google_assistant.py

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