I've seen that a lot of people (myself included) are interested in getting Plover working on itty-bitty Linux systems, mostly for the purpose of translating strokes as a component of a stand-alone steno keyboard that doesn't need Plover installed anywhere else. I haven't made such a stand-alone keyboard, but I've gotten as far as forking Plover and making a text-only tool out of it that takes raw qwerty steno strokes in from one named pipe and outputs the results on another, basically following Plover's arpeggiate functionality but with a newline in between strokes instead of a space.
https://github.com/KoiOates/ploverRun it with the launch.sh script (I haven't tried installing to see what would happen) and you'll have the named pipes /tmp/plovin and /tmp/plovout to work with. Once the program finishes loading all its pieces you can do things like 'echo "erfvuj" > /tmp/plovin' in one terminal or split and 'tail -f /tmp/plovout' in another and you'll see the word "Plover" in the second terminal (or its raw steno notation if the translation part of the pipeline still hasn't finished warming up).
The tool is basically just a modified Plover cut up to skip the GUI components and get the engine running. Then instead of using oslayer/xkeyboardcontrol.py for input and output, it loads oslayer/nonkeyboardcontrol.py, which is the meat and potatoes of the change. The other edits to the source code either do just enough to get nonkeyboardcontrol.py hooked up to the rest of the program or disable dependencies that are unnecessary to its basic functionality and that wouldn't run from a Raspberry Pi's terminal (or wouldn't run without having to install extra stuff).
Tested on a Raspberry Pi B+ straight from the terminal, running Raspbian. A lot of other stuff was already installed on that one, so on a clean Pi image there may still be other things to apt-get or pip install after cloning the repository that I've missed. When tested with cat and a short file of strokes, it outputted the translation a little slower than the blink of an eye but still faster than a human would type it.
Now I just need to find the time to teach my Raspberry Pi and Stenoboard to work together on this. I have a lot to learn and get in order before I can test this idea, but I was thinking the Stenoboard could either send completed strokes or a different byte for each key's up and down position over SPI to the Raspberry Pi (the pins are at least available to do this), and then the Pi can send the translation back, and the Stenoboard can send the translated keystrokes down the USB cable instead of its usual raw output that requires Plover on the other end. That would at least save the cost of another board to send the finished keystrokes.