With Mark II is in the works and Mark III on the roadmap, will there be software changes that may make computational requirements for Picroft higher than the capabilities of RPi 3+?
Will RPi 3+ based Picroft equal to Mark III minus screen/GUI?
My educated guess on this is that there will be no change to the current state: mycroft-core will still run on the RPI3, only resource-intensive tasks like STT/TTS will have to be “outsourced” to another computing-resource (either a local server or a cloud-based service).
The RPi 3 based hardware has several limitations, which we work around using various tactics. For example,
mimic2 is currently cloud-based (for both RPi 3 hardware and Linux installations), but it’s possible it could be on-device for Mark 2 - there’s no way we could make it in-device for RPi 3 due to the compute requirements.
We also work around other quirks on RPi 3 hardware - for instance the setup menu is used to help detect audio devices, and you need to manually edit
wpasupplicant.conf for wireless access - because the software we used previously for this on Mark 1 (which is based on Raspbian Jessie Lite) is not longer supported on Raspbian Jessie Stretch - which is what Picroft is now based on.
So the summary here is “it will be harder to do on-device capabilities that require compute, these will likely be moved to cloud services”, but other functionality is likely to continue to work.