Oh ok. Yeah, it is just not really complete yet. What this does is displays a UMAP of the embeddings for the selected feature and you can mouse over for labels.

Oh ok. Yeah, it is just not really complete yet. What this does is displays a UMAP of the embeddings for the selected feature and you can mouse over for labels. It isn't identifying clusters itself yet, finding exemplars, etc. Right now, points that are close to each other are just semantically close - as judged by UMAP. That being said, there isn't much data for this month yet. If you look at last month's data (https://nthesis.ai/public/e4883705-ec05-4e7a-83ac-6b878cc1e86e https://nthesis.ai/public/e4883705-ec05-4e7a-83ac-6b878cc1e8...) , clusters are more apparent (particularly if you view the tags instead of summary).