My goal is to find an efficient way to store and query multimodal embeddings (text + image). Which vector database has the best multimodal support? | Parse