So today I started into the basic command-and-control part of the Listener system. The original tack I took was to create escaped commands inside the text stream using the "typing" substitutions (the stuff that converts ,comma into ','):

some text \action_do_something more text

But that got rather grotty rather fast when looking at corner cases (e.g. when you want to type \action to document that mechanism). So I reworked to have two different levels of operation, the first pre-processes to find commands and splits out the text such that you get a sequence of commands-and-text while interpreting. That should allow for e.g. switching the interpretation context in-flight.

Still need to get the actual commands hooked up to do something. The meta-commands (commands related to the dictation process itself, such as "correct that" or "undo that") will be the first, after that I can look at how to make commands registered by apps get passed through and intepreted on the client side.