Comment by Adityav369
Comment by Adityav369 a day ago
Do you mean ingesting the extracted rectangles/ bounding boxes? We're actually working on bounding boxes, this is a good insight and we can add this to the product. However, the way we ingest is literally converting each page to an image then embedding that so the text, layout, diagrams are all encoded in. Would like to know what the exact use case is, can help you better
Why do you convert to image? It’s easy to turn the components of a pdf into separate items and then ingest them individually. I also imagine at some point rasterizing vectors will become a pain point for some reason.