feat(ocr): add OCR fallback for scanned PDFs and images (#20)

Author: imadreamerboy

MarkItDown.spec

@@ -8,6 +8,17 @@ hiddenimports = [
     "requests",
 ]
 hiddenimports += collect_submodules("markitdown")
+for package in (
+    "azure.ai.documentintelligence",
+    "azure.identity",
+    "pypdfium2",
+    "pypdfium2_raw",
+    "pytesseract",
+):
+    try:
+        hiddenimports += collect_submodules(package)
+    except Exception as e:
+        print(f"Warning: Could not collect hidden imports for {package}: {e}")
 
 datas = [
     ("markitdowngui/resources/markitdown-gui.ico", "markitdowngui/resources"),
@@ -21,6 +32,12 @@ try:
 except Exception as e:
     print(f"Warning: Could not collect magika data files: {e}")
 
+for package in ("pypdfium2", "pypdfium2_raw"):
+    try:
+        datas += collect_data_files(package)
+    except Exception as e:
+        print(f"Warning: Could not collect data files for {package}: {e}")
+
 a = Analysis(
     ["markitdowngui/main.py"],
     pathex=[],
@@ -30,7 +47,10 @@ a = Analysis(
     hookspath=[],
     hooksconfig={},
     runtime_hooks=[],
-    excludes=[],
+    excludes=[
+        "tkinter", "_tkinter",
+        "pytest", "_pytest", "pygments",
+    ],
     noarchive=False,
     optimize=1,
 )
@@ -67,4 +87,3 @@ coll = COLLECT(
     upx_exclude=[],
     name="MarkItDown",
 )
-

README.md

@@ -15,7 +15,8 @@ It focuses on fast multi-file conversion to Markdown with a modern Fluent-style
 - Preview modes: rendered Markdown view and raw Markdown view.
 - Save modes: export as one combined file or separate files.
 - Quick actions: copy Markdown, save output, back to queue, start over.
-- Settings for output folder, batch size, header style, table style, and theme mode (light/dark/system).
+- Optional OCR for scanned PDFs and image files, with Azure Document Intelligence first and local Tesseract fallback.
+- Settings for output folder, batch size, header style, table style, OCR, and theme mode (light/dark/system).
 - Built-in shortcuts dialog, update check action, and about dialog.
 
 ## Installation
@@ -39,6 +40,15 @@ Alternative:
 pip install -e .[dev]
 ```
 
+### OCR Notes
+
+- OCR is optional and disabled by default.
+- Local OCR requires a system `tesseract` binary. Install it from the [official Tesseract project](https://github.com/tesseract-ocr/tesseract). If it is not on your `PATH`, set the executable path in Settings.
+- Azure OCR requires an Azure Document Intelligence endpoint in Settings.
+- Azure Document Intelligence pricing includes [500 free pages per month](https://azure.microsoft.com/en-us/products/ai-foundry/tools/document-intelligence#Pricing) at the time of writing.
+- For API-key auth, set `AZURE_OCR_API_KEY`.
+- If `AZURE_OCR_API_KEY` is not set, Azure OCR falls back to Azure identity credentials supported by `DefaultAzureCredential`.
+
 ## Run the App
 
 ```sh
@@ -97,4 +107,3 @@ uv run pytest -q
 - PySide6 ([LGPLv3 License](https://www.gnu.org/licenses/lgpl-3.0.html))
 - PySide6-Fluent-Widgets / QFluentWidgets ([Project site](https://qfluentwidgets.com))
 
-

README_zh.md

@@ -13,7 +13,8 @@
 - 预览模式支持渲染视图和原始 Markdown 视图。
 - 保存模式支持合并为单文件或分别保存多个文件。
 - 常用操作：复制 Markdown、保存输出、返回队列、重新开始。
-- 设置项包括输出目录、批处理大小、标题样式、表格样式、主题模式（浅色/深色/跟随系统）。
+- 可选 OCR，支持扫描版 PDF 和图片文件，优先使用 Azure Document Intelligence，不可用时回退到本地 Tesseract。
+- 设置项包括输出目录、批处理大小、标题样式、表格样式、OCR 和主题模式（浅色/深色/跟随系统）。
 - 内置快捷键面板、检查更新入口和关于对话框。
 
 ## 安装
@@ -37,6 +38,15 @@ uv sync
 pip install -e .[dev]
 ```
 
+### OCR 说明
+
+- OCR 为可选功能，默认关闭。
+- 本地 OCR 需要系统已安装 `tesseract`。可从 [Tesseract 官方项目](https://github.com/tesseract-ocr/tesseract) 安装。如果它不在 `PATH` 中，可以在设置页里指定可执行文件路径。
+- Azure OCR 需要在设置页里填写 Azure Document Intelligence 终结点。
+- Azure Document Intelligence 定价页面目前标注有 [每月 500 页免费额度](https://azure.microsoft.com/en-us/products/ai-foundry/tools/document-intelligence#Pricing)。
+- 若使用 API Key 认证，请设置 `AZURE_OCR_API_KEY` 环境变量。
+- 如果未设置 `AZURE_OCR_API_KEY`，Azure OCR 会回退到 `DefaultAzureCredential` 支持的 Azure 身份凭据。
+
 ## 运行应用
 
 ```sh