#Converter modes

Converter modes control how source files become Markdown.

#Modes

Mode	Behaviour	Best for
`markitdown`	Use MarkItDown only for supported non-Markdown files.	Broad Office/HTML/CSV conversion.
`liteparse`	Use LiteParse only where supported.	PDFs, OCR, images, and layout-sensitive documents.
`auto`	Direct-ingest Markdown/text, try MarkItDown first, and fall back to LiteParse if output is empty. With complexity checking, complex PDFs can prefer LiteParse first.	Safe mixed folders.
`hybrid`	Direct-ingest Markdown/text, prefer LiteParse for PDFs, and use MarkItDown for broad format coverage.	Layout-heavy projects.

#Direct ingestion always comes first

These files are not sent through either converter:

.md, .markdown, .txt, .csv, .json, .jsonl, .xml, .log

Markdown files are already Markdown. Text-like files can be wrapped directly with source metadata.

#Empty-output fallback

In auto and hybrid, MarkItDown may produce no readable text for some PDFs or image-heavy files. The tool checks:

converted_text.strip() == ""

If the output is empty and LiteParse supports the file type, LiteParse is tried as a fallback. The index records:

{
  "fallback": {
    "used": true,
    "from": "markitdown",
    "to": "liteparse",
    "reason": "markitdown_empty_output"
  }
}

#PDF complexity routing

Use:

--liteparse-complexity-check

When enabled, the tool may use LiteParse's complexity check for PDFs before conversion. Complex, scanned, or OCR-heavy PDFs can be routed to LiteParse first.

#Converter metadata

Every included source records:

converter
converter_version
converter_mode
converter_options
fallback metadata
output statistics

This makes a library easier to audit and rebuild.