DocxConverter images

[rlendvai-intuitech]

Hi team,

I am using MarkItDown with an configured llm_client (GPT-4o). Image captioning works perfectly for .pptx files, where images are analyzed and described by the LLM.

However, when converting .docx files, images seem to be ignored or result in raw/truncated data:image... strings without any description.

Is LLM Vision support currently implemented for the DocxConverter (which seems to use mammoth), or is this a known limitation?

Thanks!