is the answer time dependent on the hardware where the deployment is done?

hi, is it possible to further improve an index with question/answer pairs that come from user feedback?

hi, i have a chatbot that's supposed to answer domain-specific questions (by embedding about 20manuals/books). i followed that chatbot guide on the doc page and it works.
The issue: sometimes the "conversational" part will answer instead of the "expert" part.How can i make it so that the expert part answers all the domain-specific questions?
I split my manuals by categories and created indexes based on that.
Thanks!

it's the guide i tried to follow, but it's outdated.for example the langchain agents module doesn't exist