9.4 Sizing and Choosing a Vector Store

Three numbers size a vector store: the vector count, the dimension, and the bytes per number. The previous lesson built the help-center search, 1 million document chunks of 768 dimensions shared by many tenants, on a Redis HNSW index with tenant filters. It showed how EF_RUNTIME trades recall for latency, but not how much memory the index needs or whether Redis is the right place for it.

This lesson answers both questions