Otherwise, what's the point?

if using from_documents then there is further processing in the form of chunking and encoding to get it as a node.

if using nodes then that it typically means those processing as happened already

I see ... so everything ultimately gets rendered into a node.

In that case, pre-processing nodes may indeed help optimize embedding.

Thanks!

🙂

FYI, you can also go the ingestion pipeline route!

I'm not at all familiar with this, but I'll be sure to check into it.

Thanks again 👍

no problem!