Hello everyone, I would like to ask if it’s possible to set a random seed in order replicate the index creation. I made some tests where I recrated the index from some data several times, but the UUID, the position of the items, and the subsequent similarity search is slightly different.
Hi, you can generate the UUID deterministically this way (Create objects | Weaviate - vector database).
The items are ordered by the UUID so that should fix the ordering issue too.
I am very curious about your comment about similarity search. It should be very, very close each time - I understand the vectorisation may be very slightly different time to time but it shouldn’t produce vastly differently results. How different are your results each time?