MariaDBVectorStoreEuclidian vs cosine distance type?

[kohanyirobert]

The MariaDB docs says that when creating a vector index the distance type can be specified and

valid values are cosine and euclidean (the default)

In the MariaDBVectorStore implementation in Spring AI this is not explicitly set (see here).

Contrary to this the default distance type to be used during similarity searches is cosine. Doesn't this makes things behave badly?

I just started experimenting with Spring AI and RAG so I'm no expert but this bugs me.

How does this work?