Would it be possible to add an option to never unload a model?

Here's my usecase:

• I use multiple different LLMs, for different purposes or even just for experimentation.
• But the embeddings and reranking models are always the same.
• Also, due to my coding assistant tool, the embeddings model runs quite often.

So I'd like to be able to:

• Have the embeddings and reranking models always loaded and ready to go, and
• Never unload the "main LLM" I'm using when I need to use the embeddings and reranking models

I guess I could achieve this with profiles. But it would require creating a new profile every time I add a new LLM (and I do that a lot).