Thanks! I'm not too familiar with how AtCoder's lazy segtree works, but any operation that can be done with a lazy segtree can be unified in a treap in the same way. The buttons for enabling sum/min/max and range add/set only just add code into the "value" and "lazy" structs, which is my way of writing generic segtrees.

Currently, I do range updates and queries in a segtree-style traversal for better constant. However, I don't know a way to perform a range reverse that doesn't involve two splits followed by two merges, so I'm not sure if that can be unified without increasing the constant factor by a lot.

I haven't considered generalising keys and indices, but for me I've found it convenient to have the distinction since sometimes I use both (e.g. querying index of some key, finding i'th smallest element). Some other stuff like treap "union" where you merge sorted sets only makes sense in terms of keys and not indices. There are also some weird treaps that might be too hard/not worth to generalise like the treap version of segment tree beats, which hopefully will never be used in a real problem.