I tried cycfi for another project (on Bela) and had very disappointing results… not tracking very well, and taking quite a while (musically speaking) to lock on to the pitch - and that was testing with pure sine waves! ( * )
I also tried a couple of other open source projects, and also had disappointing results.
(I cannot remember if it was cycfi, but one of them was also very unstable and would crash)
this should always be remembered, just because there are projects out there that seem to offer what we want, sometimes the ‘details’ mean they don’t really live up to our expectations
also, many of these projects are focused on desktop level platforms, so may be too cpu intensive even for the percussa.
of course, I keep an eye on these projects, and re-test occasionally (though that is effort itself!) , I also look for other projects.
I think generally, accurate pitch detection (and tracking even more so ) is pretty hard and cpu intensive. which is why many technologies for pitch shifting (like granular) avoid detection, and just shift pitch more directly.
( * ) I did have a chat with the developer at the time (last year?) , and iirc, the basic outcome was these algo are defined for quite particular use-cases, and outside of these you will get ‘mixed’ results