vector DB that stayed out of the way

[mangoquill]

Maybe I'm missing something obvious.

I'm mostly working with Llama in vLLM on mini PC. The part I'm trying to understand is consumer GPUs.

Would appreciate practical advice rather than theory.