OpenSearch changed cosine similarity score in 2.19

[quachc]

I guess this is just a heads-up, because I spend a couple hours searching before I found the cause.
OpenSearch changed the formula to calculate cosine similarity in version 2.19, apparently they didn't realize this was a breaking change when you use minScore.

opensearch-project/k-NN#2357
opensearch-project/k-NN#2561

So, as of 2.19 it's not required anymore to add a 0.5 boost to the script_score query...
Not sure if it's the same for 3.x... sadly I'm stuck on 2.x, because of the infra guys :)