Create Embeddings on a postgres database with several tables

API
1

HI guys
so i have a question. I have a postgres database that holds about information about clients. One table is a financial table, another is an incident table, the third is a client escalations table and finally the last one holds free text that is filled in the crm like notes…

these are all linked one way or another by cliend ID but the data within each table can sometimes be nested dictionaries…like under notes you could have a dictionary of , {date, text, from, to}

i want to take all of this data and create embedding on it…what i’m confised about is the best way to do it…do I?

  1. Somehow connect all this data into a single flatted dataframe? if so this will be a massive dataframe with 100+ columns

  2. Can i create an embedding for each table? if i do this, will the embedding model know that two tables are connected via a client identifier that is present in both or do i have to somehow force this connection? if so how?

  3. Any other options?

Thanks in advance