If del a collection and then create it again with a new Embedding model and the same collection name, When use this new model to query, get an error：distance between entrypoint and query node: 768 vs 1536: vector lengths don’t match.
Hello @user1454,
Welcome to our community! It’s great to have you here.
The error you’re seeing about different vector lengths indicates that there was an issue during indexing (or Corruption), where vectors of varying dimensions were inserted into the vector space. However, this has been improved in recent Weaviate versions, as incorrect vector lengths should no longer be accepted based on collection configuration.
Could you please share which Weaviate version you’re using, as well as details on how you created the collection and the method you’re using to ingest data?
thank you for your reply.the version is weaviate-client 4.9.3, i used python client v4,
self.weaviate_client.collections.delete("profile_db_202")
after del, I recreated it with a new Embedding model.
this code seems to be temporarily effective,then it will report vector lengths don’t match.
I meant the Weaviate server—are you running the latest DB version, like 1.28 or 1.29?
The image used at that time was the latest version, 6 months ago, I don’t know which version it is now.
Would you please upgrade the server to 1.29.0 and retry?