Use openai-like vlm to ocr

### Question

I tried to use openai-like vlm to ocr, and I wanted to extract images and contexts in images in PDF and transferred to markdown , however I met with error "Can not append a child with children" and the context and image could not be extracted.
I used PictureDesctiptionVLMoption, PDFpipelineoption and md_export_kwargs to improve the ocr ability,but the outcome was bad . Do I have to change the prompt or others to extract