python xata client out of sync when using .data() vector search
Hi folks, I am trying to use the xata client in python SDK search ingessted records, after editing schema while developing im noticing like complete out of sync data -- missing records, misssing columsn even though the schema has been pushed, etc, how do i fix this as it makes vector search unusable in prod.
My intuition is theres indexing happening on the table with the embeddings -- its strange cuz it was working fine til i edited the schema now ssearching is essentially a no op / does not work
2 Replies
Update -- I was right it seems like there was indexing buildup from deleting rows -- patience was key in this case a walk around the office solved the problem -- im guessing a wait for indexing implementation will help gaurd against this might be nice to throw some indexing details in the docs for folks
Hi Blaise, thanks for reaching out and sharing your findings. There is indeed a propagation delay for search replication and I agree, the docs could be clearer in this regard, we'll work on improving them.