Indexing clones, compiling omnimodels, and laying out the network.

Tip: Adjust similarity to 95% to spot near‑duplicates fast. (No more model déjà vu.)