google_assistant.py

   1 #!/usr/bin/env python3
   2
   3 # © Copyright 2021-2022, Scott Gasch
   4
   5 """A module to serve as a local client library around HTTP calls to
   6 the Google Assistant via a local gateway.
   7 """
   8
   9 import logging
  10 import warnings
  11 from dataclasses import dataclass
  12 from typing import Optional
  13
  14 import requests
  15 import speech_recognition as sr  # type: ignore
  16
  17 import config
  18
  19 logger = logging.getLogger(__name__)
  20
  21 parser = config.add_commandline_args(
  22     f"Google Assistant ({__file__})",
  23     "Args related to contacting the Google Assistant",
  24 )
  25 parser.add_argument(
  26     "--google_assistant_bridge",
  27     type=str,
  28     default="http://kiosk.house:3000",
  29     metavar="URL",
  30     help="How to contact the Google Assistant bridge",
  31 )
  32 parser.add_argument(
  33     "--google_assistant_username",
  34     type=str,
  35     metavar="GOOGLE_ACCOUNT",
  36     default="scott.gasch",
  37     help="The user account for talking to Google Assistant",
  38 )
  39
  40
  41 @dataclass
  42 class GoogleResponse:
  43     """A Google response wrapper dataclass."""
  44
  45     success: bool = False
  46     """Did the request succeed (True) or fail (False)?"""
  47
  48     response: str = ''
  49     """The response as a text string, if available."""
  50
  51     audio_url: str = ''
  52     """A URL that can be used to fetch the raw audio response."""
  53
  54     audio_transcription: Optional[str] = None
  55     """A transcription of the audio response, if available.  Otherwise
  56     None"""
  57
  58     def __repr__(self):
  59         return f"""
  60 success: {self.success}
  61 response: {self.response}
  62 audio_transcription: {self.audio_transcription}
  63 audio_url: {self.audio_url}"""
  64
  65
  66 def tell_google(cmd: str, *, recognize_speech=True) -> GoogleResponse:
  67     """Alias for ask_google."""
  68     return ask_google(cmd, recognize_speech=recognize_speech)
  69
  70
  71 def ask_google(cmd: str, *, recognize_speech=True) -> GoogleResponse:
  72     """Send a command string to Google via the google_assistant_bridge as
  73     the user google_assistant_username and return the response.  If
  74     recognize_speech is True, perform speech recognition on the audio
  75     response from Google so as to translate it into text (best effort,
  76     YMMV).  e.g.::
  77
  78         >>> google_assistant.ask_google('What time is it?')
  79         success: True
  80         response: 9:27 PM.
  81         audio_transcription: 9:27 p.m.
  82         audio_url: http://kiosk.house:3000/server/audio?v=1653971233030
  83
  84     """
  85     logging.debug("Asking google: '%s'", cmd)
  86     payload = {
  87         "command": cmd,
  88         "user": config.config['google_assistant_username'],
  89     }
  90     url = f"{config.config['google_assistant_bridge']}/assistant"
  91     r = requests.post(url, json=payload)
  92     success = False
  93     response = ""
  94     audio = ""
  95     audio_transcription: Optional[str] = ""
  96     if r.status_code == 200:
  97         j = r.json()
  98         logger.debug(j)
  99         success = bool(j["success"])
 100         response = j["response"] if success else j["error"]
 101         if success:
 102             logger.debug('Google request succeeded.')
 103             if len(response) > 0:
 104                 logger.debug("Google said: '%s'", response)
 105         audio = f"{config.config['google_assistant_bridge']}{j['audio']}"
 106         if recognize_speech:
 107             recognizer = sr.Recognizer()
 108             r = requests.get(audio)
 109             if r.status_code == 200:
 110                 raw = r.content
 111                 speech = sr.AudioData(
 112                     frame_data=raw,
 113                     sample_rate=24000,
 114                     sample_width=2,
 115                 )
 116                 try:
 117                     audio_transcription = recognizer.recognize_google(
 118                         speech,
 119                     )
 120                     logger.debug("Transcription: '%s'", audio_transcription)
 121                 except sr.UnknownValueError as e:
 122                     logger.exception(e)
 123                     msg = 'Unable to parse Google assistant\'s response.'
 124                     logger.warning(msg)
 125                     warnings.warn(msg, stacklevel=3)
 126                     audio_transcription = None
 127         return GoogleResponse(
 128             success=success,
 129             response=response,
 130             audio_url=audio,
 131             audio_transcription=audio_transcription,
 132         )
 133     else:
 134         message = f'HTTP request to {url} with {payload} failed; code {r.status_code}'
 135         logger.error(message)
 136         return GoogleResponse(
 137             success=False,
 138             response=message,
 139             audio_url=audio,
 140             audio_transcription=audio_transcription,
 141         )