Description
Problem
The embedding interface (/learn, /ask) is really nice and remarkably useful. Currently though, it seems that a user can have only one active context at a time. This is fine for small contexts where I can, say, clone some GitHub project and start with /learn the docs of the project, and then /learn -d when I'm done. But larger corpuses can take a while to tokenize. If we tokenize something we no longer think is useful, we simply have to unlearn it and start from scratch. lastly, one might imagine wanting to persist the vector db over sessions, or move between hubs, without re-learning from scratch.
Proposed Solution
Allow /learn and /ask to take an optional --db argument or something like that, specifying a (possibly persistent) path to where the vector database would live. by altering the argument, a user could optionally have distinct contexts.
Additional context
I'm still new to this feature and experimenting with it, so perhaps there are good reasons this idea does not make sense.