Support for extracting images from pdf

[Claudio9701]

Checked

• I searched existing ideas and did not find a similar one
• I added a very descriptive title
• I've clearly described the feature request and motivation for it

Feature request

Based on Issue 10423 and PR 10653 from the langchain python repo.

Using Unstructured seems to be the solutions but it's also possible to implement it in a similar way to the python version.

1. Parse PDF with {extract_images: true}
2. Convert pages to PNG images with pdf-img-convert.js
3. Extract illustrations on the pages images if detected, using a Detectron2 model fine-tuned on digitized books illustrations

Motivation

This would be useful to be able to generate image embeddings with CLIP or a similar model for multi-modal RAG.

Proposal (If applicable)

No response