Maybe you can implement an AI search feature so that it can recognise objects within images on top of the existing OCR for text search.
For example something like:
If we can have and something like a block element of collapsible toggle it will be even perfect on top of the feature already in development - Collapsible Headers.