Rasa is a popular choice for the NLU component of Rhasspy, but installing it on a Raspberry Pi was a challenge. I took up the challenge, and as a result I can finally announce a Docker image for the latest Rasa version on ARMv7.
I only tried Rasa’s demo app, and training is quite slow, so I don’t know whether this will be useful for Rhasspy, but If you’re using the Rasa NLU component in Rhasspy, give it a try with the instructions in Rhasspy’s documentation. The Docker command in the instructions should be changed to:
docker run -it -v "$(pwd):/app" -p 5005:5005 koenvervloesem/rasa:latest run --enable-api
This guy might also be of use to you as been looking at RASA of late and looks great but dunno how well it runs on a Pi4 (I am only going to use Pi as KWS anyway)
But did notice that and the thread on the Rasa forum
I have been keeping this Docker image updated to the Rasa 1.x versions, but unfortunately I haven’t had the time yet to look at migrating to Rasa 2.x. This introduces some big upgrades of the dependencies (TensorFlow, TensorFlow Addons, …), so it will take some time to investigate. If someone wants to take a stab at it, pull requests are welcome, also for incomplete attempts. I don’t have a timeframe for my own attempt.