You mean the k-means for diversity or DBSCAN for duplicates? Either way it is about 10 lines of scikit-learn code. Send me an email.