Help a programming noob understand how embedding the generated markdown file works

I've generated a markdown file that shows with images and tables like this:


The above image is the raw markdown file.


The above image is the preview of the markdown file.

Now when I want to embed the markdown file into vectors for RAG, does the embedder embed the raw markdown format (where the image is given as a reference in text/code format) or does it embed the actual image as shown in the preview?

I also want to know if it matters whether the embedder embeds the raw image text/code or the actual image. Because at the end of the day, if my final LLM answering the queries can decode the raw text/code reference of the image as the image itself and understand its contents, it doesn't really matter to me if it embeds the base64 info or the actual image.

I'm sure my understanding of the actual RAG pipeline may be incorrect, so if you could correct me anywhere possible, it would be a great help.