I think we need vectordb bench on 100M level. If you don't have 100M data, and you don't care about things like filtering and streaming insertion, I vote for P

I think we need vectordb bench on 100M level. If you don't have 100M data, and you don't care about things like filtering and streaming insertion, I vote for PGVector since SQL is convenient enough. However, for large dataset deployment, cost becomes more critical since vector search is computation intensive. Anything like es, mongodb and redis can not even share their results in the benchmark. Also, if you are looking for more fancy features rather simply ANN, purpose built vector database has faster iterations than traditional databases