
+ self.valves.TITLE_SOURCE.strip() == "chat_title"
+ or not self.valves.TITLE_SOURCE.strip()
):
title = chat_title
- elif self.valves.TITLE_SOURCE == "markdown_title":
+ elif self.valves.TITLE_SOURCE.strip() == "markdown_title":
title = self.extract_title(message_content)
- elif self.valves.TITLE_SOURCE == "ai_generated":
+ elif self.valves.TITLE_SOURCE.strip() == "ai_generated":
title = await self.generate_title_using_ai(
body, message_content, user_id, __request__
)
+ # Fallback logic
+ if not title:
+ if self.valves.TITLE_SOURCE.strip() != "chat_title" and chat_title:
+ title = chat_title
+ elif self.valves.TITLE_SOURCE.strip() != "markdown_title":
+ extracted = self.extract_title(message_content)
+ if extracted:
+ title = extracted
+
current_datetime = datetime.datetime.now()
formatted_date = current_datetime.strftime("%Y%m%d")
- if title:
- filename = f"{self.clean_filename(title)}.docx"
+ cleaned_title = self.clean_filename(title) if title else ""
+ if cleaned_title:
+ filename = f"{cleaned_title}.docx"
else:
- filename = f"{user_name}_{formatted_date}.docx"
+ clean_user = self.clean_filename(user_name)
+ filename = f"{clean_user}_{formatted_date}.docx"
- # 创建 Word 文档；若正文无一级标题，使用对话标题作为一级标题
- # 如果选择了 chat_title 且获取到了，则作为 top_heading
- # 如果选择了其他方式，title 就是文件名，也可以作为 top_heading
-
- # 保持原有逻辑：top_heading 主要是为了在文档开头补充标题
- # 这里我们尽量使用 chat_title 作为 top_heading，如果它存在的话，因为它通常是对话的主题
- # 即使文件名是 AI 生成的，文档内的标题用 chat_title 也是合理的
- # 但如果用户选择了 markdown_title，可能不希望插入 chat_title
+ # Escape filename for JS string
+ js_filename = filename.replace("\\", "\\\\").replace('"', '\\"')
top_heading = ""
if chat_title:
@@ -242,35 +420,27 @@ class Action:
elif title:
top_heading = title
+ # Create Word document; if no h1 exists, inject chat title as h1
has_h1 = bool(re.search(r"^#\s+.+$", message_content, re.MULTILINE))
-
- # Extract sources if available (for citations)
sources = (
last_assistant_message.get("sources") or body.get("sources") or []
)
-
- doc = self.markdown_to_docx(
+ doc = await self.markdown_to_docx(
message_content,
top_heading=top_heading,
has_h1=has_h1,
sources=sources,
+ event_emitter=__event_emitter__,
)
- # 保存到内存
+ # Save to memory
doc_buffer = io.BytesIO()
doc.save(doc_buffer)
doc_buffer.seek(0)
file_content = doc_buffer.read()
base64_blob = base64.b64encode(file_content).decode("utf-8")
- # Escape message_content for JavaScript template literal
- escaped_content = (
- message_content.replace("\\", "\\\\") # Escape backslashes first
- .replace("`", "\\`") # Escape backticks
- .replace("${", "\\${") # Escape template literal expressions
- )
-
- # 触发文件下载
+ # Trigger file download
if __event_call__:
await __event_call__(
{
@@ -278,183 +448,19 @@ class Action:
"data": {
"code": f"""
(async function() {{
- try {{
- // Parse document.xml to find placeholders and extract optimized code
- // We do this FIRST to get the actual code to render (which might have been optimized in Python)
-
- // Load JSZip
- if (!window.JSZip) {{
- await new Promise((resolve, reject) => {{
- const script = document.createElement("script");
- script.src = "{self.valves.MERMAID_JSZIP_URL}";
- script.onload = resolve;
- script.onerror = reject;
- document.head.appendChild(script);
- }});
- }}
+ const base64Data = "{base64_blob}";
+ const filename = "{js_filename}";
+ const mermaidUrl = "{self.valves.MERMAID_JS_URL}";
+ const jszipUrl = "{self.valves.MERMAID_JSZIP_URL}";
+ const pngScale = {float(self.valves.MERMAID_PNG_SCALE)};
+ const displayScale = {float(self.valves.MERMAID_DISPLAY_SCALE)};
+ const bgRaw = "{(self.valves.MERMAID_BACKGROUND or '').strip()}";
+ const bg = (bgRaw || "").trim();
+ const bgFill = (bg && bg.toLowerCase() !== "transparent") ? bg : "";
+ const themeBackground = bgFill || "transparent";
- const base64Data = "{base64_blob}";
- const binaryData = atob(base64Data);
- const arrayBuffer = new Uint8Array(binaryData.length);
- for (let i = 0; i < binaryData.length; i++) {{
- arrayBuffer[i] = binaryData.charCodeAt(i);
- }}
-
- const zip = new JSZip();
- await zip.loadAsync(arrayBuffer);
-
- // Parse document.xml
- const docXml = await zip.file("word/document.xml").async("string");
- const parser = new DOMParser();
- const xmlDoc = parser.parseFromString(docXml, "application/xml");
-
- const drawings = xmlDoc.getElementsByTagName("w:drawing");
- const placeholderInfo = [];
-
- for (let i = 0; i < drawings.length; i++) {{
- const drawing = drawings[i];
- const docPr = drawing.getElementsByTagName("wp:docPr")[0];
- if (docPr) {{
- const descr = docPr.getAttribute("descr");
- if (descr && descr.startsWith("MERMAID_SRC:")) {{
- const encodedCode = descr.substring("MERMAID_SRC:".length);
- const code = decodeURIComponent(encodedCode);
-
- // Find the blip and extent to replace
- const parent = drawing.parentNode; // w:r usually, or w:drawing parent
- // We need to find a:blip and wp:extent within this drawing
- const blip = drawing.getElementsByTagName("a:blip")[0];
- const extent = drawing.getElementsByTagName("wp:extent")[0];
-
- if (blip && extent) {{
- const rId = blip.getAttribute("r:embed");
- placeholderInfo.push({{ rId, extent, code }});
- }}
- }}
- }}
- }}
-
- if (placeholderInfo.length === 0) {{
- console.log("No Mermaid placeholders found in DOCX.");
- // Just download the file as is
- const blob = new Blob([arrayBuffer], {{type: "application/vnd.openxmlformats-officedocument.wordprocessingml.document"}});
- const url = URL.createObjectURL(blob);
- const a = document.createElement("a");
- a.style.display = "none";
- a.href = url;
- a.download = "{filename}";
- document.body.appendChild(a);
- a.click();
- URL.revokeObjectURL(url);
- document.body.removeChild(a);
- return;
- }}
-
- console.log(`Found ${{placeholderInfo.length}} Mermaid placeholders.`);
-
- // Load Mermaid
- if (!window.mermaid) {{
- await new Promise((resolve, reject) => {{
- const script = document.createElement("script");
- script.src = "{self.valves.MERMAID_JS_URL}";
- script.onload = resolve;
- script.onerror = reject;
- document.head.appendChild(script);
- }});
- }}
-
- mermaid.initialize({{
- startOnLoad: false,
- theme: 'default',
- }});
-
- // Read rels XML once
- const relsXml = await zip.file("word/_rels/document.xml.rels").async("string");
- const relsDoc = parser.parseFromString(relsXml, "application/xml");
- const relationships = relsDoc.getElementsByTagName("Relationship");
- const rIdToPath = {{}};
-
- for (let i = 0; i < relationships.length; i++) {{
- const rel = relationships[i];
- rIdToPath[rel.getAttribute("Id")] = rel.getAttribute("Target");
- }}
-
- // Render and replace
- console.log(`Processing ${{placeholderInfo.length}} diagrams...`);
-
- for (let i = 0; i < placeholderInfo.length; i++) {{
- const {{ rId, extent, code }} = placeholderInfo[i];
- const imagePath = "word/" + rIdToPath[rId];
-
- console.log(`Block ${{i + 1}}/${{placeholderInfo.length}}: Rendering and replacing at ${{imagePath}}`);
-
- // Render SVG
- const id = "mermaid-export-" + i;
- const {{ svg }} = await mermaid.render(id, code);
-
- // Convert SVG to PNG
- const canvas = document.createElement("canvas");
- const ctx = canvas.getContext("2d");
- const img = new Image();
-
- // Get SVG dimensions
- const svgMatch = svg.match(/viewBox="[^"]*\s+[^"]*\s+([^"\s]+)\s+([^"\s]+)"/);
- let width = 800;
- let height = 600;
- if (svgMatch) {{
- width = parseFloat(svgMatch[1]);
- height = parseFloat(svgMatch[2]);
- }}
-
- // Scale up for better quality
- const scale = {self.valves.MERMAID_PNG_SCALE};
- canvas.width = width * scale;
- canvas.height = height * scale;
-
- await new Promise((resolve, reject) => {{
- img.onload = resolve;
- img.onerror = reject;
- img.src = "data:image/svg+xml;base64," + btoa(unescape(encodeURIComponent(svg)));
- }});
-
- ctx.scale(scale, scale);
- ctx.drawImage(img, 0, 0, width, height);
-
- const pngDataUrl = canvas.toDataURL("image/png");
- const pngBase64 = pngDataUrl.split(",")[1];
-
- // Replace image in ZIP
- zip.file(imagePath, pngBase64, {{base64: true}});
-
- // Update dimensions in document.xml (EMUs)
- // 1 inch = 914400 EMUs, 1 pixel ≈ 9525 EMUs at 96 DPI
- // Max width: ~6 inches (page width minus margins)
- const maxWidthEmu = 5486400; // 6 inches
- const displayScale = {self.valves.MERMAID_DISPLAY_SCALE};
- let emuWidth = Math.round(width * 9525 * displayScale);
- let emuHeight = Math.round(height * 9525 * displayScale);
-
- // Scale down if too wide
- if (emuWidth > maxWidthEmu) {{
- const scaleFactor = maxWidthEmu / emuWidth;
- emuWidth = maxWidthEmu;
- emuHeight = Math.round(emuHeight * scaleFactor);
- }}
-
- extent.setAttribute("cx", emuWidth);
- extent.setAttribute("cy", emuHeight);
- }}
-
- // Serialize updated XML
- const serializer = new XMLSerializer();
- const newDocXml = serializer.serializeToString(xmlDoc);
- zip.file("word/document.xml", newDocXml);
-
- // Generate final blob
- const finalBlob = await zip.generateAsync({{type: "blob"}});
- const filename = "{filename}";
-
- const url = URL.createObjectURL(finalBlob);
+ function downloadBlob(blob, filename) {{
+ const url = URL.createObjectURL(blob);
const a = document.createElement("a");
a.style.display = "none";
a.href = url;
@@ -463,9 +469,362 @@ class Action:
a.click();
URL.revokeObjectURL(url);
document.body.removeChild(a);
+ }}
+
+ async function loadScript(url, globalName) {{
+ if (globalName && window[globalName]) return;
+ await new Promise((resolve, reject) => {{
+ const script = document.createElement("script");
+ script.src = url;
+ script.onload = resolve;
+ script.onerror = reject;
+ document.head.appendChild(script);
+ }});
+ }}
+
+ function decodeBase64ToUint8Array(b64) {{
+ const binary = atob(b64);
+ const bytes = new Uint8Array(binary.length);
+ for (let i = 0; i < binary.length; i++) bytes[i] = binary.charCodeAt(i);
+ return bytes;
+ }}
+
+ function parseViewBox(vb) {{
+ if (!vb) return null;
+ const parts = vb.trim().split(/\\s+/).map(Number);
+ if (parts.length !== 4 || parts.some((n) => !isFinite(n))) return null;
+ return {{ minX: parts[0], minY: parts[1], width: parts[2], height: parts[3] }};
+ }}
+
+ function normalizeSvgForWord(svgText) {{
+ const parser = new DOMParser();
+ const doc = parser.parseFromString(svgText, "image/svg+xml");
+ const svgEl = doc.documentElement;
+ if (!svgEl || svgEl.tagName.toLowerCase() !== "svg") return svgText;
+
+ // Pad viewBox a little to reduce clipping in Word.
+ const vb0 = parseViewBox(svgEl.getAttribute("viewBox"));
+ if (vb0 && vb0.width > 0 && vb0.height > 0) {{
+ const minDim = Math.min(vb0.width, vb0.height);
+ let pad = Math.max(8.0, minDim * 0.02);
+ pad = Math.min(pad, 24.0);
+ const vb = {{
+ minX: vb0.minX - pad,
+ minY: vb0.minY - pad,
+ width: vb0.width + 2 * pad,
+ height: vb0.height + 2 * pad,
+ }};
+ svgEl.setAttribute("viewBox", `${{vb.minX}} ${{vb.minY}} ${{vb.width}} ${{vb.height}}`);
+ }}
+
+ const vb = parseViewBox(svgEl.getAttribute("viewBox"));
+ const widthAttr = (svgEl.getAttribute("width") || "").trim();
+ const heightAttr = (svgEl.getAttribute("height") || "").trim();
+ const widthPct = widthAttr.endsWith("%");
+ const heightPct = heightAttr.endsWith("%");
+ if (vb && vb.width > 0 && vb.height > 0 && (!widthAttr || !heightAttr || widthPct || heightPct)) {{
+ svgEl.setAttribute("width", `${{vb.width}}`);
+ svgEl.setAttribute("height", `${{vb.height}}`);
+ }}
+
+ svgEl.removeAttribute("style");
+ svgEl.setAttribute("preserveAspectRatio", "xMidYMid meet");
+ svgEl.setAttribute("overflow", "visible");
+
+ const removeNode = (n) => {{
+ try {{ n && n.parentNode && n.parentNode.removeChild(n); }} catch (_e) {{}}
+ }};
+
+ // Remove Mermaid/OWUI background rectangles to avoid \"white box\" rendering in Word dark mode.
+ svgEl
+ .querySelectorAll('rect[data-owui-bg=\"1\"], rect.background, rect[class~=\"background\"], rect#background')
+ .forEach(removeNode);
+ try {{
+ const isWhiteish = (fill) => {{
+ const f = (fill || "").trim().toLowerCase();
+ return (
+ f === "white" ||
+ f === "#fff" ||
+ f === "#ffffff" ||
+ f === "rgb(255,255,255)" ||
+ f === "rgb(255, 255, 255)"
+ );
+ }};
+ const nearly = (a, b) => Math.abs(a - b) <= 1e-3;
+ const rectMatches = (r, box) => {{
+ if (!box) return false;
+ const x = parseFloat(r.getAttribute("x") || "0");
+ const y = parseFloat(r.getAttribute("y") || "0");
+ const w = parseFloat(r.getAttribute("width") || "");
+ const h = parseFloat(r.getAttribute("height") || "");
+ if (!isFinite(x) || !isFinite(y) || !isFinite(w) || !isFinite(h)) return false;
+ return (
+ nearly(x, box.minX) &&
+ nearly(y, box.minY) &&
+ nearly(w, box.width) &&
+ nearly(h, box.height)
+ );
+ }};
+ const vbNow = parseViewBox(svgEl.getAttribute("viewBox"));
+ svgEl.querySelectorAll("rect[fill]").forEach((r) => {{
+ const fill = r.getAttribute("fill");
+ if (!isWhiteish(fill)) return;
+ if (rectMatches(r, vb0) || rectMatches(r, vbNow)) removeNode(r);
+ }});
+ }} catch (_e) {{}}
+ try {{
+ const vbCanvas = parseViewBox(svgEl.getAttribute(\"viewBox\")) || vb0 || vb;
+ if (vbCanvas) {{
+ const existing = svgEl.querySelector('rect[data-owui-canvas=\"1\"]');
+ const rect = existing || doc.createElementNS(\"http://www.w3.org/2000/svg\", \"rect\");
+ rect.setAttribute(\"data-owui-canvas\", \"1\");
+ rect.setAttribute(\"x\", `${{vbCanvas.minX}}`);
+ rect.setAttribute(\"y\", `${{vbCanvas.minY}}`);
+ rect.setAttribute(\"width\", `${{vbCanvas.width}}`);
+ rect.setAttribute(\"height\", `${{vbCanvas.height}}`);
+ rect.setAttribute(\"fill\", \"#FFFFFF\");
+ // Word quirk: without a full-canvas rect with *non-zero* opacity, Word will often
+ // only offer \"Convert to Shape\" when clicking on an actual stroke/fill (not empty space).
+ // We keep this rect nearly transparent and non-interactive.
+ rect.setAttribute(\"fill-opacity\", \"0.001\");
+ rect.setAttribute(\"stroke\", \"none\");
+ rect.setAttribute(\"stroke-opacity\", \"0\");
+ rect.setAttribute(\"pointer-events\", \"none\");
+ if (!existing) {{
+ const first = svgEl.firstChild;
+ svgEl.insertBefore(rect, first);
+ }}
+ }}
+ }} catch (_e) {{}}
+
+ return new XMLSerializer().serializeToString(svgEl);
+ }}
+
+ function getMaxWidthEmu(xmlDoc) {{
+ try {{
+ const sects = xmlDoc.getElementsByTagName("w:sectPr");
+ const sect = sects && sects.length ? sects[sects.length - 1] : null;
+ if (!sect) return 5486400; // 6 in
+ const pgSz = sect.getElementsByTagName("w:pgSz")[0];
+ const pgMar = sect.getElementsByTagName("w:pgMar")[0];
+ if (!pgSz || !pgMar) return 5486400;
+ const pageW = parseInt(pgSz.getAttribute("w:w") || "", 10);
+ const left = parseInt(pgMar.getAttribute("w:left") || "", 10);
+ const right = parseInt(pgMar.getAttribute("w:right") || "", 10);
+ if (!isFinite(pageW) || !isFinite(left) || !isFinite(right)) return 5486400;
+ const twips = Math.max(1, pageW - left - right);
+ return Math.round(twips * 635); // 1 twip = 635 EMU
+ }} catch (_e) {{
+ return 5486400;
+ }}
+ }}
+
+ function getChildByTag(parent, tag) {{
+ const nodes = parent.getElementsByTagName(tag);
+ return nodes && nodes.length ? nodes[0] : null;
+ }}
+
+ try {{
+ await loadScript(jszipUrl, "JSZip");
+ await loadScript(mermaidUrl, "mermaid");
+
+ // Mermaid init: disable htmlLabels to keep SVG Word-friendly; PNG fallback still included.
+ try {{
+ window.mermaid.initialize({{
+ startOnLoad: false,
+ theme: "default",
+ themeVariables: {{
+ background: themeBackground,
+ fontFamily: "Calibri, Segoe UI, Arial, sans-serif",
+ fontSize: "10pt",
+ }},
+ themeCSS: ".slice {{ font-size: 10pt !important; }}\\n.legend text {{ font-size: 10pt !important; }}\\n.pieTitleText {{ font-size: 10pt !important; }}",
+ fontFamily: "Calibri, Segoe UI, Arial, sans-serif",
+ securityLevel: "strict",
+ flowchart: {{ htmlLabels: false }},
+ }});
+ }} catch (_e) {{
+ // Ignore and proceed with defaults.
+ }}
+
+ const bytes = decodeBase64ToUint8Array(base64Data);
+ const zip = new window.JSZip();
+ await zip.loadAsync(bytes);
+
+ const docXml = await zip.file("word/document.xml").async("string");
+ const relsXml = await zip.file("word/_rels/document.xml.rels").async("string");
+ const parser = new DOMParser();
+ const xmlDoc = parser.parseFromString(docXml, "application/xml");
+ const relsDoc = parser.parseFromString(relsXml, "application/xml");
+
+ // Build rId -> target path mapping
+ const rels = relsDoc.getElementsByTagName("Relationship");
+ const rIdToTarget = {{}};
+ for (let i = 0; i < rels.length; i++) {{
+ const rel = rels[i];
+ const id = rel.getAttribute("Id");
+ const target = rel.getAttribute("Target");
+ if (id && target) rIdToTarget[id] = target;
+ }}
+
+ const maxWidthEmu = getMaxWidthEmu(xmlDoc);
+ const maxWidthEmuScaled = Math.max(1, Math.round(maxWidthEmu * Math.min(1.0, Math.max(0.1, displayScale || 1.0))));
+
+ const drawings = xmlDoc.getElementsByTagName("w:drawing");
+ const placeholders = [];
+
+ for (let i = 0; i < drawings.length; i++) {{
+ const drawing = drawings[i];
+ const docPr = getChildByTag(drawing, "wp:docPr");
+ if (!docPr) continue;
+ const descr = docPr.getAttribute("descr") || "";
+ if (!descr.startsWith("MERMAID_SRC:")) continue;
+ const encoded = descr.substring("MERMAID_SRC:".length);
+ const code = decodeURIComponent(encoded);
+
+ const blip = getChildByTag(drawing, "a:blip");
+ const ridPng = blip ? blip.getAttribute("r:embed") : null;
+ const svgBlip = getChildByTag(drawing, "asvg:svgBlip");
+ const ridSvg = svgBlip ? svgBlip.getAttribute("r:embed") : null;
+
+ const container = getChildByTag(drawing, "wp:inline") || getChildByTag(drawing, "wp:anchor");
+ const extent = container ? getChildByTag(container, "wp:extent") : null;
+
+ const xfrm = getChildByTag(drawing, "a:xfrm");
+ const xfrmExt = xfrm ? getChildByTag(xfrm, "a:ext") : null;
+
+ placeholders.push({{ code, ridPng, ridSvg, extent, xfrmExt, svgBlip }});
+ }}
+
+ if (!placeholders.length) {{
+ const blob = new Blob([bytes], {{ type: "application/vnd.openxmlformats-officedocument.wordprocessingml.document" }});
+ downloadBlob(blob, filename);
+ return;
+ }}
+
+ // Phase 1: Render all Mermaid diagrams sequentially (mermaid needs DOM)
+ const renderResults = [];
+ for (let i = 0; i < placeholders.length; i++) {{
+ const item = placeholders[i];
+ try {{
+ const id = "owui-mermaid-" + i;
+ const rendered = await window.mermaid.render(id, item.code);
+ let svgText = rendered && rendered.svg ? rendered.svg : rendered;
+ if (!svgText || typeof svgText !== "string") throw new Error("Mermaid returned empty SVG");
+
+ svgText = normalizeSvgForWord(svgText);
+ const hasForeignObject = /<foreignObject\\b/i.test(svgText);
+ if (hasForeignObject && item.svgBlip) {{
+ try {{ item.svgBlip.parentNode && item.svgBlip.parentNode.removeChild(item.svgBlip); }} catch (_e) {{}}
+ item.ridSvg = null;
+ }}
+
+ const svgDoc = new DOMParser().parseFromString(svgText, "image/svg+xml");
+ const svgEl = svgDoc.documentElement;
+ const vb = parseViewBox(svgEl && svgEl.getAttribute ? svgEl.getAttribute("viewBox") : null);
+ const ratio = vb && vb.width > 0 && vb.height > 0 ? (vb.width / vb.height) : (4/3);
+
+ const widthEmu = maxWidthEmuScaled;
+ const heightEmu = Math.max(1, Math.round(widthEmu / ratio));
+
+ renderResults.push({{ item, svgText, widthEmu, heightEmu, success: true }});
+ }} catch (err) {{
+ console.error("Mermaid render failed for block", i, err);
+ renderResults.push({{ item, svgText: null, widthEmu: 0, heightEmu: 0, success: false }});
+ }}
+ }}
+
+ // Phase 2: Convert SVG to PNG in parallel for performance
+ async function svgToPng(svgText, targetWidthPx, targetHeightPx) {{
+ const canvas = document.createElement("canvas");
+ const ctx = canvas.getContext("2d");
+ const scale = Math.max(1.0, pngScale || 1.0);
+ canvas.width = Math.round(targetWidthPx * scale);
+ canvas.height = Math.round(targetHeightPx * scale);
+ ctx.setTransform(1, 0, 0, 1, 0, 0);
+ if (bgFill) {{
+ ctx.fillStyle = bgFill;
+ ctx.fillRect(0, 0, canvas.width, canvas.height);
+ }}
+ ctx.scale(scale, scale);
+
+ const img = new Image();
+ await new Promise((resolve, reject) => {{
+ img.onload = resolve;
+ img.onerror = reject;
+ img.src = "data:image/svg+xml;base64," + btoa(unescape(encodeURIComponent(svgText)));
+ }});
+
+ ctx.drawImage(img, 0, 0, targetWidthPx, targetHeightPx);
+ const pngDataUrl = canvas.toDataURL("image/png");
+ return pngDataUrl.split(",")[1];
+ }}
+
+ // Create PNG conversion promises for parallel execution
+ const pngPromises = renderResults.map(async (result, i) => {{
+ if (!result.success || !result.svgText) return null;
+ const {{ item, widthEmu, heightEmu }} = result;
+ if (!item.ridPng || !rIdToTarget[item.ridPng]) return null;
+
+ const targetWidthPx = Math.max(1, Math.round(widthEmu / 9525));
+ const targetHeightPx = Math.max(1, Math.round(heightEmu / 9525));
+
+ try {{
+ const pngBase64 = await svgToPng(result.svgText, targetWidthPx, targetHeightPx);
+ return {{ index: i, pngBase64, path: "word/" + rIdToTarget[item.ridPng] }};
+ }} catch (err) {{
+ console.error("PNG conversion failed for block", i, err);
+ return null;
+ }}
+ }});
+
+ // Wait for all PNG conversions to complete
+ const pngResults = await Promise.all(pngPromises);
+
+ // Phase 3: Update ZIP with all results
+ for (let i = 0; i < renderResults.length; i++) {{
+ const result = renderResults[i];
+ if (!result.success) continue;
+
+ const {{ item, svgText, widthEmu, heightEmu }} = result;
+
+ // Update extent in XML
+ if (item.extent) {{
+ item.extent.setAttribute("cx", `${{widthEmu}}`);
+ item.extent.setAttribute("cy", `${{heightEmu}}`);
+ }}
+ if (item.xfrmExt) {{
+ item.xfrmExt.setAttribute("cx", `${{widthEmu}}`);
+ item.xfrmExt.setAttribute("cy", `${{heightEmu}}`);
+ }}
+
+ // Write SVG part
+ if (item.ridSvg && rIdToTarget[item.ridSvg]) {{
+ zip.file("word/" + rIdToTarget[item.ridSvg], svgText);
+ }}
+ }}
+
+ // Write PNG files from parallel results
+ for (const pngResult of pngResults) {{
+ if (pngResult && pngResult.pngBase64) {{
+ zip.file(pngResult.path, pngResult.pngBase64, {{ base64: true }});
+ }}
+ }}
+
+ const newDocXml = new XMLSerializer().serializeToString(xmlDoc);
+ zip.file("word/document.xml", newDocXml);
+
+ const finalBlob = await zip.generateAsync({{
+ type: "blob",
+ compression: "DEFLATE",
+ compressionOptions: {{ level: 6 }},
+ }});
+ downloadBlob(finalBlob, filename);
}} catch (error) {{
- console.error('Export failed:', error);
- alert('导出失败: ' + error.message);
+ console.error("Export pipeline failed:", error);
+ const bytes = decodeBase64ToUint8Array(base64Data);
+ const blob = new Blob([bytes], {{ type: "application/vnd.openxmlformats-officedocument.wordprocessingml.document" }});
+ downloadBlob(blob, filename);
}}
}})();
"""
@@ -476,29 +835,36 @@ class Action:
await __event_emitter__(
{
"type": "status",
- "data": {"description": "Word 文档已导出", "done": True},
+ "data": {
+ "description": self._get_msg("exported"),
+ "done": True,
+ },
}
)
await self._send_notification(
- __event_emitter__, "success", f"已成功导出为 {filename}"
+ __event_emitter__,
+ "success",
+ self._get_msg("success", filename=filename),
)
- return {"message": "下载事件已触发"}
+ return {"message": "Download triggered"}
except Exception as e:
- print(f"Error exporting to Word: {str(e)}")
+ logger.exception(f"Error exporting to Word: {str(e)}")
await __event_emitter__(
{
"type": "status",
"data": {
- "description": f"导出失败: {str(e)}",
+ "description": self._get_msg("export_failed", error=str(e)),
"done": True,
},
}
)
await self._send_notification(
- __event_emitter__, "error", f"导出 Word 文档时出错: {str(e)}"
+ __event_emitter__,
+ "error",
+ self._get_msg("error_export", error=str(e)),
)
async def generate_title_using_ai(
@@ -532,17 +898,17 @@ class Action:
return ""
def extract_title(self, content: str) -> str:
- """从 Markdown 内容提取一级/二级标题"""
+ """Extract title from Markdown h1/h2 only"""
lines = content.split("\n")
for line in lines:
- # 仅匹配 h1-h2 标题
+ # Match h1-h2 headings only
match = re.match(r"^#{1,2}\s+(.+)$", line.strip())
if match:
return match.group(1).strip()
return ""
def extract_chat_title(self, body: dict) -> str:
- """从请求体中提取会话标题"""
+ """Extract chat title from common payload fields."""
if not isinstance(body, dict):
return ""
@@ -563,7 +929,7 @@ class Action:
return ""
def extract_chat_id(self, body: dict, metadata: Optional[dict]) -> str:
- """从 body 或 metadata 中提取 chat_id"""
+ """Extract chat_id from body or metadata"""
if isinstance(body, dict):
chat_id = body.get("chat_id") or body.get("id")
if isinstance(chat_id, str) and chat_id.strip():
@@ -582,19 +948,21 @@ class Action:
return ""
async def fetch_chat_title(self, chat_id: str, user_id: str = "") -> str:
- """根据 chat_id 从数据库获取标题"""
+ """Fetch chat title from database by chat_id"""
if not chat_id:
return ""
def _load_chat():
if user_id:
- return Chats.get_chat_by_id_and_user_id(id=chat_id, user_id=user_id)
+ chat = Chats.get_chat_by_id_and_user_id(id=chat_id, user_id=user_id)
+ if chat:
+ return chat
return Chats.get_chat_by_id(chat_id)
try:
chat = await asyncio.to_thread(_load_chat)
except Exception as exc:
- logger.warning(f"加载聊天 {chat_id} 失败: {exc}")
+ logger.warning(f"Failed to load chat {chat_id}: {exc}")
return ""
if not chat:
@@ -605,12 +973,488 @@ class Action:
return title.strip() if isinstance(title, str) else ""
def clean_filename(self, name: str) -> str:
- """清理文件名中的非法字符"""
- return re.sub(r'[\\/*?:"<>|]', "", name).strip()[:50]
+ """Clean illegal characters from filename and strip emoji."""
+ if not isinstance(name, str):
+ return ""
+
+ def _is_emoji_codepoint(codepoint: int) -> bool:
+ # Common emoji ranges + flag regional indicators.
+ return (
+ 0x1F000 <= codepoint <= 0x1FAFF
+ or 0x1F1E6 <= codepoint <= 0x1F1FF
+ or 0x2600 <= codepoint <= 0x26FF
+ or 0x2700 <= codepoint <= 0x27BF
+ or 0x2300 <= codepoint <= 0x23FF
+ or 0x2B00 <= codepoint <= 0x2BFF
+ )
+
+ def _is_emoji_modifier(codepoint: int) -> bool:
+ # VS15/VS16, ZWJ, keycap, skin tones, and tag characters used in some emoji sequences.
+ return (
+ codepoint in (0x200D, 0xFE0E, 0xFE0F, 0x20E3)
+ or 0x1F3FB <= codepoint <= 0x1F3FF
+ or 0xE0020 <= codepoint <= 0xE007F
+ )
+
+ without_emoji = "".join(
+ ch
+ for ch in name
+ if not (_is_emoji_codepoint(ord(ch)) or _is_emoji_modifier(ord(ch)))
+ )
+ cleaned = re.sub(r'[\\/*?:"<>|]', "", without_emoji)
+ cleaned = re.sub(r"\s+", " ", cleaned).strip().strip(".")
+ return cleaned[:50].strip()
+
+ def _max_embed_image_bytes(self) -> int:
+ mb = getattr(self.valves, "MAX_EMBED_IMAGE_MB", 20)
+ try:
+ mb_i = int(mb)
+ except Exception:
+ mb_i = 20
+ mb_i = max(1, mb_i)
+ return mb_i * 1024 * 1024
+
+ def _extract_owui_api_file_id(self, url: str) -> Optional[str]:
+ if not isinstance(url, str) or not url:
+ return None
+ m = _OWUI_API_FILE_ID_RE.search(url)
+ if not m:
+ return None
+ fid = (m.group("id") or "").strip()
+ return fid or None
+
+ def _read_file_bytes_limited(self, path: Path, max_bytes: int) -> Optional[bytes]:
+ try:
+ if not path.exists():
+ return None
+ try:
+ size = path.stat().st_size
+ if size > max_bytes:
+ return None
+ except Exception:
+ pass
+ with path.open("rb") as f:
+ data = f.read(max_bytes + 1)
+ if len(data) > max_bytes:
+ return None
+ return data
+ except Exception:
+ return None
+
+ def _decode_base64_limited(self, b64: str, max_bytes: int) -> Optional[bytes]:
+ if not isinstance(b64, str):
+ return None
+ s = re.sub(r"\s+", "", b64.strip())
+ if not s:
+ return None
+
+ # Rough pre-check: base64 expands by ~4/3. Avoid decoding clearly oversized payloads.
+ est = (len(s) * 3) // 4
+ if est > max_bytes:
+ return None
+
+ pad = (-len(s)) % 4
+ if pad:
+ s = s + ("=" * pad)
+ try:
+ out = base64.b64decode(s, validate=False)
+ except (binascii.Error, ValueError):
+ return None
+ if len(out) > max_bytes:
+ return None
+ return out
+
+ def _image_bytes_from_data_url(self, url: str, max_bytes: int) -> Optional[bytes]:
+ if not isinstance(url, str):
+ return None
+ m = _DATA_IMAGE_URL_RE.match(url.strip())
+ if not m:
+ return None
+ b64 = m.group("b64") or ""
+ return self._decode_base64_limited(b64, max_bytes)
+
+ def _image_bytes_from_owui_file_id(
+ self, file_id: str, max_bytes: int
+ ) -> Optional[bytes]:
+ if not file_id or Files is None:
+ return None
+ try:
+ file_obj = Files.get_file_by_id(file_id)
+ except Exception:
+ return None
+ if not file_obj:
+ return None
+
+ # Common patterns across Open WebUI versions / storage backends.
+ data_field = getattr(file_obj, "data", None)
+ if isinstance(data_field, dict):
+ blob_value = data_field.get("bytes")
+ if isinstance(blob_value, (bytes, bytearray)):
+ raw = bytes(blob_value)
+ return raw if len(raw) <= max_bytes else None
+ for key in ("b64", "base64", "data"):
+ inline = data_field.get(key)
+ if isinstance(inline, str) and inline.strip():
+ return self._decode_base64_limited(inline, max_bytes)
+
+ for attr in ("path", "file_path", "absolute_path"):
+ candidate = getattr(file_obj, attr, None)
+ if isinstance(candidate, str) and candidate.strip():
+ raw = self._read_file_bytes_limited(Path(candidate), max_bytes)
+ if raw is not None:
+ return raw
+
+ for attr in ("content", "blob", "data"):
+ raw = getattr(file_obj, attr, None)
+ if isinstance(raw, (bytes, bytearray)):
+ b = bytes(raw)
+ return b if len(b) <= max_bytes else None
+
+ return None
+
+ def _add_image_placeholder(self, paragraph, alt: str, reason: str):
+ label = (alt or "").strip() or "image"
+ msg = f"[{label} not embedded: {reason}]"
+ self._add_text_run(paragraph, msg, bold=False, italic=False, strike=False)
+
+ def _try_embed_image(
+ self, paragraph, image_bytes: bytes
+ ) -> Tuple[bool, Optional[str]]:
+ if not image_bytes:
+ return False, "empty image bytes"
+ try:
+ run = paragraph.add_run()
+ width = None
+ if self._active_doc is not None:
+ try:
+ width = self._available_block_width(self._active_doc)
+ except Exception:
+ width = None
+ run.add_picture(cast(Any, io.BytesIO(image_bytes)), width=width)
+ return True, None
+ except Exception as e:
+ return False, str(e)
+
+ def _embed_markdown_image(self, paragraph, alt: str, url: str):
+ max_bytes = self._max_embed_image_bytes()
+ u = (url or "").strip()
+ if not u:
+ self._add_image_placeholder(paragraph, alt, "missing URL")
+ return
+
+ image_bytes: Optional[bytes] = None
+ if u.lower().startswith("data:"):
+ image_bytes = self._image_bytes_from_data_url(u, max_bytes)
+ if image_bytes is None:
+ self._add_image_placeholder(
+ paragraph,
+ alt,
+ f"invalid data URL or exceeds {self.valves.MAX_EMBED_IMAGE_MB}MB",
+ )
+ return
+ else:
+ file_id = self._extract_owui_api_file_id(u)
+ if not file_id:
+ # External images are not fetched; treat as non-embeddable.
+ self._add_image_placeholder(paragraph, alt, "external URL")
+ return
+ image_bytes = self._image_bytes_from_owui_file_id(file_id, max_bytes)
+ if image_bytes is None:
+ self._add_image_placeholder(
+ paragraph, alt, f"file unavailable ({file_id})"
+ )
+ return
+
+ success, error_msg = self._try_embed_image(paragraph, image_bytes)
+ if not success:
+ self._add_image_placeholder(
+ paragraph, alt, f"unsupported image type: {error_msg}"
+ )
+
+ async def markdown_to_docx(
+ self,
+ markdown_text: str,
+ top_heading: str = "",
+ has_h1: bool = False,
+ sources: Optional[List[dict]] = None,
+ event_emitter: Optional[Callable] = None,
+ ) -> Document:
+ """
+ Convert Markdown text to Word document
+ Supports: headings, paragraphs, bold, italic, code blocks, lists, tables, links
+ Additionally: Mermaid fenced blocks (```mermaid) rendered client-side via Mermaid.js (SVG+PNG),
+ LaTeX math to Word equations, and OpenWebUI citations to References.
+ """
+ doc = Document()
+ self._active_doc = doc
+ try:
+ self._mermaid_figure_counter = 0
+ self._mermaid_placeholder_counter = 0
+ self._caption_style_name = None
+ self._citation_anchor_by_index = {}
+ self._citation_refs = self._build_citation_refs(sources or [])
+ self._bookmark_id_counter = 1
+ for ref in self._citation_refs:
+ self._citation_anchor_by_index[ref.idx] = ref.anchor
+
+ # Set default fonts
+ self.set_document_default_font(doc)
+
+ # If there is no h1 in content, prepend chat title as h1 when provided
+ if top_heading and not has_h1:
+ self.add_heading(doc, top_heading, 1)
+
+ lines = markdown_text.split("\n")
+ i = 0
+ in_code_block = False
+ code_block_content = []
+ code_block_info_raw = ""
+ code_block_lang = ""
+ code_block_attrs: List[str] = []
+ in_math_block = False
+ math_block_delim = ""
+ math_block_lines: List[str] = []
+ in_list = False
+ list_items = []
+ list_type = None # 'ordered' or 'unordered'
+
+ total_lines = len(lines)
+ last_update_time = time.time()
+
+ while i < len(lines):
+ # Update status every 2 seconds
+ if event_emitter and time.time() - last_update_time > 2.0:
+ progress = int((i / total_lines) * 100)
+ await event_emitter(
+ {
+ "type": "status",
+ "data": {
+ "description": f"{self._get_msg('converting')} ({progress}%)",
+ "done": False,
+ },
+ }
+ )
+ last_update_time = time.time()
+
+ line = lines[i]
+
+ # Handle display math blocks (\[...\] or $$...$$)
+ if not in_code_block and self.valves.MATH_ENABLE:
+ single_line = self._extract_single_line_math(line)
+ if single_line is not None:
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+ self._add_display_equation(doc, single_line)
+ i += 1
+ continue
+
+ if not in_math_block:
+ stripped = line.strip()
+ if stripped in (r"\[", "$$"):
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+ in_math_block = True
+ math_block_delim = stripped
+ math_block_lines = []
+ i += 1
+ continue
+ else:
+ stripped = line.strip()
+ close = r"\]" if math_block_delim == r"\[" else "$$"
+ if stripped == close:
+ in_math_block = False
+ latex = "\n".join(math_block_lines).strip()
+ self._add_display_equation(doc, latex)
+ math_block_delim = ""
+ math_block_lines = []
+ i += 1
+ continue
+ math_block_lines.append(line)
+ i += 1
+ continue
+
+ # Handle code blocks
+ if line.strip().startswith("```"):
+ if not in_code_block:
+ # Process pending list first
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ in_code_block = True
+ code_block_info_raw = line.strip()[3:].strip()
+ code_block_lang, code_block_attrs = self._parse_fence_info(
+ code_block_info_raw
+ )
+ code_block_content = []
+ else:
+ # End code block
+ in_code_block = False
+ code_text = "\n".join(code_block_content)
+ if code_block_lang.lower() == "mermaid":
+ self._insert_mermaid_placeholder(doc, code_text)
+ else:
+ self.add_code_block(doc, code_text, code_block_lang)
+ code_block_content = []
+ code_block_info_raw = ""
+ code_block_lang = ""
+ code_block_attrs = []
+ i += 1
+ continue
+
+ if in_code_block:
+ code_block_content.append(line)
+ i += 1
+ continue
+
+ # Handle tables
+ if line.strip().startswith("|") and line.strip().endswith("|"):
+ # Process pending list first
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ table_lines = []
+ while i < len(lines) and lines[i].strip().startswith("|"):
+ table_lines.append(lines[i])
+ i += 1
+ self.add_table(doc, table_lines)
+ continue
+
+ # Handle headings
+ header_match = re.match(r"^(#{1,6})\s+(.+)$", line.strip())
+ if header_match:
+ # Process pending list first
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ level = len(header_match.group(1))
+ text = header_match.group(2)
+ self.add_heading(doc, text, level)
+ i += 1
+ continue
+
+ # Handle unordered lists
+ unordered_match = re.match(r"^(\s*)[-*+]\s+(.+)$", line)
+ if unordered_match:
+ if not in_list or list_type != "unordered":
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = True
+ list_type = "unordered"
+ indent = len(unordered_match.group(1)) // 2
+ list_items.append((indent, unordered_match.group(2)))
+ i += 1
+ continue
+
+ # Handle ordered lists
+ ordered_match = re.match(r"^(\s*)\d+[.)]\s+(.+)$", line)
+ if ordered_match:
+ if not in_list or list_type != "ordered":
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = True
+ list_type = "ordered"
+ indent = len(ordered_match.group(1)) // 2
+ list_items.append((indent, ordered_match.group(2)))
+ i += 1
+ continue
+
+ # Handle blockquotes
+ if line.strip().startswith(">"):
+ # Process pending list first
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ # Collect consecutive quote lines
+ blockquote_lines = []
+ while i < len(lines) and lines[i].strip().startswith(">"):
+ # Remove leading > and optional space
+ quote_line = re.sub(r"^>\s?", "", lines[i])
+ blockquote_lines.append(quote_line)
+ i += 1
+ self.add_blockquote(doc, "\n".join(blockquote_lines))
+ continue
+
+ # Handle horizontal rules
+ if re.match(r"^[-*_]{3,}$", line.strip()):
+ # Process pending list first
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ self.add_horizontal_rule(doc)
+ i += 1
+ continue
+
+ # Handle empty lines
+ if not line.strip():
+ # End list
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+ i += 1
+ continue
+
+ # Handle normal paragraphs
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+ list_items = []
+ in_list = False
+
+ self.add_paragraph(doc, line)
+ i += 1
+
+ # Process remaining list
+ if in_list and list_items:
+ self.add_list_to_doc(doc, list_items, list_type)
+
+ # If math block wasn't closed, render it as plain text for robustness.
+ if in_math_block and math_block_lines:
+ self.add_paragraph(doc, r"\[")
+ for l in math_block_lines:
+ self.add_paragraph(doc, l)
+ self.add_paragraph(doc, r"\]")
+
+ if self._citation_refs:
+ self._add_references_section(doc)
+
+ return doc
+ finally:
+ self._active_doc = None
+
+ def _extract_single_line_math(self, line: str) -> Optional[str]:
+ s = line.strip()
+ # \[ ... \]
+ m = re.match(r"^\\\[(.*)\\\]$", s)
+ if m:
+ return m.group(1).strip()
+ # $$ ... $$
+ m = re.match(r"^\$\$(.*)\$\$$", s)
+ if m:
+ return m.group(1).strip()
+ return None
def _strip_reasoning_blocks(self, text: str) -> str:
"""
Strip model reasoning blocks from assistant Markdown before export.
+
+ OpenWebUI can include reasoning as interleaved <details type=\"reasoning\">...</details>
+ (and sometimes <think>/<analysis> blocks). These should never be exported into DOCX.
"""
if not text:
return text
@@ -654,49 +1498,7 @@ class Action:
xml = f'<m:oMathPara xmlns:m="{m_ns}" xmlns:w="{w_ns}">{omml}</m:oMathPara>'
return parse_xml(xml)
- def _add_inline_equation(
- self,
- paragraph,
- latex: str,
- bold: bool = False,
- italic: bool = False,
- strike: bool = False,
- ):
- latex = (latex or "").strip()
- if not latex:
- return
-
- if not LATEX_MATH_AVAILABLE:
- self._add_text_run(
- paragraph, f"\\({latex}\\)", bold=bold, italic=italic, strike=strike
- )
- return
-
- try:
- mathml = latex_to_mathml(latex)
- omml = mathml2omml.convert(mathml)
- o_math = self._omml_oMath_element(omml)
- run = paragraph.add_run()
- run.bold = bold
- run.italic = italic
- run.font.strike = strike
- cast(Any, run)._r.append(o_math)
- except Exception as exc:
- logger.warning(f"Inline math conversion failed; keeping literal: {exc}")
- self._add_text_run(
- paragraph, f"\\({latex}\\)", bold=bold, italic=italic, strike=strike
- )
-
- def _omml_oMath_element(self, omml: str):
- # Ensure the OMML element declares the math namespace so parse_xml works.
- m_ns = "http://schemas.openxmlformats.org/officeDocument/2006/math"
- s = (omml or "").strip()
- if s.startswith("<m:oMath>") and s.endswith("</m:oMath>"):
- inner = s[len("<m:oMath>") : -len("</m:oMath>")]
- s = f'<m:oMath xmlns:m="{m_ns}">{inner}</m:oMath>'
- elif s.startswith("<m:oMath") and "xmlns:m=" not in s.split(">", 1)[0]:
- s = s.replace("<m:oMath", f'<m:oMath xmlns:m="{m_ns}"', 1)
- return parse_xml(s)
+ # (Math warning paragraphs removed)
def _build_citation_refs(self, sources: List[dict]) -> List[_CitationRef]:
citation_idx_map: Dict[str, int] = {}
@@ -804,7 +1606,7 @@ class Action:
cast(Any, paragraph)._p.append(hyperlink)
def _add_references_section(self, doc: Document):
- self.add_heading(doc, "参考资料", 2)
+ self.add_heading(doc, self._get_msg("references"), 2)
for ref in self._citation_refs:
para = doc.add_paragraph(style="List Number")
@@ -817,6 +1619,760 @@ class Action:
para, ref.title, bold=False, italic=False, strike=False
)
+ def _parse_fence_info(self, info_raw: str) -> Tuple[str, List[str]]:
+ parts = [p for p in (info_raw or "").split() if p.strip()]
+ if not parts:
+ return "", []
+ return parts[0], parts[1:]
+
+ def _normalize_mermaid_text(self, source: str) -> str:
+ text = (source or "").replace("\r\n", "\n").replace("\r", "\n").strip()
+ return text + "\n"
+
+ def _prepare_mermaid_for_js(self, source: str) -> str:
+ """
+ Prepare Mermaid source for client-side rendering:
+ - strip title directives (caption already carries it),
+ """
+ text = self._strip_mermaid_title_for_render(source)
+ return text
+
+ def _png_with_text_chunk(self, png_bytes: bytes, keyword: str, value: str) -> bytes:
+ """
+ Ensure placeholder PNGs stay distinct in the DOCX package:
+ python-docx may deduplicate identical image bytes into one media part.
+ We insert a small tEXt chunk so each placeholder is unique, without changing
+ dimensions or requiring external imaging libraries.
+ """
+ if not png_bytes.startswith(b"\x89PNG\r\n\x1a\n"):
+ return png_bytes
+
+ keyword_b = (keyword or "owui").encode("latin-1", errors="ignore")[:79]
+ keyword_b = keyword_b.replace(b"\x00", b"") or b"owui"
+ value_b = (value or "").encode("latin-1", errors="ignore")
+ data = keyword_b + b"\x00" + value_b
+ chunk_type = b"tEXt"
+ crc = zlib.crc32(chunk_type + data) & 0xFFFFFFFF
+ chunk = (
+ struct.pack("!I", len(data)) + chunk_type + data + struct.pack("!I", crc)
+ )
+
+ out = bytearray()
+ out.extend(png_bytes[:8])
+ offset = 8
+ inserted = False
+ while offset + 8 <= len(png_bytes):
+ length = struct.unpack("!I", png_bytes[offset : offset + 4])[0]
+ ctype = png_bytes[offset + 4 : offset + 8]
+ chunk_total = 12 + length
+ if offset + chunk_total > len(png_bytes):
+ break
+ if ctype == b"IEND" and not inserted:
+ out.extend(chunk)
+ inserted = True
+ out.extend(png_bytes[offset : offset + chunk_total])
+ offset += chunk_total
+ if ctype == b"IEND":
+ break
+ if not inserted:
+ return png_bytes
+ return bytes(out)
+
+ def _make_mermaid_placeholder_png(self, seed: str) -> bytes:
+ return self._png_with_text_chunk(_TRANSPARENT_1PX_PNG, "owui", seed)
+
+ def _dummy_mermaid_svg_bytes(self) -> bytes:
+ return (
+ '<svg xmlns="http://www.w3.org/2000/svg" width="1" height="1" viewBox="0 0 1 1"></svg>'
+ ).encode("utf-8")
+
+ def _insert_mermaid_placeholder(self, doc: Document, mermaid_source: str):
+ caption_title: Optional[str] = (
+ self._extract_mermaid_title(mermaid_source)
+ if self.valves.MERMAID_CAPTIONS_ENABLE
+ else None
+ )
+
+ source_for_render = mermaid_source
+ if self.valves.MERMAID_OPTIMIZE_LAYOUT:
+ source_for_render = re.sub(
+ r"^(graph|flowchart)\s+LR\b",
+ r"\1 TD",
+ source_for_render,
+ flags=re.MULTILINE | re.IGNORECASE,
+ )
+ source_for_render = self._prepare_mermaid_for_js(source_for_render)
+
+ self._mermaid_placeholder_counter += 1
+ seed = hashlib.sha256(
+ f"{self._mermaid_placeholder_counter}\n{source_for_render}".encode(
+ "utf-8", errors="replace"
+ )
+ ).hexdigest()[:16]
+ png_bytes = self._make_mermaid_placeholder_png(seed)
+
+ try:
+ shape = doc.add_picture(cast(Any, io.BytesIO(png_bytes)))
+ except Exception as e:
+ logger.warning(f"Failed to add Mermaid placeholder image: {e}")
+ self.add_paragraph(doc, f"[Mermaid placeholder failed: {e}]")
+ return
+ try:
+ doc.paragraphs[-1].alignment = WD_ALIGN_PARAGRAPH.CENTER
+ except Exception:
+ pass
+
+ # Attach a dummy SVG part so we can later overwrite it client-side (SVG+PNG).
+ self._attach_svg_blip(doc, shape, self._dummy_mermaid_svg_bytes())
+
+ try:
+ encoded = quote(source_for_render)
+ inline = shape._inline
+ docPr = inline.docPr
+ docPr.set("descr", f"MERMAID_SRC:{encoded}")
+ docPr.set("title", "Mermaid Diagram Placeholder")
+ except Exception as exc:
+ logger.warning(f"Failed to annotate Mermaid placeholder: {exc}")
+
+ self._add_mermaid_caption(doc, caption_title)
+
+ def _extract_mermaid_title(self, source: str) -> Optional[str]:
+ lines = self._normalize_mermaid_text(source).split("\n")
+ header_found = False
+ for raw in lines:
+ line = raw.strip()
+ if not line:
+ continue
+ if line.startswith("%%{") and line.endswith("}%%"):
+ continue
+ if line.startswith("%%"):
+ continue
+ # diagram header line
+ if not header_found:
+ header_found = True
+ # Mermaid beta/diagram headers can embed a title on the header line, e.g.:
+ # - radar-beta title Foo
+ # - xychart-beta title: "Foo"
+ mt = re.match(
+ r"^(?P<header>\S.*?)(?:\s+title\s*:?\s+)(?P<title>.+)$",
+ line,
+ re.IGNORECASE,
+ )
+ if mt:
+ title = (mt.group("title") or "").strip().strip('"').strip("'")
+ if title:
+ return title
+ continue
+
+ # title "Foo" / title Foo
+ m = re.match(r'^title\s*:?\s+"(.+)"\s*$', line, re.IGNORECASE)
+ if m:
+ return m.group(1).strip()
+ m = re.match(r"^title\s*:?\s+(.+)$", line, re.IGNORECASE)
+ if m:
+ return m.group(1).strip().strip('"').strip("'")
+ return None
+
+ def _strip_mermaid_title_for_render(self, source: str) -> str:
+ """
+ Removes Mermaid title directives from the source before rendering.
+ Captions already carry the title.
+ """
+ lines = self._normalize_mermaid_text(source).split("\n")
+ out: List[str] = []
+ header_found = False
+ title_stripped = False
+ meaningful_after_header = False
+
+ for raw in lines:
+ line = raw.rstrip("\n")
+ stripped = line.strip()
+
+ if not stripped:
+ out.append(line)
+ continue
+
+ if stripped.startswith("%%{") and stripped.endswith("}%%"):
+ out.append(line)
+ continue
+ if stripped.startswith("%%"):
+ out.append(line)
+ continue
+
+ if not header_found:
+ header_found = True
+ # Some Mermaid diagram headers can embed a title on the header line, e.g.:
+ # - radar-beta title Foo
+ # - xychart-beta title: "Foo"
+ mt = re.match(
+ r"^(?P<header>\S.*?)(?:\s+title\s*:?\s+)(?P<title>.+)$",
+ stripped,
+ re.IGNORECASE,
+ )
+ if mt:
+ cleaned = (mt.group("header") or "").strip()
+ out.append(cleaned if cleaned else stripped)
+ title_stripped = True
+ continue
+ out.append(line)
+ continue
+
+ if not title_stripped and not meaningful_after_header:
+ # Strip a standalone title directive line early in the diagram.
+ if re.match(r'^title\s*:?\s+(".+"|.+)$', stripped, re.IGNORECASE):
+ title_stripped = True
+ continue
+
+ # Consider this a meaningful content line after header.
+ meaningful_after_header = True
+ out.append(line)
+
+ return "\n".join(out).strip() + "\n"
+
+ def _ensure_caption_style(self, doc: Document) -> str:
+ if self._caption_style_name is not None:
+ return self._caption_style_name
+
+ style_name = (self.valves.MERMAID_CAPTION_STYLE or "").strip()
+ if style_name == "":
+ # Empty means: do not apply a caption style.
+ self._caption_style_name = ""
+ return ""
+
+ # Prefer existing style if present.
+ try:
+ _ = doc.styles[style_name]
+ self._caption_style_name = style_name
+ return style_name
+ except KeyError:
+ pass
+
+ # If user requested "Caption" but it's missing, create a safe custom style name.
+ if style_name.lower() == "caption":
+ style_name = "OWUI Caption"
+
+ try:
+ _ = doc.styles[style_name]
+ self._caption_style_name = style_name
+ return style_name
+ except KeyError:
+ pass
+
+ try:
+ style = doc.styles.add_style(style_name, WD_STYLE_TYPE.PARAGRAPH)
+ style.font.name = "Calibri"
+ style.font.size = Pt(9)
+ style.font.color.rgb = RGBColor(80, 80, 80)
+ style.paragraph_format.space_before = Pt(2)
+ style.paragraph_format.space_after = Pt(8)
+ self._caption_style_name = style_name
+ return style_name
+ except Exception:
+ self._caption_style_name = "Normal"
+ return "Normal"
+
+ def _add_mermaid_caption(self, doc: Document, title: Optional[str]):
+ if not self.valves.MERMAID_CAPTIONS_ENABLE:
+ return
+
+ # Use configured prefix, or auto-detect from user language
+ prefix = (self.valves.MERMAID_CAPTION_PREFIX or "").strip()
+ if prefix == "":
+ prefix = self._get_msg("figure_prefix")
+
+ if prefix == "" and not title:
+ return
+
+ self._mermaid_figure_counter += 1
+ if prefix == "":
+ caption = title or ""
+ else:
+ base = f"{prefix} {self._mermaid_figure_counter}"
+ caption = f"{base}: {title}" if title else base
+ if caption == "":
+ return
+
+ para = doc.add_paragraph()
+ style_name = self._ensure_caption_style(doc)
+ if style_name:
+ para.style = style_name
+ para.alignment = WD_ALIGN_PARAGRAPH.CENTER
+ self.add_formatted_text(para, caption)
+
+ def _available_block_width(self, doc: Document):
+ section = doc.sections[0]
+ return section.page_width - section.left_margin - section.right_margin
+
+ def _attach_svg_blip(self, doc: Document, inline_shape: Any, svg_bytes: bytes):
+ if not svg_bytes:
+ return
+
+ try:
+ pkg = doc.part.package
+ partname = pkg.next_partname("/word/media/image%d.svg")
+ from docx.opc.part import Part
+
+ svg_part = Part(partname, "image/svg+xml", svg_bytes)
+ rid_svg = doc.part.relate_to(svg_part, RT.IMAGE)
+
+ inline = inline_shape._inline
+ blips = inline.xpath(".//a:blip")
+ if not blips:
+ return
+ blip = blips[0]
+
+ existing = blip.xpath(".//asvg:svgBlip")
+ if existing:
+ existing[0].set(qn("r:embed"), rid_svg)
+ return
+
+ extLst = OxmlElement("a:extLst")
+ ext = OxmlElement("a:ext")
+ ext.set("uri", "{96DAC541-7B7A-43D3-8B79-37D633B846F1}")
+
+ svgBlip = OxmlElement("asvg:svgBlip")
+ svgBlip.set(qn("r:embed"), rid_svg)
+ ext.append(svgBlip)
+ extLst.append(ext)
+ blip.append(extLst)
+ except Exception as exc:
+ logger.warning(f"Failed to attach SVG blip; keeping PNG fallback: {exc}")
+
+ # (Mermaid warning paragraphs removed)
+
+ def set_document_default_font(self, doc: Document):
+ """Set document default font using configured fonts."""
+ style = doc.styles["Normal"]
+ font = style.font
+ font.name = self.valves.FONT_LATIN
+ font.size = Pt(11)
+ # Set Asian font
+ style._element.rPr.rFonts.set(qn("w:eastAsia"), self.valves.FONT_ASIAN)
+
+ # Set paragraph format
+ paragraph_format = style.paragraph_format
+ paragraph_format.line_spacing_rule = WD_LINE_SPACING.ONE_POINT_FIVE
+ paragraph_format.space_after = Pt(6)
+
+ def add_heading(self, doc: Document, text: str, level: int):
+ """Add heading"""
+ # Word heading levels start from 0, Markdown from 1
+ heading_level = min(level, 9) # Word supports up to Heading 9
+ heading = doc.add_heading(level=heading_level)
+
+ # Parse and add formatted text
+ self.add_formatted_text(heading, text)
+
+ def add_paragraph(self, doc: Document, text: str):
+ """Add paragraph with inline formatting support"""
+ paragraph = doc.add_paragraph()
+ self.add_formatted_text(paragraph, text)
+
+ def add_formatted_text(self, paragraph, text: str):
+ """
+ Parse Markdown inline formatting and add to paragraph.
+ Supports: bold, italic, inline code, links, strikethrough, auto-link URLs,
+ and inline LaTeX math \\(...\\) when MATH_ENABLE is on.
+ """
+ self._add_inline_segments(
+ paragraph, text or "", bold=False, italic=False, strike=False
+ )
+
+ def _add_text_run(self, paragraph, s: str, bold: bool, italic: bool, strike: bool):
+ if not s:
+ return
+ run = paragraph.add_run(s)
+ if bold:
+ run.bold = True
+ if italic:
+ run.italic = True
+ if strike:
+ run.font.strike = True
+
+ def _add_inline_code(self, paragraph, s: str):
+ if s == "":
+ return
+
+ def _add_code_run(chunk: str):
+ if not chunk:
+ return
+ run = paragraph.add_run(chunk)
+ run.font.name = self.valves.FONT_CODE
+ run._element.rPr.rFonts.set(qn("w:eastAsia"), self.valves.FONT_CODE)
+ run.font.size = Pt(10)
+ shading = OxmlElement("w:shd")
+ shading.set(qn("w:fill"), "E8E8E8")
+ run._element.rPr.append(shading)
+
+ i = 0
+ for m in _AUTO_URL_RE.finditer(s):
+ start, end = m.span()
+ if start > i:
+ _add_code_run(s[i:start])
+
+ raw = m.group(0)
+ trimmed = raw
+ while trimmed and trimmed[-1] in ".,;:!?)]}":
+ trimmed = trimmed[:-1]
+ suffix = raw[len(trimmed) :]
+
+ normalized = self._normalize_url(trimmed)
+ if normalized:
+ self._add_hyperlink_code(
+ paragraph, display_text=trimmed, url=normalized
+ )
+ else:
+ _add_code_run(raw)
+
+ if suffix:
+ _add_code_run(suffix)
+
+ i = end
+
+ if i < len(s):
+ _add_code_run(s[i:])
+
+ def _add_hyperlink_code(self, paragraph, display_text: str, url: str):
+ u = self._normalize_url(url)
+ if not u:
+ self._add_inline_code(paragraph, display_text)
+ return
+
+ part = getattr(paragraph, "part", None)
+ if part is None or not hasattr(part, "relate_to"):
+ self._add_inline_code(paragraph, display_text)
+ return
+
+ r_id = part.relate_to(u, RT.HYPERLINK, is_external=True)
+
+ hyperlink = OxmlElement("w:hyperlink")
+ hyperlink.set(qn("r:id"), r_id)
+
+ new_run = OxmlElement("w:r")
+ rPr = OxmlElement("w:rPr")
+
+ rFonts = OxmlElement("w:rFonts")
+ rFonts.set(qn("w:ascii"), self.valves.FONT_CODE)
+ rFonts.set(qn("w:hAnsi"), self.valves.FONT_CODE)
+ rFonts.set(qn("w:eastAsia"), self.valves.FONT_CODE)
+ rPr.append(rFonts)
+
+ sz = OxmlElement("w:sz")
+ sz.set(qn("w:val"), "20") # 10pt
+ rPr.append(sz)
+ sz_cs = OxmlElement("w:szCs")
+ sz_cs.set(qn("w:val"), "20")
+ rPr.append(sz_cs)
+
+ shading = OxmlElement("w:shd")
+ shading.set(qn("w:fill"), "E8E8E8")
+ rPr.append(shading)
+
+ new_run.append(rPr)
+
+ t = OxmlElement("w:t")
+ t.text = display_text
+ new_run.append(t)
+
+ hyperlink.append(new_run)
+ cast(Any, paragraph)._p.append(hyperlink)
+
+ def _add_inline_segments(
+ self, paragraph, text: str, bold: bool, italic: bool, strike: bool
+ ):
+ i = 0
+ n = len(text)
+
+ def next_special(start: int) -> int:
+ candidates = []
+ for ch in ("`", "!", "[", "*", "_", "~", "$", "\\"):
+ idx = text.find(ch, start)
+ if idx != -1:
+ candidates.append(idx)
+ idx = text.find(r"\(", start)
+ if idx != -1:
+ candidates.append(idx)
+ idx = text.find("http://", start)
+ if idx != -1:
+ candidates.append(idx)
+ idx = text.find("https://", start)
+ if idx != -1:
+ candidates.append(idx)
+ idx = text.find("www.", start)
+ if idx != -1:
+ candidates.append(idx)
+ return min(candidates) if candidates else n
+
+ while i < n:
+ # Markdown image: 
+ if text.startswith("![", i):
+ close = text.find("]", i + 2)
+ if close != -1 and close + 1 < n and text[close + 1] == "(":
+ close_paren = text.find(")", close + 2)
+ if close_paren != -1:
+ alt = text[i + 2 : close]
+ url = text[close + 2 : close_paren].strip()
+ # Allow angle-bracket wrapped URLs: 
+ if url.startswith("<") and url.endswith(">") and len(url) >= 2:
+ url = url[1:-1].strip()
+ self._embed_markdown_image(paragraph, alt=alt, url=url)
+ i = close_paren + 1
+ continue
+
+ if text[i] == "`":
+ j = text.find("`", i + 1)
+ if j != -1:
+ self._add_inline_code(paragraph, text[i + 1 : j])
+ i = j + 1
+ continue
+
+ if text.startswith(r"\(", i):
+ j = text.find(r"\)", i + 2)
+ if j != -1:
+ self._add_inline_equation(
+ paragraph,
+ text[i + 2 : j],
+ bold=bold,
+ italic=italic,
+ strike=strike,
+ )
+ i = j + 2
+ continue
+
+ # Handle backslash escapes
+ if text[i] == "\\":
+ if i + 1 < n:
+ ch = text[i + 1]
+ # Standard Markdown escapes + $ for math
+ if ch in "\\`*_{}[]()#+-.!|$":
+ self._add_text_run(paragraph, ch, bold, italic, strike)
+ i += 2
+ continue
+ # Keep other backslashes literal
+ self._add_text_run(paragraph, "\\", bold, italic, strike)
+ i += 1
+ continue
+
+ # Handle long run of underscores (fill-in-the-blank)
+ if text[i] == "_":
+ run_len = 0
+ while i + run_len < n and text[i + run_len] == "_":
+ run_len += 1
+ if run_len >= 4:
+ self._add_text_run(
+ paragraph, text[i : i + run_len], bold, italic, strike
+ )
+ i += run_len
+ continue
+
+ # Handle long run of asterisks (separator/mask)
+ if text[i] == "*":
+ run_len = 0
+ while i + run_len < n and text[i + run_len] == "*":
+ run_len += 1
+ if run_len >= 4:
+ self._add_text_run(
+ paragraph, text[i : i + run_len], bold, italic, strike
+ )
+ i += run_len
+ continue
+
+ # Handle long run of tildes (separator)
+ if text[i] == "~":
+ run_len = 0
+ while i + run_len < n and text[i + run_len] == "~":
+ run_len += 1
+ if run_len >= 4:
+ self._add_text_run(
+ paragraph, text[i : i + run_len], bold, italic, strike
+ )
+ i += run_len
+ continue
+
+ # Inline $...$ math (conservative parsing)
+ if (
+ text[i] == "$"
+ and self.valves.MATH_ENABLE
+ and self.valves.MATH_INLINE_DOLLAR_ENABLE
+ ):
+ # Avoid treating $$ as inline math here (block math uses $$ on its own line).
+ if text.startswith("$$", i):
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+
+ # Markdown-ish heuristics to reduce false positives:
+ # - Do not allow whitespace right after opening or right before closing
+ # - Avoid cases like "USD$5" where opening is attached to an alnum
+ if i + 1 >= n or text[i + 1].isspace():
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+ if i > 0 and text[i - 1].isalnum():
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+
+ j = i + 1
+ while True:
+ j = text.find("$", j)
+ if j == -1:
+ break
+ # Skip escaped dollars inside: "\$"
+ if j > 0 and text[j - 1] == "\\":
+ j += 1
+ continue
+ break
+
+ if j != -1:
+ inner = text[i + 1 : j]
+ if (
+ inner
+ and "\n" not in inner
+ and not inner[0].isspace()
+ and not inner[-1].isspace()
+ ):
+ # Treat "$5" as currency more often than math.
+ if _CURRENCY_NUMBER_RE.match(inner) and (
+ i == 0 or text[i - 1].isspace()
+ ):
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+ # Disallow digit immediately following the closing $ (common in prices like "$5.00" already handled above).
+ if j + 1 < n and text[j + 1].isdigit():
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+ self._add_inline_equation(
+ paragraph, inner, bold=bold, italic=italic, strike=strike
+ )
+ i = j + 1
+ continue
+
+ self._add_text_run(paragraph, "$", bold, italic, strike)
+ i += 1
+ continue
+
+ if text.startswith("~~", i):
+ j = text.find("~~", i + 2)
+ if j != -1:
+ self._add_inline_segments(
+ paragraph,
+ text[i + 2 : j],
+ bold=bold,
+ italic=italic,
+ strike=True,
+ )
+ i = j + 2
+ continue
+
+ if text.startswith("**", i):
+ j = text.find("**", i + 2)
+ if j != -1:
+ self._add_inline_segments(
+ paragraph,
+ text[i + 2 : j],
+ bold=True,
+ italic=italic,
+ strike=strike,
+ )
+ i = j + 2
+ continue
+
+ if text.startswith("__", i):
+ j = text.find("__", i + 2)
+ if j != -1:
+ self._add_inline_segments(
+ paragraph,
+ text[i + 2 : j],
+ bold=True,
+ italic=italic,
+ strike=strike,
+ )
+ i = j + 2
+ continue
+
+ if text[i] == "*" and (i + 1 >= n or text[i + 1] != "*"):
+ j = text.find("*", i + 1)
+ if j != -1:
+ self._add_inline_segments(
+ paragraph,
+ text[i + 1 : j],
+ bold=bold,
+ italic=True,
+ strike=strike,
+ )
+ i = j + 1
+ continue
+
+ if text[i] == "_" and (i + 1 >= n or text[i + 1] != "_"):
+ j = text.find("_", i + 1)
+ if j != -1:
+ self._add_inline_segments(
+ paragraph,
+ text[i + 1 : j],
+ bold=bold,
+ italic=True,
+ strike=strike,
+ )
+ i = j + 1
+ continue
+
+ if text[i] == "[":
+ close = text.find("]", i + 1)
+ if close != -1 and close + 1 < n and text[close + 1] == "(":
+ close_paren = text.find(")", close + 2)
+ if close_paren != -1:
+ label = text[i + 1 : close]
+ url = text[close + 2 : close_paren]
+ self._add_hyperlink(paragraph, label, url)
+ i = close_paren + 1
+ continue
+ # Citation marker like [12] -> internal link to References.
+ if close != -1:
+ inner = text[i + 1 : close].strip()
+ if inner.isdigit():
+ idx = int(inner)
+ anchor = self._citation_anchor_by_index.get(idx)
+ if anchor:
+ self._add_internal_hyperlink(paragraph, f"[{idx}]", anchor)
+ i = close + 1
+ continue
+
+ m = _AUTO_URL_RE.match(text, i)
+ if m:
+ raw = m.group(0)
+ trimmed = raw
+ while trimmed and trimmed[-1] in ".,;:!?)]}":
+ trimmed = trimmed[:-1]
+ suffix = raw[len(trimmed) :]
+
+ normalized = self._normalize_url(trimmed)
+ if normalized:
+ # Display the original (trimmed) text; use normalized URL as the target.
+ self._add_hyperlink(
+ paragraph, trimmed, normalized, display_text=trimmed
+ )
+ else:
+ self._add_text_run(paragraph, raw, bold, italic, strike)
+ i += len(raw)
+ continue
+
+ if suffix:
+ self._add_text_run(paragraph, suffix, bold, italic, strike)
+ i += len(raw)
+ continue
+
+ j = next_special(i)
+ if j == i:
+ # Unmatched special character; treat literally to avoid infinite loops.
+ self._add_text_run(paragraph, text[i], bold, italic, strike)
+ i += 1
+ else:
+ self._add_text_run(paragraph, text[i:j], bold, italic, strike)
+ i = j
+
def _normalize_url(self, url: str) -> str:
u = (url or "").strip()
if u.lower().startswith("www."):
@@ -871,618 +2427,107 @@ class Action:
hyperlink.append(new_run)
cast(Any, paragraph)._p.append(hyperlink)
- def _add_text_run(self, paragraph, s: str, bold: bool, italic: bool, strike: bool):
- if not s:
- return
- run = paragraph.add_run(s)
- if bold:
- run.bold = True
- if italic:
- run.italic = True
- if strike:
- run.font.strike = True
-
- # Set Chinese font (copying from existing add_paragraph logic)
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "宋体")
-
- def _add_inline_code(self, paragraph, s: str):
- if s == "":
- return
-
- # Simple inline code without URL parsing for now, or copy full logic if needed.
- # For now, just basic styling to match existing.
- run = paragraph.add_run(s)
- run.font.name = "Consolas"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimHei")
- run.font.size = Pt(10)
- shading = OxmlElement("w:shd")
- shading.set(qn("w:fill"), "E8E8E8")
- run._element.rPr.append(shading)
-
- def _add_hyperlink_code(self, paragraph, display_text: str, url: str):
- u = self._normalize_url(url)
- if not u:
- self._add_inline_code(paragraph, display_text)
- return
-
- part = getattr(paragraph, "part", None)
- if part is None or not hasattr(part, "relate_to"):
- self._add_inline_code(paragraph, display_text)
- return
-
- r_id = part.relate_to(u, RT.HYPERLINK, is_external=True)
-
- hyperlink = OxmlElement("w:hyperlink")
- hyperlink.set(qn("r:id"), r_id)
-
- new_run = OxmlElement("w:r")
- rPr = OxmlElement("w:rPr")
-
- rFonts = OxmlElement("w:rFonts")
- rFonts.set(qn("w:ascii"), "Consolas")
- rFonts.set(qn("w:hAnsi"), "Consolas")
- rFonts.set(qn("w:eastAsia"), "SimHei")
- rPr.append(rFonts)
-
- sz = OxmlElement("w:sz")
- sz.set(qn("w:val"), "20") # 10pt
- rPr.append(sz)
-
- shading = OxmlElement("w:shd")
- shading.set(qn("w:fill"), "E8E8E8")
- rPr.append(shading)
-
- new_run.append(rPr)
-
- t = OxmlElement("w:t")
- t.text = display_text
- new_run.append(t)
-
- hyperlink.append(new_run)
- cast(Any, paragraph)._p.append(hyperlink)
-
- def markdown_to_docx(
+ def _add_inline_equation(
self,
- markdown_text: str,
- top_heading: str = "",
- has_h1: bool = False,
- sources: Optional[List[dict]] = None,
- ) -> Document:
- """
- 将 Markdown 文本转换为 Word 文档
- 支持：标题、段落、粗体、斜体、代码块、列表、表格、链接、原生数学公式、引用和移除思考过程。
- """
- doc = Document()
+ paragraph,
+ latex: str,
+ bold: bool = False,
+ italic: bool = False,
+ strike: bool = False,
+ ):
+ latex = (latex or "").strip()
+ if not latex:
+ return
- # 设置默认中文字体
- self.set_document_default_font(doc)
+ if not self.valves.MATH_ENABLE or not LATEX_MATH_AVAILABLE:
+ self._add_text_run(
+ paragraph, f"\\({latex}\\)", bold=bold, italic=italic, strike=strike
+ )
+ return
- # 构建引用
- self._citation_refs = self._build_citation_refs(sources)
+ try:
+ mathml = latex_to_mathml(latex)
+ omml = mathml2omml.convert(mathml)
+ o_math = self._omml_oMath_element(omml)
+ run = paragraph.add_run()
+ run.bold = bold
+ run.italic = italic
+ run.font.strike = strike
+ cast(Any, run)._r.append(o_math)
+ except Exception as exc:
+ logger.warning(f"Inline math conversion failed; keeping literal: {exc}")
+ self._add_text_run(
+ paragraph, f"\\({latex}\\)", bold=bold, italic=italic, strike=strike
+ )
- # 移除思考过程
- markdown_text = self._strip_reasoning_blocks(markdown_text)
-
- # 若正文无一级标题且有对话标题，则作为一级标题写入
- if top_heading and not has_h1:
- self.add_heading(doc, top_heading, 1)
-
- lines = markdown_text.split("\n")
- i = 0
- in_code_block = False
- code_block_content = []
- code_block_lang = ""
- in_list = False
- list_items = []
- list_type = None # 'ordered' or 'unordered'
-
- while i < len(lines):
- line = lines[i]
-
- # 处理代码块
- if line.strip().startswith("```"):
- if not in_code_block:
- # 先处理之前积累的列表
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- in_code_block = True
- code_block_info_raw = line.strip()[3:].strip()
- code_block_lang, code_block_attrs = self._parse_fence_info(
- code_block_info_raw
- )
- code_block_content = []
- else:
- # 代码块结束
- in_code_block = False
- code_text = "\n".join(code_block_content)
-
- # 检查是否为 Mermaid 或 Flowchart
- mermaid_langs = {
- "mermaid",
- "flowchart",
- "sequence",
- "gantt",
- "class",
- "state",
- "pie",
- "er",
- "journey",
- "gitgraph",
- "mindmap",
- }
-
- if code_block_lang.lower() in mermaid_langs:
- # 创建 Mermaid 块对象
- block = _MermaidFenceBlock(
- info_raw=code_block_info_raw,
- language=code_block_lang,
- attrs=code_block_attrs,
- source=code_text,
- )
- # Handle Mermaid diagram
- if code_block_lang == "mermaid":
- # Optimize layout if enabled
- if self.valves.MERMAID_OPTIMIZE_LAYOUT:
- # Replace LR with TD for graph and flowchart
- code_text = re.sub(
- r"^(graph|flowchart)\s+LR\b",
- r"\1 TD",
- code_text,
- flags=re.MULTILINE | re.IGNORECASE,
- )
-
- self._insert_mermaid_placeholder(doc, code_text)
- else:
- # Insert Placeholder using the block object
- self._insert_mermaid_placeholder(doc, block)
- else:
- self.add_code_block(doc, code_text, code_block_lang)
-
- code_block_content = []
- code_block_lang = ""
- i += 1
- continue
-
- if in_code_block:
- code_block_content.append(line)
- i += 1
- continue
-
- # 处理数学块: $$...$$ 或 \[...\]
- # 简单检测: 如果行以 $$ 或 \[ 开头，则视为数学块开始
- stripped_line = line.strip()
- if stripped_line.startswith("$$") or stripped_line.startswith("\\["):
- # 先处理之前积累的列表
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- # 检查是否为单行块，如 $$ E=mc^2 $$
- if (
- stripped_line.startswith("$$")
- and stripped_line.endswith("$$")
- and len(stripped_line) > 4
- ) or (
- stripped_line.startswith("\\[")
- and stripped_line.endswith("\\]")
- and len(stripped_line) > 4
- ):
- # 提取内容
- if stripped_line.startswith("$$"):
- math_content = stripped_line[2:-2]
- else:
- math_content = stripped_line[2:-2]
- self._add_display_equation(doc, math_content)
- i += 1
- continue
-
- # 多行数学块
- math_lines = []
- # 移除开头标记
- if stripped_line.startswith("$$"):
- current_line_content = stripped_line[2:]
- end_marker = "$$"
- else:
- current_line_content = stripped_line[2:]
- end_marker = "\\]"
-
- if current_line_content.strip():
- math_lines.append(current_line_content)
-
- i += 1
- block_closed = False
- while i < len(lines):
- next_line = lines[i]
- if next_line.strip().endswith(end_marker):
- # 找到结束标记
- content_before_end = next_line.strip()[: -len(end_marker)]
- if content_before_end.strip():
- math_lines.append(content_before_end)
- block_closed = True
- i += 1
- break
- math_lines.append(next_line)
- i += 1
-
- self._add_display_equation(doc, "\n".join(math_lines))
- continue
-
- # 处理表格
- if line.strip().startswith("|") and line.strip().endswith("|"):
- # 先处理之前积累的列表
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- table_lines = []
- while i < len(lines) and lines[i].strip().startswith("|"):
- table_lines.append(lines[i])
- i += 1
- self.add_table(doc, table_lines)
- continue
-
- # 处理标题
- header_match = re.match(r"^(#{1,6})\s+(.+)$", line.strip())
- if header_match:
- # 先处理之前积累的列表
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- level = len(header_match.group(1))
- text = header_match.group(2)
- self.add_heading(doc, text, level)
- i += 1
- continue
-
- # 处理无序列表
- unordered_match = re.match(r"^(\s*)[-*+]\s+(.+)$", line)
- if unordered_match:
- if not in_list or list_type != "unordered":
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = True
- list_type = "unordered"
- indent = len(unordered_match.group(1)) // 2
- list_items.append((indent, unordered_match.group(2)))
- i += 1
- continue
-
- # 处理有序列表
- ordered_match = re.match(r"^(\s*)\d+[.)]\s+(.+)$", line)
- if ordered_match:
- if not in_list or list_type != "ordered":
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = True
- list_type = "ordered"
- indent = len(ordered_match.group(1)) // 2
- list_items.append((indent, ordered_match.group(2)))
- i += 1
- continue
-
- # 处理引用块
- if line.strip().startswith(">"):
- # 先处理之前积累的列表
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- # 收集连续的引用行
- blockquote_lines = []
- while i < len(lines) and lines[i].strip().startswith(">"):
- # 移除开头的 > 和可能的空格
- quote_line = re.sub(r"^>\s?", "", lines[i])
- blockquote_lines.append(quote_line)
- i += 1
- self.add_blockquote(doc, "\n".join(blockquote_lines))
- continue
-
- # 处理水平分割线
- if re.match(r"^[-*_]{3,}$", line.strip()):
- # 先处理之前积累的列表
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- self.add_horizontal_rule(doc)
- i += 1
- continue
-
- # 处理空行
- if not line.strip():
- # 列表结束
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
- i += 1
- continue
-
- # 处理普通段落
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
- list_items = []
- in_list = False
-
- self.add_paragraph(doc, line)
- i += 1
-
- # 处理剩余的列表
- if in_list and list_items:
- self.add_list_to_doc(doc, list_items, list_type)
-
- # 添加参考资料章节
- if self._citation_refs:
- self._add_references_section(doc)
-
- return doc
-
- def set_document_default_font(self, doc: Document):
- """设置文档默认字体，确保中英文都正常显示"""
- # 设置正文样式
- style = doc.styles["Normal"]
- font = style.font
- font.name = "Times New Roman" # 英文字体
- font.size = Pt(11)
-
- # 设置中文字体
- style._element.rPr.rFonts.set(qn("w:eastAsia"), "宋体")
-
- # 设置段落格式
- paragraph_format = style.paragraph_format
- paragraph_format.line_spacing_rule = WD_LINE_SPACING.ONE_POINT_FIVE
- paragraph_format.space_after = Pt(6)
-
- def add_heading(self, doc: Document, text: str, level: int):
- """添加标题"""
- # Word 标题级别从 0 开始，Markdown 从 1 开始
- heading_level = min(level, 9) # Word 最多支持 Heading 9
- heading = doc.add_heading(level=heading_level)
-
- # 解析并添加格式化文本
- self.add_formatted_text(heading, text)
-
- # 设置中文字体
- for run in heading.runs:
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "黑体")
- run.font.color.rgb = RGBColor(0, 0, 0)
-
- def add_paragraph(self, doc: Document, text: str):
- """添加段落，支持内联格式"""
- paragraph = doc.add_paragraph()
- self.add_formatted_text(paragraph, text)
-
- # 设置中文字体
- for run in paragraph.runs:
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "宋体")
-
- def add_formatted_text(self, paragraph, text: str):
- """
- 解析 Markdown 内联格式并添加到段落
- 支持：粗体、斜体、行内代码、链接、删除线、行内公式、引用
- """
- # 定义格式化模式
- patterns = [
- # Inline Math \( ... \)
- (r"\\\((.+?)\\\)", {"math": True}),
- # Inline Math $...$ (single dollar signs, non-greedy)
- (r"(?<!\$)\$(?!\$)([^$]+?)\$(?!\$)", {"math": True}),
- # Citations [1], [2], etc.
- (r"\[(\d+)\]", {"citation": True}),
- # 粗斜体 ***text*** 或 ___text___
- (r"\*\*\*(.+?)\*\*\*|___(.+?)___", {"bold": True, "italic": True}),
- # 粗体 **text** 或 __text__
- (r"\*\*(.+?)\*\*|__(.+?)__", {"bold": True}),
- # 斜体 *text* 或 _text_
- (
- r"(?<!\*)\*(?!\*)(.+?)(?<!\*)\*(?!\*)|(?<!_)_(?!_)(.+?)(?<!_)_(?!_)",
- {"italic": True},
- ),
- # 行内代码 `code`
- (r"`([^`]+)`", {"code": True}),
- # 链接 [text](url)
- (r"\[([^\]]+)\]\(([^)]+)\)", {"link": True}),
- # 删除线 ~~text~~
- (r"~~(.+?)~~", {"strike": True}),
- ]
-
- # 收集所有匹配项
- all_matches = []
-
- for pattern, style in patterns:
- for match in re.finditer(pattern, text):
- # 获取匹配的文本内容
- groups = match.groups()
- matched_text = next((g for g in groups if g is not None), "")
-
- # Special handling for citations to ensure they map to valid refs
- if style.get("citation"):
- try:
- idx = int(matched_text)
- # Only treat as citation if we have a corresponding reference
- # Check if idx exists in our refs list (1-based index)
- if not any(r.idx == idx for r in self._citation_refs):
- continue
- except ValueError:
- continue
-
- all_matches.append(
- {
- "start": match.start(),
- "end": match.end(),
- "text": matched_text,
- "style": style,
- "full_match": match.group(0),
- "url": (
- groups[1] if style.get("link") and len(groups) > 1 else None
- ),
- }
- )
-
- # 按位置排序
- all_matches.sort(key=lambda x: x["start"])
-
- # 移除重叠的匹配项
- filtered_matches = []
- last_end = 0
- for m in all_matches:
- if m["start"] >= last_end:
- filtered_matches.append(m)
- last_end = m["end"]
-
- # 构建最终文本
- pos = 0
- for match in filtered_matches:
- # 添加匹配项之前的普通文本
- if match["start"] > pos:
- plain_text = text[pos : match["start"]]
- if plain_text:
- self._add_text_run(paragraph, plain_text, False, False, False)
-
- # 添加格式化文本
- style = match["style"]
- run_text = match["text"]
-
- if style.get("math"):
- self._add_inline_equation(paragraph, run_text)
- elif style.get("citation"):
- idx = int(run_text)
- # Find the anchor for this index
- ref = next((r for r in self._citation_refs if r.idx == idx), None)
- if ref:
- self._add_internal_hyperlink(paragraph, f"[{idx}]", ref.anchor)
- else:
- self._add_text_run(paragraph, f"[{idx}]", False, False, False)
- elif style.get("link"):
- # 处理链接
- self._add_hyperlink(paragraph, run_text, match["url"])
- elif style.get("code"):
- # 行内代码
- self._add_inline_code(paragraph, run_text)
- else:
- # For bold/italic/strike, check if the text contains inline math
- # Pattern for inline math: \(...\) or $...$
- math_pattern = r"(\\\((.+?)\\\)|\$([^$]+?)\$)"
- math_matches = list(re.finditer(math_pattern, run_text))
-
- if math_matches:
- # Process text with inline math
- text_pos = 0
- for math_match in math_matches:
- # Add text before math
- if math_match.start() > text_pos:
- before_text = run_text[text_pos : math_match.start()]
- self._add_text_run(
- paragraph,
- before_text,
- bold=style.get("bold", False),
- italic=style.get("italic", False),
- strike=style.get("strike", False),
- )
- # Add inline equation with formatting
- latex_content = math_match.group(2) or math_match.group(3)
- self._add_inline_equation(
- paragraph,
- latex_content,
- bold=style.get("bold", False),
- italic=style.get("italic", False),
- strike=style.get("strike", False),
- )
- text_pos = math_match.end()
- # Add remaining text after last math
- if text_pos < len(run_text):
- self._add_text_run(
- paragraph,
- run_text[text_pos:],
- bold=style.get("bold", False),
- italic=style.get("italic", False),
- strike=style.get("strike", False),
- )
- else:
- self._add_text_run(
- paragraph,
- run_text,
- bold=style.get("bold", False),
- italic=style.get("italic", False),
- strike=style.get("strike", False),
- )
-
- pos = match["end"]
-
- if pos < len(text):
- self._add_text_run(paragraph, text[pos:], False, False, False)
+ def _omml_oMath_element(self, omml: str):
+ # Ensure the OMML element declares the math namespace so parse_xml works.
+ m_ns = "http://schemas.openxmlformats.org/officeDocument/2006/math"
+ s = (omml or "").strip()
+ if s.startswith("<m:oMath>") and s.endswith("</m:oMath>"):
+ inner = s[len("<m:oMath>") : -len("</m:oMath>")]
+ s = f'<m:oMath xmlns:m="{m_ns}">{inner}</m:oMath>'
+ elif s.startswith("<m:oMath") and "xmlns:m=" not in s.split(">", 1)[0]:
+ s = s.replace("<m:oMath", f'<m:oMath xmlns:m="{m_ns}"', 1)
+ return parse_xml(s)
def add_code_block(self, doc: Document, code: str, language: str = ""):
- """添加代码块，支持语法高亮"""
- # 语法高亮颜色映射 (基于常见的 IDE 配色)
+ """Add code block with syntax highlighting"""
+ # Token color mapping (based on common IDE themes)
TOKEN_COLORS = {
- Token.Keyword: RGBColor(0, 92, 197), # macOS 风格蓝 - 关键字
+ Token.Keyword: RGBColor(0, 92, 197), # macOS blue - keywords
Token.Keyword.Constant: RGBColor(0, 92, 197),
Token.Keyword.Declaration: RGBColor(0, 92, 197),
Token.Keyword.Namespace: RGBColor(0, 92, 197),
Token.Keyword.Type: RGBColor(0, 92, 197),
- Token.Name.Function: RGBColor(0, 0, 0), # 函数名保持黑色
- Token.Name.Class: RGBColor(38, 82, 120), # 深青蓝 - 类名
- Token.Name.Decorator: RGBColor(170, 51, 0), # 暖橙 - 装饰器
- Token.Name.Builtin: RGBColor(0, 110, 71), # 墨绿 - 内置
- Token.String: RGBColor(196, 26, 22), # 红色 - 字符串
- Token.String.Doc: RGBColor(109, 120, 133), # 灰 - 文档字符串
- Token.Comment: RGBColor(109, 120, 133), # 灰 - 注释
+ Token.Name.Function: RGBColor(0, 0, 0), # Functions stay black
+ Token.Name.Class: RGBColor(38, 82, 120), # Deep cyan-blue - classes
+ Token.Name.Decorator: RGBColor(170, 51, 0), # Warm orange - decorators
+ Token.Name.Builtin: RGBColor(0, 110, 71), # Deep green - builtins
+ Token.String: RGBColor(196, 26, 22), # Red - strings
+ Token.String.Doc: RGBColor(109, 120, 133), # Gray - docstrings
+ Token.Comment: RGBColor(109, 120, 133), # Gray - comments
Token.Comment.Single: RGBColor(109, 120, 133),
Token.Comment.Multiline: RGBColor(109, 120, 133),
- Token.Number: RGBColor(28, 0, 207), # 靛蓝 - 数字
+ Token.Number: RGBColor(28, 0, 207), # Indigo - numbers
Token.Number.Integer: RGBColor(28, 0, 207),
Token.Number.Float: RGBColor(28, 0, 207),
- Token.Operator: RGBColor(90, 99, 120), # 灰蓝 - 运算符
- Token.Punctuation: RGBColor(0, 0, 0), # 黑色 - 标点
+ Token.Operator: RGBColor(90, 99, 120), # Gray-blue - operators
+ Token.Punctuation: RGBColor(0, 0, 0), # Black - punctuation
}
def get_token_color(token_type):
- """递归查找 token 颜色"""
+ """Recursively find token color"""
while token_type:
if token_type in TOKEN_COLORS:
return TOKEN_COLORS[token_type]
token_type = token_type.parent
return None
- # 添加语言标签（如果有）
+ # Add language label if available
if language:
lang_para = doc.add_paragraph()
lang_para.paragraph_format.space_before = Pt(6)
lang_para.paragraph_format.space_after = Pt(0)
lang_para.paragraph_format.left_indent = Cm(0.5)
lang_run = lang_para.add_run(language.upper())
- lang_run.font.name = "Consolas"
+ lang_run.font.name = self.valves.FONT_CODE
lang_run.font.size = Pt(8)
lang_run.font.color.rgb = RGBColor(100, 100, 100)
lang_run.font.bold = True
- # 添加代码块段落
+ # Add code block paragraph
paragraph = doc.add_paragraph()
paragraph.paragraph_format.left_indent = Cm(0.5)
paragraph.paragraph_format.space_before = Pt(3) if language else Pt(6)
paragraph.paragraph_format.space_after = Pt(6)
- # 添加浅灰色背景
+ # Add light gray background
shading = OxmlElement("w:shd")
shading.set(qn("w:fill"), "F7F7F7")
paragraph._element.pPr.append(shading)
- # 尝试使用 Pygments 进行语法高亮
+ # Try to use Pygments for syntax highlighting
if PYGMENTS_AVAILABLE and language:
try:
lexer = get_lexer_by_name(language, stripall=False)
@@ -1495,115 +2540,46 @@ class Action:
if not token_value:
continue
run = paragraph.add_run(token_value)
- run.font.name = "Consolas"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimHei")
+ run.font.name = self.valves.FONT_CODE
+ run._element.rPr.rFonts.set(qn("w:eastAsia"), self.valves.FONT_CODE)
run.font.size = Pt(10)
- # 应用颜色
+ # Apply color
color = get_token_color(token_type)
if color:
run.font.color.rgb = color
- # 关键字加粗
+ # Bold keywords
if token_type in Token.Keyword:
run.font.bold = True
else:
- # 无语法高亮，纯文本显示
+ # No syntax highlighting, plain text display
run = paragraph.add_run(code)
- run.font.name = "Consolas"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "SimHei")
+ run.font.name = self.valves.FONT_CODE
+ run._element.rPr.rFonts.set(qn("w:eastAsia"), self.valves.FONT_CODE)
run.font.size = Pt(10)
- def _insert_mermaid_placeholder(
- self, doc: Document, block: Union[_MermaidFenceBlock, str]
- ):
- self._mermaid_figure_counter += 1
- if isinstance(block, str):
- code = block
- else:
- code = block.source
-
- # 为每个占位符创建唯一的透明 PNG
- # 通过改变图片尺寸来确保 python-docx 不会重用同一个图片文件
- # 使用 figure_counter 来创建不同尺寸（1x1, 1x2, 1x3, ...）
- from PIL import Image
-
- # 创建一个透明图片，尺寸为 1 x counter（确保每个都不同）
- img = Image.new("RGBA", (1, self._mermaid_figure_counter), (0, 0, 0, 0))
- image_stream = io.BytesIO()
- img.save(image_stream, format="PNG")
- image_stream.seek(0)
-
- # 添加居中段落
- paragraph = doc.add_paragraph()
- paragraph.alignment = WD_ALIGN_PARAGRAPH.CENTER
- run = paragraph.add_run()
-
- # 添加图片 (默认大小，稍后 JS 会调整)
- picture = run.add_picture(image_stream, width=Inches(1))
-
- # 设置 Alt Text (描述) 为 "MERMAID_SRC:<encoded_code>"
- # 这是 Python 和 JS 之间的魔法链接
- import urllib.parse
-
- encoded_code = urllib.parse.quote(code)
-
- # 访问底层 XML 设置 docPr descr
- inline = picture._inline
- docPr = inline.docPr
-
- # 使用 .set() 确保属性写入 XML
- docPr.set("descr", f"MERMAID_SRC:{encoded_code}")
- docPr.set("title", "Mermaid Diagram Placeholder")
-
- # 添加标题
- if self.valves.MERMAID_CAPTIONS_ENABLE:
- self._add_mermaid_caption(doc, self._mermaid_figure_counter)
-
- def _add_mermaid_caption(self, doc: Document, figure_number: int):
- if not self._caption_style_name:
- self._ensure_caption_style(doc)
- self._caption_style_name = self.valves.MERMAID_CAPTION_STYLE
-
- caption_text = f"{self.valves.MERMAID_CAPTION_PREFIX} {figure_number}"
- paragraph = doc.add_paragraph(caption_text, style=self._caption_style_name)
- paragraph.alignment = WD_ALIGN_PARAGRAPH.CENTER
- paragraph.paragraph_format.keep_with_next = False
-
- def _ensure_caption_style(self, doc: Document):
- style_name = self.valves.MERMAID_CAPTION_STYLE
- styles = doc.styles
- if style_name not in styles:
- style = styles.add_style(style_name, WD_STYLE_TYPE.PARAGRAPH)
- style.base_style = styles["Normal"]
- style.next_paragraph_style = styles["Normal"]
- font = style.font
- font.name = "Times New Roman"
- font.size = Pt(10)
- font.italic = True
- font.color.rgb = RGBColor(0x55, 0x55, 0x55) # 深灰色
-
- def _parse_fence_info(self, info_raw: str) -> Tuple[str, List[str]]:
- parts = info_raw.split()
- if not parts:
- return "", []
- lang = parts[0]
- attrs = parts[1:]
- return lang, attrs
-
def add_table(self, doc: Document, table_lines: List[str]):
- """添加 Markdown 表格，支持智能列宽、对齐和单元格内格式化"""
+ """Add Markdown table with sane Word sizing/spacing, alignment, and hyperlinks/math support in cells."""
if len(table_lines) < 2:
return
- header_fill = "F2F2F2"
- zebra_fill = "FBFBFB"
+ def _validate_hex(c: str, default: str) -> str:
+ c = c.strip().lstrip("#")
+ if re.fullmatch(r"[0-9A-Fa-f]{6}", c):
+ return c
+ return default
+
+ header_fill = _validate_hex(self.valves.TABLE_HEADER_COLOR, "F2F2F2")
+ zebra_fill = _validate_hex(self.valves.TABLE_ZEBRA_COLOR, "FBFBFB")
def _split_row(line: str) -> List[str]:
+ # Keep empty cells, trim surrounding pipes.
raw = line.strip().strip("|")
return [c.strip() for c in raw.split("|")]
def _is_separator_row(cells: List[str]) -> bool:
+ # Markdown separator: --- / :--- / ---: / :---:
if not cells:
return False
ok = 0
@@ -1648,10 +2624,10 @@ class Action:
table.alignment = WD_TABLE_ALIGNMENT.LEFT
cast(Any, table).autofit = False
- # 单元格边距
+ # Cell margins (twips): smaller padding for compact tables.
self._set_table_cell_margins(table, top=60, bottom=60, left=90, right=90)
- # 列宽按内容比例分配
+ # Column widths: proportional to content, bounded, then normalized to page width.
available_width = int(self._available_block_width(doc))
min_col = max(int(Inches(0.55)), available_width // max(1, num_cols * 3))
@@ -1699,7 +2675,7 @@ class Action:
def _fill_cell(cell, text: str, align: WD_ALIGN_PARAGRAPH, bold: bool = False):
cell.text = ""
parts = [
- p for p in re.split(r"(?:<br\s*/?>|\\n)", text or "") if p is not None
+ p for p in re.split(r"(?:<br\s*/?>|\n)", text or "") if p is not None
]
if not parts:
parts = [""]
@@ -1709,12 +2685,10 @@ class Action:
self.add_formatted_text(para, part)
for run in para.runs:
run.font.size = Pt(9)
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "宋体")
if bold:
run.bold = True
- # 表头行
+ # Header row
header_row = table.rows[0]
self._set_table_header_row_repeat(header_row)
for ci in range(num_cols):
@@ -1727,7 +2701,7 @@ class Action:
bold=True,
)
- # 数据行
+ # Body rows
for ri, row_data in enumerate(body, start=1):
row = table.rows[ri]
for ci in range(num_cols):
@@ -1740,10 +2714,6 @@ class Action:
aligns[ci] if ci < len(aligns) else WD_ALIGN_PARAGRAPH.LEFT,
)
- def _available_block_width(self, doc: Document):
- section = doc.sections[0]
- return section.page_width - section.left_margin - section.right_margin
-
def _set_table_cell_margins(
self, table, top: int, bottom: int, left: int, right: int
):
@@ -1770,35 +2740,30 @@ class Action:
def add_list_to_doc(
self, doc: Document, items: List[Tuple[int, str]], list_type: str
):
- """添加列表"""
+ """Add list"""
for indent, text in items:
paragraph = doc.add_paragraph()
if list_type == "unordered":
- # 无序列表使用项目符号
+ # Unordered list with bullets
paragraph.style = "List Bullet"
else:
- # 有序列表使用编号
+ # Ordered list with numbers
paragraph.style = "List Number"
- # 设置缩进
+ # Set indent
paragraph.paragraph_format.left_indent = Cm(0.5 * (indent + 1))
- # 添加格式化文本
+ # Add formatted text
self.add_formatted_text(paragraph, text)
- # 设置字体
- for run in paragraph.runs:
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "宋体")
-
def add_horizontal_rule(self, doc: Document):
- """添加水平分割线"""
+ """Add horizontal rule"""
paragraph = doc.add_paragraph()
paragraph.paragraph_format.space_before = Pt(12)
paragraph.paragraph_format.space_after = Pt(12)
- # 添加底部边框作为分割线
+ # Add bottom border as horizontal rule
pPr = paragraph._element.get_or_add_pPr()
pBdr = OxmlElement("w:pBdr")
bottom = OxmlElement("w:bottom")
@@ -1810,35 +2775,33 @@ class Action:
pPr.append(pBdr)
def add_blockquote(self, doc: Document, text: str):
- """添加引用块，带有左侧边框和灰色背景"""
+ """Add blockquote with left border and gray background"""
for line in text.split("\n"):
paragraph = doc.add_paragraph()
paragraph.paragraph_format.left_indent = Cm(1.0)
paragraph.paragraph_format.space_before = Pt(3)
paragraph.paragraph_format.space_after = Pt(3)
- # 添加左侧边框
+ # Add left border
pPr = paragraph._element.get_or_add_pPr()
pBdr = OxmlElement("w:pBdr")
left = OxmlElement("w:left")
left.set(qn("w:val"), "single")
- left.set(qn("w:sz"), "24") # 边框粗细
- left.set(qn("w:space"), "4") # 边框与文字间距
- left.set(qn("w:color"), "CCCCCC") # 灰色边框
+ left.set(qn("w:sz"), "24") # Border thickness
+ left.set(qn("w:space"), "4") # Space between border and text
+ left.set(qn("w:color"), "CCCCCC") # Gray border
pBdr.append(left)
pPr.append(pBdr)
- # 添加浅灰色背景
+ # Add light gray background
shading = OxmlElement("w:shd")
shading.set(qn("w:fill"), "F9F9F9")
pPr.append(shading)
- # 添加格式化文本
+ # Add formatted text
self.add_formatted_text(paragraph, line)
- # 设置字体为斜体灰色
+ # Set font to italic gray
for run in paragraph.runs:
- run.font.name = "Times New Roman"
- run._element.rPr.rFonts.set(qn("w:eastAsia"), "楷体")
- run.font.color.rgb = RGBColor(85, 85, 85) # 深灰色文字
+ run.font.color.rgb = RGBColor(85, 85, 85) # Dark gray text
run.italic = True