Hello,
Are there best practice recommendations for creating vector embeddings for existing columns on a table? For example let's say you have a table around US Postal Addresses, with each component of the address as its own column. I've seen some documentation mention concatenate the data together and create/store embeddings of this string into a single VECTOR column.
In testing this strategy however I've noticed certain questions don't return accurate results unless you have a single new vector column per each column of standard data.
Somewhat of a related question. What do most find to be the most efficient way to manage embeddings for data changes? Update triggers? Asynchronous job of some sorts? Other? Obviously there's going to be quite a bit of factors to consider; however, I'm just looking for more general discussion on what folks have found to be the most efficient for their environments.
Thanks in advance for the assistance!