Just wanted to check the general flow...

[konductors]

Hi.

Would I be correct in thinking this:

1. User enters a query
2. Cactus/ the edge LLM processes the query
3. The local LLM's response is returned and a complexity / confidence value is assigned to it
4. IF confidence is below a specified value THEN a cloud LLM is called
5. The user's response is then returned to them from the cloud LLM

So effectively whenever a user sends a request, it is "pre-processed" by the local LLM, which generates its response and grades the complexity / confidence. Then it hands off to a cloud LLM if necessary? i.e; the user's request is processed twice in the case of complex queries which go on to need cloud processing?

Thanks