Wakeword + stt new problem

1

Hello,
I have a weird problem.
I updated to 2.5.7.
Using external MQTT + Porcupine + PyAudio + Kaldi.
Wakeword works and makes the beep but my command is totally ignored.
When I try with the Wake Up button my command works perfectly.

From the log tab
[WARNING:2020-10-16 15:29:36,808] rhasspyserver_hermes: Dialogue management is disabled. ASR will NOT be automatically enabled.
[DEBUG:2020-10-16 15:29:36,807] rhasspyserver_hermes: <- HotwordDetected(model_id=’/profiles/en/raven/default/example-1.wav’, model_version=’’, model_type=‘personal’, current_sensitivity=0.5, site_id=‘IF-MB-CLIENT’, session_id=None, send_audio_captured=None, lang=None)

Any idea?

2

I seem to be having similar problems.

After the 2.5.7 update, I’ve also noticed the PiZero’s seem to take a lot longer to save and restart when making any changes.

3

Hmmm… digging a little deeper, it looks like the session id is None coming from the satellite wakeword. Running the recognition from the web ui, it’s getting a session id and is working correctly.

4

This is expected. The session is created by the dialogue manager after it receives the wake word detection message and the feedback sound playback has finished.

@dror-israel Can you post your configuration and a complete log?

5

server

{
    "dialogue": {
        "satellite_site_ids": "IF-MB-CLIENT",
        "system": "rhasspy"
    },
    "intent": {
        "fuzzywuzzy": {
            "min_confidence": "0.5"
        },
        "satellite_site_ids": "IF-MB-CLIENT",
        "system": "fuzzywuzzy"
    },
    "mqtt": {
        "enabled": "true",
        "host": "192.168.1.***",
        "password": "***",
        "site_id": "IF-RHASSPY-SERVER",
        "username": "***"
    },
    "speech_to_text": {
        "satellite_site_ids": "IF-MB-CLIENT",
        "system": "kaldi"
    },
    "text_to_speech": {
        "satellite_site_ids": "IF-MB-CLIENT",
        "system": "picotts"
    }
}

satellite

{
    "intent": {
        "system": "hermes"
    },
    "microphone": {
        "pyaudio": {
            "device": "0",
            "udp_audio_port": "12202"
        },
        "system": "pyaudio"
    },
    "mqtt": {
        "enabled": "true",
        "host": "192.168.1.***",
        "password": "***",
        "site_id": "IF-MB-CLIENT",
        "username": "***"
    },
    "sounds": {
        "system": "aplay"
    },
    "speech_to_text": {
        "system": "hermes"
    },
    "text_to_speech": {
        "system": "hermes"
    },
    "wake": {
        "porcupine": {
            "sensitivity": "0.45",
            "udp_audio": "12202"
        },
        "system": "porcupine"
    }
}
6

Where can I find the full log?
For the server or the satellite?

7

I’d say both server and satellite.
As to where the logs are, it depends if you are running Docker (docker logs) or not.

8

I have it on a Docker

9

Then: https://docs.docker.com/config/containers/logging/

10

satellite

[DEBUG:2020-10-16 19:05:19,248] rhasspywake_porcupine_hermes: -> HotwordDetected(model_id='/usr/lib/rhasspy/rhasspy-wake-porcupine-hermes/rhasspywake_porcupine_hermes/porcupine/resources/keyword_files/raspberrypi/porcupine.ppn', model_version='', model_type='personal', current_sensitivity=0.45, site_id='IF-MB-CLIENT', session_id=None, send_audio_captured=None, lang=None)
[DEBUG:2020-10-16 19:05:19,249] rhasspywake_porcupine_hermes: Publishing 309 bytes(s) to hermes/hotword/porcupine/detected
[DEBUG:2020-10-16 19:05:19,286] rhasspyserver_hermes: <- HotwordDetected(model_id='/usr/lib/rhasspy/rhasspy-wake-porcupine-hermes/rhasspywake_porcupine_hermes/porcupine/resources/keyword_files/raspberrypi/porcupine.ppn', model_version='', model_type='personal', current_sensitivity=0.45, site_id='IF-MB-CLIENT', session_id=None, send_audio_captured=None, lang=None)
[WARNING:2020-10-16 19:05:19,287] rhasspyserver_hermes: Dialogue management is disabled. ASR will NOT be automatically enabled.
[DEBUG:2020-10-16 19:05:19,331] rhasspywake_porcupine_hermes: <- HotwordToggleOff(site_id='IF-MB-CLIENT', reason=<HotwordToggleReason.PLAY_AUDIO: 'playAudio'>)
[DEBUG:2020-10-16 19:05:19,332] rhasspywake_porcupine_hermes: Disabled
[DEBUG:2020-10-16 19:05:19,386] rhasspyspeakers_cli_hermes: <- AudioPlayBytes(83948 byte(s))
[DEBUG:2020-10-16 19:05:19,387] rhasspyspeakers_cli_hermes: ['aplay', '-q', '-t', 'wav']
[DEBUG:2020-10-16 19:05:19,933] rhasspyspeakers_cli_hermes: -> AudioPlayFinished(id='1f0db880-8c64-4ee8-918f-92e207496f61', session_id='')
[DEBUG:2020-10-16 19:05:19,935] rhasspyspeakers_cli_hermes: Publishing 63 bytes(s) to hermes/audioServer/IF-MB-CLIENT/playFinished
[DEBUG:2020-10-16 19:05:20,011] rhasspywake_porcupine_hermes: <- HotwordToggleOn(site_id='IF-MB-CLIENT', reason=<HotwordToggleReason.PLAY_AUDIO: 'playAudio'>)
[DEBUG:2020-10-16 19:05:20,012] rhasspywake_porcupine_hermes: Enabled
[DEBUG:2020-10-16 19:05:20,067] rhasspywake_porcupine_hermes: Receiving audio
[DEBUG:2020-10-16 19:11:11,829] rhasspyprofile.download: speech_to_text.system pocketsphinx hermes = False
[DEBUG:2020-10-16 19:11:11,829] rhasspyprofile.download: speech_to_text.system kaldi hermes = False
[DEBUG:2020-10-16 19:11:11,829] rhasspyprofile.download: speech_to_text.system deepspeech hermes = False
[DEBUG:2020-10-16 19:11:11,830] rhasspyprofile.download: speech_to_text.pocketsphinx.open_transcription True False = False
[DEBUG:2020-10-16 19:11:11,831] rhasspyprofile.download: speech_to_text.kaldi.open_transcription True False = False
[DEBUG:2020-10-16 19:11:11,831] rhasspyprofile.download: speech_to_text.deepspeech.open_transcription True False = False
[DEBUG:2020-10-16 19:11:11,832] rhasspyprofile.download: speech_to_text.pocketsphinx.mix_weight >0 0 = False
[DEBUG:2020-10-16 19:11:11,833] rhasspyprofile.download: speech_to_text.kaldi.mix_weight >0 0 = False
[DEBUG:2020-10-16 19:11:11,833] rhasspyprofile.download: speech_to_text.deepspeech.mix_weight >0 0 = False
11

server

[DEBUG:2020-10-16 19:05:19,281] rhasspydialogue_hermes: <- HotwordDetected(model_id='/usr/lib/rhasspy/rhasspy-wake-porcupine-hermes/rhasspywake_porcupine_hermes/porcupine/resources/keyword_files/raspberrypi/porcupine.ppn', model_version='', model_type='personal', current_sensitivity=0.45, site_id='IF-MB-CLIENT', session_id=None, send_audio_captured=None, lang=None)
[DEBUG:2020-10-16 19:05:19,284] rhasspydialogue_hermes: Playing WAV /usr/lib/rhasspy/etc/wav/beep_hi.wav
[DEBUG:2020-10-16 19:05:19,287] rhasspydialogue_hermes: -> HotwordToggleOff(site_id='IF-MB-CLIENT', reason=<HotwordToggleReason.PLAY_AUDIO: 'playAudio'>)
[DEBUG:2020-10-16 19:05:19,288] rhasspydialogue_hermes: Publishing 49 bytes(s) to hermes/hotword/toggleOff
[DEBUG:2020-10-16 19:05:19,291] rhasspydialogue_hermes: -> AsrToggleOff(site_id='IF-MB-CLIENT', reason=<AsrToggleReason.PLAY_AUDIO: 'playAudio'>)
[DEBUG:2020-10-16 19:05:19,291] rhasspydialogue_hermes: Publishing 49 bytes(s) to hermes/asr/toggleOff
[DEBUG:2020-10-16 19:05:19,293] rhasspydialogue_hermes: -> AudioPlayBytes(83948 byte(s)) to hermes/audioServer/IF-MB-CLIENT/playBytes/1f0db880-8c64-4ee8-918f-92e207496f61
[DEBUG:2020-10-16 19:05:19,295] rhasspydialogue_hermes: Waiting for playFinished (timeout=1.2012925170068027)
[DEBUG:2020-10-16 19:05:19,354] rhasspyasr_kaldi_hermes: <- AsrToggleOff(site_id='IF-MB-CLIENT', reason=<AsrToggleReason.PLAY_AUDIO: 'playAudio'>)
[DEBUG:2020-10-16 19:05:19,355] rhasspyasr_kaldi_hermes: Disabled (AsrToggleReason.PLAY_AUDIO)
[DEBUG:2020-10-16 19:05:19,968] rhasspydialogue_hermes: <- AudioPlayFinished(id='1f0db880-8c64-4ee8-918f-92e207496f61', session_id='')
[DEBUG:2020-10-16 19:05:19,968] rhasspytts_cli_hermes: <- AudioPlayFinished(id='1f0db880-8c64-4ee8-918f-92e207496f61', session_id='')
[DEBUG:2020-10-16 19:05:19,972] rhasspydialogue_hermes: -> HotwordToggleOn(site_id='IF-MB-CLIENT', reason=<HotwordToggleReason.PLAY_AUDIO: 'playAudio'>)
[DEBUG:2020-10-16 19:05:19,972] rhasspydialogue_hermes: Publishing 49 bytes(s) to hermes/hotword/toggleOn
[DEBUG:2020-10-16 19:05:19,975] rhasspydialogue_hermes: -> AsrToggleOn(site_id='IF-MB-CLIENT', reason=<AsrToggleReason.PLAY_AUDIO: 'playAudio'>)
[DEBUG:2020-10-16 19:05:19,976] rhasspydialogue_hermes: Publishing 49 bytes(s) to hermes/asr/toggleOn
[WARNING:2020-10-16 19:05:19,978] rhasspydialogue_hermes: No session for id IF-MB-CLIENT-porcupine-d97f59c7-de31-4789-bc75-e1e55b0e1362
[DEBUG:2020-10-16 19:05:20,055] rhasspyasr_kaldi_hermes: <- AsrToggleOn(site_id='IF-MB-CLIENT', reason=<AsrToggleReason.PLAY_AUDIO: 'playAudio'>)
[DEBUG:2020-10-16 19:05:20,056] rhasspyasr_kaldi_hermes: Enabled
[DEBUG:2020-10-16 19:11:11,849] rhasspyprofile.download: speech_to_text.system pocketsphinx kaldi = False
[DEBUG:2020-10-16 19:11:11,850] rhasspyprofile.download: speech_to_text.system kaldi kaldi = True
[DEBUG:2020-10-16 19:11:11,851] rhasspyprofile.download: Skipping kaldi/base_dictionary.txt (/profiles/en/kaldi/base_dictionary.txt)
[DEBUG:2020-10-16 19:11:11,852] rhasspyprofile.download: Skipping kaldi/g2p.fst (/profiles/en/kaldi/g2p.fst)
[DEBUG:2020-10-16 19:11:11,853] rhasspyprofile.download: Skipping kaldi/g2p.corpus (/profiles/en/kaldi/g2p.corpus)
[DEBUG:2020-10-16 19:11:11,854] rhasspyprofile.download: Skipping kaldi/model/conf/mfcc.conf (/profiles/en/kaldi/model/conf/mfcc.conf)
[DEBUG:2020-10-16 19:11:11,854] rhasspyprofile.download: Skipping kaldi/model/conf/mfcc_hires.conf (/profiles/en/kaldi/model/conf/mfcc_hires.conf)
[DEBUG:2020-10-16 19:11:11,855] rhasspyprofile.download: Skipping kaldi/model/conf/online_cmvn.conf (/profiles/en/kaldi/model/conf/online_cmvn.conf)
[DEBUG:2020-10-16 19:11:11,856] rhasspyprofile.download: Skipping kaldi/model/extractor/final.dubm (/profiles/en/kaldi/model/extractor/final.dubm)
[DEBUG:2020-10-16 19:11:11,857] rhasspyprofile.download: Skipping kaldi/model/extractor/final.ie (/profiles/en/kaldi/model/extractor/final.ie)
[DEBUG:2020-10-16 19:11:11,858] rhasspyprofile.download: Skipping kaldi/model/extractor/final.mat (/profiles/en/kaldi/model/extractor/final.mat)
[DEBUG:2020-10-16 19:11:11,859] rhasspyprofile.download: Skipping kaldi/model/extractor/global_cmvn.stats (/profiles/en/kaldi/model/extractor/global_cmvn.stats)
[DEBUG:2020-10-16 19:11:11,860] rhasspyprofile.download: Skipping kaldi/model/extractor/online_cmvn.conf (/profiles/en/kaldi/model/extractor/online_cmvn.conf)
[DEBUG:2020-10-16 19:11:11,861] rhasspyprofile.download: Skipping kaldi/model/extractor/splice_opts (/profiles/en/kaldi/model/extractor/splice_opts)
[DEBUG:2020-10-16 19:11:11,862] rhasspyprofile.download: Skipping kaldi/model/model/cmvn_opts (/profiles/en/kaldi/model/model/cmvn_opts)
[DEBUG:2020-10-16 19:11:11,862] rhasspyprofile.download: Skipping kaldi/model/model/den.fst (/profiles/en/kaldi/model/model/den.fst)
[DEBUG:2020-10-16 19:11:11,863] rhasspyprofile.download: Skipping kaldi/model/model/final.mdl (/profiles/en/kaldi/model/model/final.mdl)
[DEBUG:2020-10-16 19:11:11,864] rhasspyprofile.download: Skipping kaldi/model/model/normalization.fst (/profiles/en/kaldi/model/model/normalization.fst)
[DEBUG:2020-10-16 19:11:11,865] rhasspyprofile.download: Skipping kaldi/model/model/tree (/profiles/en/kaldi/model/model/tree)
[DEBUG:2020-10-16 19:11:11,866] rhasspyprofile.download: Skipping kaldi/model/phones/extra_questions.txt (/profiles/en/kaldi/model/phones/extra_questions.txt)
[DEBUG:2020-10-16 19:11:11,867] rhasspyprofile.download: Skipping kaldi/model/phones/nonsilence_phones.txt (/profiles/en/kaldi/model/phones/nonsilence_phones.txt)
[DEBUG:2020-10-16 19:11:11,868] rhasspyprofile.download: Skipping kaldi/model/phones/optional_silence.txt (/profiles/en/kaldi/model/phones/optional_silence.txt)
[DEBUG:2020-10-16 19:11:11,868] rhasspyprofile.download: Skipping kaldi/model/phones/silence_phones.txt (/profiles/en/kaldi/model/phones/silence_phones.txt)
[DEBUG:2020-10-16 19:11:11,869] rhasspyprofile.download: speech_to_text.system deepspeech kaldi = False
[DEBUG:2020-10-16 19:11:11,870] rhasspyprofile.download: speech_to_text.pocketsphinx.open_transcription True False = False
[DEBUG:2020-10-16 19:11:11,871] rhasspyprofile.download: speech_to_text.kaldi.open_transcription True False = False
[DEBUG:2020-10-16 19:11:11,872] rhasspyprofile.download: speech_to_text.deepspeech.open_transcription True False = False
[DEBUG:2020-10-16 19:11:11,872] rhasspyprofile.download: speech_to_text.pocketsphinx.mix_weight >0 0 = False
[DEBUG:2020-10-16 19:11:11,873] rhasspyprofile.download: speech_to_text.kaldi.mix_weight >0 0 = False
[DEBUG:2020-10-16 19:11:11,874] rhasspyprofile.download: speech_to_text.deepspeech.mix_weight >0 0 = False
12

This is clearly an anomaly but I don’t think it is caused by a bug in Rhasspy. The dialogue manager should create a new session but it seems to log this warning instead and stop. I cannot get my head around how this can happen by looking at the dialogue manager code…

With docker ps -a on both server and satellite can you…

  • check that you do not have multiple Rhasspy container running on the server and on the satellite
  • check that both server and satellite are running the latest image (2.5.7)
13

I’m having the same issue and verified mine are all with one image and on the latest.
I also have 2 satellites that have this behavior.

I don’t know if this helps, but they seem to be running really slow, it takes a much longer amount of time to save the config and restart after editing.

In testing through the webpage(s)…
I’ve noticed I’m still not able to pickup the Site ID in Homeassistant (when running the intent handling on the satellite).
My other Satellite still doesn’t seem to pass intents through the server to Homeassistant.

14

Only 1 instance for each and both on 2.5.7…

15

Seems other users have the same kind of problem…

@dror-israel Thanks for your feedback. Is the satellite on an ARM CPU device like a Raspberry Pi?

@dror-israel @Platup Did you retrain after upgrading to 2.5.7? If not, can you retrain and try again? If it still does not work, can you rollback to version 2.5.6, retrain and see if everything goes back to normal?

I’m starting to suspect that something might be wrong with the ARM docker image…

16

I’m trying a fresh install.
Both on rpi 4 / 3B+

17

No luck with re-training/re-starting. I’ll try to rollback later.
Don’t know if this helps, but the site_id fixes in 2.5.7 don’t seem to be there. Testing from the Satellites webpages, it seems to be the same behavior I was seeing on 2.5.6.

18

One more interesting behavior…
After restarting the satellites, I got 1 “successful” response (from the satellite handling it’s own HASS intents). And on the 2nd satellite, it did try to send correctly to the server, but the server didn’t process the intent (like on 2.5.6).
After the first success on each, they will only wake and behave like above.

19

Same behavior for me either

20

Rolled back
It is working now