I have done some work incorporating induced Dirichlet priors for the thresholds, enabling e.g. using an ordered Beta regression model for a response strictly between 0 and 1 (for whatever reason) or a case where we only have observed ones or zeroes, but not both, together with values between 0 and 1.
The code is available here induced_dirichlet_ordbeta.R (github.com) and builds upon discussion in this thread: Dirichlet prior on ordinal regression cutpoints in brms
I did also some small changes to @saudiwin’s code for handling a modeled phi.