> The key is to segment the DB using other data before doing the embedding search By segmenting data, is this as simple as adding a further condition to the SQ

> The key is to segment the DB using other data before doing the embedding search By segmenting data, is this as simple as adding a further condition to the SQL? E.g. if your embeddings belonged to a "client" you would start with WHERE client_id - x? > How you chunk text matters. Chunking by paragraph is much better than naive chunking Would love some more info on this. SO if you had a 400 word slab of text, it would be better to create m3 or 4 embeddings from that?