I am not really satisfied with the MyCroft software, however the hardware is neat and based on a raspberry pi. When I read about Snips digital voice assistant I wanted to test it and had a Mark 1 laying around. Just download a raspian and put it onto a SD card, which you replace the MyCroft one with. Actually installing Snips with:
npm install -g snips-sam
sam connect raspberrypi
sam init
sam install assistant -i proj_XXX

wasn't too hard. However the sound made me some headache. From the original SD card I could find out, that I had to modify /boot/config.txt and add:
dtparam=i2c_arm=on
dtparam=i2s=on
dtparam=spi=on
dtoverlay=pi3-disable-bt
dtoverlay=pi3-miniuart-bt
dtoverlay=rpi-proto

After reboot and enabling all audio outputs via alsamixer (a bit hidden), I was able to speak first time to snips. However in a ridiculous high pitch and too fast output. Playing around with sample rates in alsa didn't bring success, what worked was to switch to pulseaudio:
apt install pulseaudio
cat << EOF >>/etc/systemd/system/pulseaudio.service
[Unit]
Description=PulseAudio system-wide server
#network only required for remote sinks
After=network-online.target
Wants=network-online.target
[Service]
Type=forking
PIDFile=/var/run/pulse/pid
ExecStart=/usr/bin/pulseaudio --daemonize --system --realtime --log-target=journal
ExecStop=/usr/bin/pulseaudio -k
[Install]
WantedBy=multi-user.target
EOF
systemctl --global disable pulseaudio.service pulseaudio.socket
systemctl enable --now pulseaudio
usermod -G pulse-access -a pi
usermod -G pulse-access -a _snips
systemctl restart snips-audio-server
cat << EOF >> /etc/asound.conf
pcm.pulse {
type pulse
}
ctl.pulse {
type pulse
}
pcm.default pulse
ctl.default pulse
echo default-sample-rate = 44100 >> /etc/pulse/daemon.conf