google_assistant.py

   1 #!/usr/bin/env python3
   2
   3 """A module to serve as a local client library around HTTP calls to
   4 the Google Assistant via a local gateway."""
   5
   6 import logging
   7 import warnings
   8 from dataclasses import dataclass
   9 from typing import Optional
  10
  11 import requests
  12 import speech_recognition as sr  # type: ignore
  13
  14 import config
  15
  16 logger = logging.getLogger(__name__)
  17
  18 parser = config.add_commandline_args(
  19     f"Google Assistant ({__file__})",
  20     "Args related to contacting the Google Assistant",
  21 )
  22 parser.add_argument(
  23     "--google_assistant_bridge",
  24     type=str,
  25     default="http://kiosk.house:3000",
  26     metavar="URL",
  27     help="How to contact the Google Assistant bridge",
  28 )
  29 parser.add_argument(
  30     "--google_assistant_username",
  31     type=str,
  32     metavar="GOOGLE_ACCOUNT",
  33     default="scott.gasch",
  34     help="The user account for talking to Google Assistant",
  35 )
  36
  37
  38 @dataclass
  39 class GoogleResponse:
  40     """A response wrapper."""
  41
  42     success: bool = False
  43     response: str = ''
  44     audio_url: str = ''
  45     audio_transcription: Optional[str] = None  # None if not available.
  46
  47     def __repr__(self):
  48         return f"""
  49 success: {self.success}
  50 response: {self.response}
  51 audio_transcription: {self.audio_transcription}
  52 audio_url: {self.audio_url}"""
  53
  54
  55 def tell_google(cmd: str, *, recognize_speech=True) -> GoogleResponse:
  56     """Alias for ask_google."""
  57     return ask_google(cmd, recognize_speech=recognize_speech)
  58
  59
  60 def ask_google(cmd: str, *, recognize_speech=True) -> GoogleResponse:
  61     """Send a command string to Google via the google_assistant_bridge as the
  62     user google_assistant_username and return the response.  If recognize_speech
  63     is True, perform speech recognition on the audio response from Google so as
  64     to translate it into text (best effort, YMMV).
  65     """
  66     logging.debug("Asking google: '%s'", cmd)
  67     payload = {
  68         "command": cmd,
  69         "user": config.config['google_assistant_username'],
  70     }
  71     url = f"{config.config['google_assistant_bridge']}/assistant"
  72     r = requests.post(url, json=payload)
  73     success = False
  74     response = ""
  75     audio = ""
  76     audio_transcription: Optional[str] = ""
  77     if r.status_code == 200:
  78         j = r.json()
  79         logger.debug(j)
  80         success = bool(j["success"])
  81         response = j["response"] if success else j["error"]
  82         if success:
  83             logger.debug('Google request succeeded.')
  84             if len(response) > 0:
  85                 logger.debug("Google said: '%s'", response)
  86         audio = f"{config.config['google_assistant_bridge']}{j['audio']}"
  87         if recognize_speech:
  88             recognizer = sr.Recognizer()
  89             r = requests.get(audio)
  90             if r.status_code == 200:
  91                 raw = r.content
  92                 speech = sr.AudioData(
  93                     frame_data=raw,
  94                     sample_rate=24000,
  95                     sample_width=2,
  96                 )
  97                 try:
  98                     audio_transcription = recognizer.recognize_google(
  99                         speech,
 100                     )
 101                     logger.debug("Transcription: '%s'", audio_transcription)
 102                 except sr.UnknownValueError as e:
 103                     logger.exception(e)
 104                     msg = 'Unable to parse Google assistant\'s response.'
 105                     logger.warning(msg)
 106                     warnings.warn(msg, stacklevel=3)
 107                     audio_transcription = None
 108         return GoogleResponse(
 109             success=success,
 110             response=response,
 111             audio_url=audio,
 112             audio_transcription=audio_transcription,
 113         )
 114     else:
 115         message = f'HTTP request to {url} with {payload} failed; code {r.status_code}'
 116         logger.error(message)
 117         return GoogleResponse(
 118             success=False,
 119             response=message,
 120             audio_url=audio,
 121             audio_transcription=audio_transcription,
 122         )