jasper-docker

An armhf Docker image for Jasper, the open source platform for developing always-on, voice-controlled applications

NOTE This Dockerfile builds an Ubuntu 15.04 armhf image optimized for the Raspberry Pi 2 NOTE Please use Dockerfile.amd64 if you would like to build for that platform

IMPORTANT Please read the configuration instructions carefully, and view the Jasper website for further information: http://jasperproject.github.io/documentation/installation/

How to use this Dockerfile

Just pull the image from Docker Hub

The easiest and fastest way to get Jasper is to download a prebuilt image:

docker pull danielchalef/armhf-jasper-docker

Or build your own image as follows:

$ git clone https://github.com/danielchalef/jasper-docker.git
$ cd jasper-docker
$ docker build -t jasper-docker .

and then run the image:

Jasper needs access to your sound devices, which requires sharing the host's devices with your docker container. Note that the C and D in the device names below mean Card and Device respectively. See Gotchas/Default devices below on how to determine the correct PCM devices to share.

$ docker run -ti --privileged \
      -v /dev/snd/pcmC0D0p:/dev/snd/pcmC0D0p \
      -v /dev/snd/pcmC1D0c:/dev/snd/pcmC1D0c \
      -v /dev/snd/controlC0:/dev/snd/controlC0 \
      -v /dev/snd/controlC1:/dev/snd/controlC1 \
      danielchalef/armhf-jasper-docker /bin/bash

You will need to configure Jasper per the Jasper website. This includes running client/populate.py and setting your TTS and STT configurations (see below).

Using this image for your own app

If you'd like to use this image as a base for your own Jasper-based apps, make the first line of your custom Dockerfile:

FROM danielchalef/armhf-jasper-docker:latest

Why Ubuntu?

I favor Ubuntu over Raspian as:

• packages are typically newer
• and many of the interesting scientific and math tools for machine learning, computer vision, and speech to text are available as deb packages and do not require compilation from source.

If you're interested in using an Ubuntu 15.04 Core armhf docker image for your own apps, you can pull mine by:

docker pull danielchalef/armhf-ubuntu-core

Compilation

Where source code has been compiled, the gcc optimization flags -mtune=cortex-a7 -mfpu=neon-vfpv4 have been used. If you intend on using this Dockerfile for armhf devices other than the Raspberry Pi 2, you will need modify these flags accordingly.

Text to Speech, and Speech to Text Configuration

Dependencies for Google TTS and STT have been installed. If you'd like to use other services / apps, please see the Jasper website for required dependencies and configuration. Importantly, since you're using Ubuntu, you should be able to find many of the dependencies are already available as deb packages and will not need to manually compile them from source.

Things that tripped me up, and may do the same to you

Getting Jasper using the right sound devices was a real PITA. It's likely you will be using a USB microphone, and either USB speakers or the local 3.5mm jack. Unfortunately the ALSA device numbers may vary, and you'll need to tell ALSA how to expose them (and possibly hack Jasper source code in order for it to find your audio output - see below).

Unmuting your microphone

Firstly, ensure that your mike is unmuted. You can use alsamixer for this.

Ensure ALSA is configured with the right default devices

Determine the location of your sound devices:

• Input devices

$ arecord -l
card 1: Device [USB PnP Sound Device], device 0: USB Audio [USB Audio]
  Subdevices: 1/1
  Subdevice #0: subdevice #0

• Output devices

$ aplay -l
**** List of PLAYBACK Hardware Devices ****
card 0: ALSA [bcm2835 ALSA], device 0: bcm2835 ALSA [bcm2835 ALSA]
  Subdevices: 8/8
  Subdevice #0: subdevice #0
  Subdevice #1: subdevice #1
  Subdevice #2: subdevice #2
  Subdevice #3: subdevice #3
  Subdevice #4: subdevice #4
  Subdevice #5: subdevice #5
  Subdevice #6: subdevice #6
  Subdevice #7: subdevice #7
card 0: ALSA [bcm2835 ALSA], device 1: bcm2835 ALSA [bcm2835 IEC958/HDMI]
  Subdevices: 1/1
  Subdevice #0: subdevice #0

Note the card and device numbers. Now we'll need to tell ALSA which devices to use as defaults. Create either a ~/.asoundrc or /etc/asound.conf file with the contents below. For more on ALSA config files: http://alsa.opensrc.org/Asoundrc

 pcm.!default {
         type asym
         playback.pcm {
                 type plug
                 slave.pcm "hw:0,0"
         }
         capture.pcm {
                 type plug
                 slave.pcm "hw:1,0"
         } 
 }

Ensure Jasper is using the right output device

To get TTS working (if Jasper is silent), you may need to change the ALSA device Jasper uses for TTS. Unfortunately this can only be done by modifying the source code:

File client/tts.py line number 76

cmd = ['aplay', '-D', 'plughw:0,0', str(filename)]