google_assistant.py

   1 #!/usr/bin/env python3
   2
   3 import logging
   4 from typing import NamedTuple
   5 import sys
   6
   7 import requests
   8 import speech_recognition as sr  # type: ignore
   9
  10 import config
  11
  12 logger = logging.getLogger(__name__)
  13
  14 parser = config.add_commandline_args(
  15     f"Google Assistant ({__file__})",
  16     "Args related to contacting the Google Assistant",
  17 )
  18 parser.add_argument(
  19     "--google_assistant_bridge",
  20     type=str,
  21     default="http://kiosk.house:3000",
  22     metavar="URL",
  23     help="How to contact the Google Assistant bridge"
  24 )
  25 parser.add_argument(
  26     "--google_assistant_username",
  27     type=str,
  28     metavar="GOOGLE_ACCOUNT",
  29     default="scott.gasch",
  30     help="The user account for talking to Google Assistant"
  31 )
  32
  33
  34 class GoogleResponse(NamedTuple):
  35     success: bool
  36     response: str
  37     audio_url: str
  38     audio_transcription: str
  39
  40     def __repr__(self):
  41         return f"""
  42 success: {self.success}
  43 response: {self.response}
  44 audio_transcription: {self.audio_transcription}
  45 audio_url: {self.audio_url}"""
  46
  47
  48 def tell_google(cmd: str, *, recognize_speech=True) -> GoogleResponse:
  49     """Alias for ask_google."""
  50     return ask_google(cmd, recognize_speech=recognize_speech)
  51
  52
  53 def ask_google(cmd: str, *, recognize_speech=True) -> GoogleResponse:
  54     """Send a command string to Google via the google_assistant_bridge as the
  55     user google_assistant_username and return the response.  If recognize_speech
  56     is True, perform speech recognition on the audio response from Google so as
  57     to translate it into text (best effort, YMMV).
  58     """
  59     logging.debug(f"Asking google: '{cmd}'")
  60     payload = {
  61         "command": cmd,
  62         "user": config.config['google_assistant_username'],
  63     }
  64     url = f"{config.config['google_assistant_bridge']}/assistant"
  65     r = requests.post(url, json=payload)
  66     success = False
  67     response = ""
  68     audio = ""
  69     audio_transcription = ""
  70     if r.status_code == 200:
  71         j = r.json()
  72         success = bool(j["success"])
  73         response = j["response"] if success else j["error"]
  74         if success:
  75             logger.debug('Google request succeeded.')
  76             if len(response) > 0:
  77                 logger.debug(f"Google said: '{response}'")
  78         audio = f"{config.config['google_assistant_bridge']}{j['audio']}"
  79         if recognize_speech:
  80             recognizer = sr.Recognizer()
  81             r = requests.get(audio)
  82             if r.status_code == 200:
  83                 raw = r.content
  84                 speech = sr.AudioData(
  85                     frame_data=raw,
  86                     sample_rate=24000,
  87                     sample_width=2,
  88                 )
  89                 try:
  90                     audio_transcription = recognizer.recognize_google(
  91                         speech,
  92                     )
  93                     logger.debug(f"Transcription: '{audio_transcription}'")
  94                 except sr.UnknownValueError as e:
  95                     logger.exception(e)
  96                     logger.warning('Unable to parse Google assistant\'s response.')
  97                     audio_transcription = None
  98         return GoogleResponse(
  99             success=success,
 100             response=response,
 101             audio_url=audio,
 102             audio_transcription=audio_transcription,
 103         )
 104     else:
 105         message = f'HTTP request to {url} with {payload} failed; code {r.status_code}'
 106         logger.error(message)
 107         return GoogleResponse(
 108             success=False,
 109             response=message,
 110             audio_url=audio,
 111             audio_transcription=audio_transcription,
 112         )
 113         sys.exit(-1)