hugvey/hugvey/speech/google.py

222 lines
8.4 KiB
Python

import json
import logging
import os
import pyaudio
import re
import socket
import os
from threading import Thread
from google.cloud.speech import enums
from google.cloud.speech import types
from google.auth.credentials import Credentials
from google.cloud import speech
import asyncio
import threading
import queue
import uuid
from hugvey.communication import LOG_BS
import audioop
mainLogger = logging.getLogger("hugvey")
logger = mainLogger.getChild("google")
class RequireRestart(Exception):
pass
class GoogleVoiceClient(object):
def __init__(self, hugvey, src_rate, credential_file, language_code = "en_GB"):
self.logger = mainLogger.getChild(f"{hugvey.id}").getChild('google')
self.src_rate = src_rate
self.hugvey = hugvey
self.language_code = language_code
os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = credential_file
# Create a thread-safe buffer of audio data
self.buffer = queue.Queue()
self.isRunning = threading.Event()
self.toBeShutdown = False
self.target_rate = 16000
self.cv_laststate = None
self.restart = False
self.task = threading.Thread(target=self.run, name=f"hugvey#{self.hugvey.id}v")
self.task.setDaemon(True)
self.task.start()
self.subsequentMutedFrames = 0
self.lastNonFinalTranscript = None
def pause(self):
self.isRunning.clear()
self.restart = True
def resume(self):
self.buffer = queue.Queue() # have a clear queue when resuming
self.isRunning.set()
def generator(self):
self.logger.debug('start generator')
while not self.toBeShutdown and self.isRunning.is_set():
try:
# set a timeout, as not to wait infinitely for the buffer when
# we actually want to restart
yield self.buffer.get(timeout=.3)
except queue.Empty as e:
self.logger.debug('empty mic buffer - restart?')
# print(self.isRunning.isSet())
self.logger.info('stop generator')
self.restart = False # don't trigger double restart
# raise RequireRestart("Restart required (generator)")
def setLanguage(self, language_code):
if self.language_code == language_code:
return
self.logger.info("Change language from {} to {}".format(self.language_code, language_code))
self.language_code = language_code
self.restart = True
def run(self):
self.isRunning.set()
self.speech_client = speech.SpeechClient()
config = types.RecognitionConfig(
encoding=enums.RecognitionConfig.AudioEncoding.LINEAR16,
sample_rate_hertz=self.src_rate,
language_code=self.language_code)
self.streaming_config = types.StreamingRecognitionConfig(
config=config,
interim_results=True)
while not self.toBeShutdown:
try:
self.logger.info("wait for Google Voice")
if not self.isRunning.wait(timeout=1):
continue # re-ceck toBeShutdown
self.logger.info("Starting Google Voice")
audio_generator = self.generator()
requests = (types.StreamingRecognizeRequest(audio_content=content)
for content in audio_generator)
responses = self.speech_client.streaming_recognize(
self.streaming_config, requests)
self.logger.info("Starting voice loop")
for response in responses:
if not response.results:
self.logger.debug('...')
continue
"""Iterates through server responses and prints them.
The responses passed is a generator that will block until a response
is provided by the server.
Each response may contain multiple results, and each result may contain
multiple alternatives; for details, see https://goo.gl/tjCPAU. Here we
print only the transcription for the top alternative of the top result.
In this case, responses are provided for interim results as well. If the
response is an interim one, print a line feed at the end of it, to allow
the next result to overwrite it, until the response is a final one. For the
final one, print a newline to preserve the finalized transcription.
"""
# The `results` list is consecutive. For streaming, we only care about
# the first result being considered, since once it's `is_final`, it
# moves on to considering the next utterance.
result = response.results[0]
if not result.alternatives:
continue
# Display the transcription of the top alternative.
transcript = result.alternatives[0].transcript
# self.logger.debug("Text: ".format(transcript))
if not result.is_final:
self.logger.debug(f"Text: {transcript}")
self.lastNonFinalTranscript = transcript
else:
self.logger.info(f"Text: {transcript}")
self.lastNonFinalTranscript = None
if result.is_final:
self.logger.info("native final")
msg = {
"event": "speech",
"is_final": result.is_final,
"transcript": transcript.strip(),
}
self.hugvey.queueEvent(msg)
if self.restart:
self.restart = False
raise RequireRestart("Restart required")
if self.toBeShutdown:
self.logger.warn("Stopping voice loop")
break
except RequireRestart as e:
self.restart = False
self.logger.warn("Restart Google Voice. Language: {}".format(self.language_code))
except Exception as e:
self.logger.critical(f"Crashed Google Voice: {e}")
# make sure we always send a 'final' transcript.
if self.lastNonFinalTranscript is not None:
msg = {
"event": "speech",
"is_final": True,
"transcript": self.lastNonFinalTranscript.strip(),
}
self.hugvey.queueEvent(msg)
self.logger.warn("Stop google run()") # finish means wrapping of hugvey#3v thread
def receive(self, chunk):
if not self.task.isAlive():
raise Exception("Voice thread died")
if audioop.max(chunk, 2) == 0:
# mic is muted on client side.
self.subsequentMutedFrames += 1
# self.logger.debug("Muted")
if self.subsequentMutedFrames > 4 and self.isRunning.is_set():
self.logger.info("Pause muted stream!")
self.pause()
return
self.subsequentMutedFrames = 0
# self.logger.debug("We have mic!")
if not self.isRunning.is_set():
self.logger.info("Resume voice")
self.resume()
if not self.isRunning.is_set():
# logger.log(LOG_BS, "Don't put to queue if google is paused")
return
if self.src_rate == self.target_rate:
data = chunk
else:
data, self.cv_laststate = audioop.ratecv(chunk, 2, 1, self.src_rate, self.target_rate, self.cv_laststate)
self.buffer.put_nowait(data)
def shutdown(self):
self.toBeShutdown = True
self.hugvey = None
def triggerStart(self):
pass
def __del__(self):
self.logger.warn("Destroyed google object")