What LLM do you use for this? Self hosted or cloud?

This uses the OpenAI API currently. I tried Gemini for embeddings for a little while but didn't seem materially better.

Looks beautiful. How difficult would it be if I want to develop a plugin - Mood detection, Chat Helper (LLM linked), ...

Does this support table detection and extraction?

Yes, it's experimental at the moment: https://docs.vlm.run/guides/doc-ai/guide-visual-grounding