Feature Request: Add OCR support for vertical layout ancient Chinese books/documents

Description

I would like to request support for OCR processing of vertical layout ancient Chinese books and historical documents. Currently, when processing vertical layout ancient Chinese book PDFs, MarkItDown outputs empty documents without any error messages or warnings.

Current Behavior

• Processing vertical layout ancient Chinese book PDFs results in empty output
• No error messages or warnings are displayed
• The conversion appears to complete successfully but produces no content

Expected Behavior

• MarkItDown should be able to recognize and extract text from vertical layout documents
• Text should be properly oriented and structured in the markdown output
• If OCR fails, appropriate error messages should be provided

Use Case

Ancient Chinese books and historical documents are typically written in vertical columns reading from right to left. These documents are important for:

• Academic research in Chinese studies
• Historical document digitization projects
• Cultural preservation efforts
• Educational material preparation

Environment

• Operating System: Ubuntu 24.04
• Python Version: 3.12
• MarkItDown Version: [Please specify your version]

Possible Solutions

1. Add support for vertical text detection in the OCR pipeline
2. Implement text orientation detection and correction
3. Add specific handling for Chinese/CJK vertical text layouts
4. Provide configuration options for text direction (vertical/horizontal)

Additional Context

This feature would greatly benefit researchers and institutions working with historical Chinese documents, making MarkItDown more suitable for multilingual and multicultural document processing workflows.

Would appreciate any guidance on whether this is feasible or if there are any workarounds available.

Start by reproducing the empty output with a vertical-layout ancient Chinese PDF on Ubuntu 24.04 and inspect the existing OCR pipeline and document-conversion entry points; the issue names no files or tests. Done should include extracted, correctly oriented Markdown text for vertical documents and clear error or warning output when OCR cannot process them.