https://docs.llamaindex.ai/en/latest/

I have no idea tbh. Is that a setting on the vecs client?

Looking at the source code, if it is the client, it might need to be updated to pass in a client rather than creating one for you https://github.com/run-llama/llama_index/blob/7c27b3125d5e44e2757a8922f7462257a4e70faa/llama_index/vector_stores/supabase.py#L42

Although if you want in-memory vectors, there are a ton of options too (our default vector store, qdrant, chroma, etc.)

thanks. Another question: do the in-memory vector stores behave the same as the supabase vector store? I'm not sure if the semantic search algorithm is the same among them.

I'm sure the search is slightly different. But most implement the same algorithm as everyone else (HNSW, etc,)

I've never noticed a difference between vector stores lol