This is pretty cool. The title would be a bit more descriptive if it were “Fast Rolling Quantile Filters for Python”, since the high-pass/low-pass filter functionality seems to be the focus. The README mentions it uses binary heaps - if you’re willing to accept some (bounded) approximation, then it should be possible to reduce memory usage and somewhat reduce runtime by using a sketching data structure like Dunning’s…
You can also just get/use exact answers with the `accumulation_tree` package that Python `tdigest` impl uses, though for large windows a B-tree is faster than a red-black ranked/instrumented binary tree. Either tree also scales better than heap pairs to multiple quantiles per moving window (like 5%, 50%, 95%). Python's `blist` package may not have the exact API you need to build from, but it is very close. (`blist` also does not optimize for the "histogramming case" where there can be many duplicate values, IIRC. Instead each value gets its own independent slot in the data structure.)