It would be good to reduce the required end of sentence silence (preferably without messing with the model) and make them configurable individually at runtime.
You can currently put something like:
--endpoint.rule2.min-trailing-silence=0.25
--endpoint.rule3.min-trailing-silence=0.25
--endpoint.rule4.min-trailing-silence=0.3
at the end of your <model>/conf/model.conf before running numen to change
the silence timeout of commanding and transcribing as one.
Perhaps numen could support the environment variables NUMEN_COMMAND_TIMEOUT
and NUMEN_TRANSCRIBE_TIMEOUT and actions commandtimeout and transcribetimeout.
maybe instead call the actions
csilenceandtsilence(now that there's going to betmodel).