i got an agent as a fast-api-server using the langgraph postgres backend. I changed the embedding model and its dimensions, now i get errors… I looked in the code of PostgresBaseStore and the Migrations and i see no migrations of vectors, when the embedding has changed. Moreover, all vectors are stored in one table, so all vectors should come from the same embedding?
I wonder what are the best practices concenring embeddings and migration of embedded data?
For me it’s not a problem using one embedding globally, but not being able to change the embedding dimension at all seems to be a flaw for long-running agents.
Is the only possiblity to write a migration script myself?
Or is this a feature just langgraph-cloud offers 
Anyway if no solution is available i would open a feature request