Snips is an AI-powered voice assistant - Alexa alternative

Feature Requests
21 
2017-07-19 08:37:32 ERROR (Thread-1) [homeassistant.util.yaml] invalid key: "OrderedDict([('objectColor', None)])"
  in "/home/homeassistant/.homeassistant/configuration.yaml", line 235, column 0
2017-07-19 08:37:32 ERROR (MainThread) [homeassistant.bootstrap] Error loading /home/homeassistant/.homeassistant/configuration.yaml: invalid key: "OrderedDict([('objectColor', None)])"
  in "/home/homeassistant/.homeassistant/configuration.yaml", line 235, column 0
2017-07-19 08:37:35 ERROR (Thread-11) [homeassistant.util.yaml] invalid key: "OrderedDict([('objectColor', None)])"
  in "/home/homeassistant/.homeassistant/configuration.yaml", line 235, column 0
2017-07-19 08:37:35 ERROR (MainThread) [homeassistant.components] invalid key: "OrderedDict([('objectColor', None)])"
  in "/home/homeassistant/.homeassistant/configuration.yaml", line 235, column 0

I have received this exact same error in my many attempts at incorporating snips.

22

I took out colour_name

snpis still not working since my MQTT has username and passwors, hence snips cannot connect.

No reply from developer, Project cancelled

1 Like
23

I tested it out, however since I’m not a native english speaker, the engine seems to have trouble processing my voice. Anyone else got problems with the NLP?

24

I’m going to wait until I can use my own mqtt broker and change the hot word. Any idea when will these be implemented? Thanks.

25

is already implemented, I think only without username and password

26

I set it up within hass.io
there you can simply set the mqtt server you want. Otherwise you could always setup a mqtt bridge.

Hass.io - Snips addon

27

Thanks guys. So how was it compare to alexa? Is the response fast? Is the speech recognition accurate?

1 Like
28

There has been a recent update in snips.ai which has improved a lot of things:

  • modular architecture
  • docker less installation (via apt package manager), easier to configure
  • more hotwords (jarvis, chappie)
  • ability for dialogues

I do not own alexa, but the hotword detection is very fast (<1 sec) and accurate. The language processing takes a bit longer (4-6 sec) but is still relatively fast and quite accurate.

I am using it with the PS3eye microphone.

29

Thanks. But I don’t think waiting for 4-6 seconds for a response is ideal. Where do you install this? Is it on a Raspberry Pi?

30

Yes, I run it on a Rasbperry Pi 3 and have Home Assistant (RasPi 2) listen to MQTT.

EDIT: About recognition speed - the 4-6 seconds is the whole process including your command. So from when you stop speaking to the execution (e.g. playing music) is only about 2-4 seconds.

31

in which scenario this time waiting can be not ideal or acceptable ?

32

It works well for me. Here is an example that took 12 seconds. This is start to finish so includes my saying ‘Play rhiana by fleetwood mac’ and also the custom tts (the audio) playing ‘OK, playing music by fleetwood mac’ (yes, it missed the song).

[01:39:52] [Hotword] detected on site default
[01:39:52] [Asr] was asked to stop listening on site default
[01:39:52] [Hotword] was asked to toggle itself 'off' on site default
[01:39:52] [Dialogue] session with id 'fc3ede8e-a674-4cc1-b335-79fdc53f1c2b' was started on site default
[01:39:52] [Asr] was asked to listen on site default
[01:39:59] [Asr] captured text "play rhianna by fleetwood mac" in 6.5s
[01:39:59] [Asr] was asked to stop listening on site default
[01:39:59] [Nlu] was asked to parse input play rhianna by fleetwood mac
[01:39:59] [Nlu] detected intent user_rZEKlr8M7M3__playArtist with probability 0.829 for input "play rhianna by fleetwood mac"
              Slots ->
                 artist_name -> fleetwood mac
[01:39:59] [Dialogue] New intent detected user_rZEKlr8M7M3__playArtist with probability 0.829
              Slots ->
                 artist_name -> fleetwood mac
[01:40:01] [AudioServer] was asked to play a wav of 120.4 kB with id 'QV9QDQ4XAD51G2JHOY7MTRIE3K9AQJFT' on site default
[01:40:04] [AudioServer] finished playing wav with id 'QV9QDQ4XAD51G2JHOY7MTRIE3K9AQJFT'
[01:40:04] [Dialogue] session with id 'fc3ede8e-a674-4cc1-b335-79fdc53f1c2b' was ended on site default. The session was ended because one of the component didn't respond in a ti
mely manner
[01:40:04] [Asr] was asked to stop listening on site default
[01:40:04] [Hotword] was asked to toggle itself 'on' on site default