Oryx recommendation for Item-Item CF

Question

Recently I cloned oryx and started working with it, but apparently it has only one type of recommendation algorithm implemented (ALS-SVD). For the type of data that I have this algorithm gives terrible results, I checked it with Mahout. I am wondering anybody has any experience on extending Oryx with other types of CF algorithms, like Item-item (specially for unary data using LLR similarity). If I want to do it by myself what is the starting point after cloning the code. tnx