[FEATURE] Image embeddings

[davide445]

Describe the feature you'd like As for text, having the possibility to parse a directory and generate embeddings of the included images using specialized models such as CLIP or Florence-2 (using the current integrations with various services like DeepInfra or LocalAI), saving the resulsts on the vectordb + the image reference.

At chat time the text need to be sent to both text and image embeddings modeles, perform semantic search on the db, pass matching text vectors and images to a multimodal LLM able to use both.

This will greatly facilitate the usage of documents containing images (the majority of them).

Additional context Add any other context or screenshots about the