def test_run():
stages = [
mock.MagicMock(),
mock.MagicMock(),
mock.MagicMock(),
]
pipeline = SpeechPipeline(mock.MagicMock(), stages=stages)
@pipeline.event
def on_step(context):
pipeline.stop()
pipeline.run()
def create(spokestack_id: str,
spokestack_secret: str,
sample_rate: int = 16000,
frame_width: int = 20,
**kwargs) -> SpeechPipeline:
"""
Args:
spokestack_id (str): spokestack API id.
spokestack_secret (str): Spokestack API secret.
sample_rate (int): sample rate of the audio (Hz).
frame_width (int): width of the audio frame: 10, 20, or 30 (ms).
Returns:
SpeechPipeline instance with profile configuration.
"""
pipeline = SpeechPipeline(
input_source=PyAudioInput(sample_rate=sample_rate,
frame_width=frame_width,
**kwargs),
stages=[
VoiceActivityDetector(sample_rate=sample_rate,
frame_width=frame_width,
**kwargs),
VoiceActivityTrigger(),
ActivationTimeout(frame_width=frame_width, **kwargs),
CloudSpeechRecognizer(spokestack_id=spokestack_id,
spokestack_secret=spokestack_secret,
sample_rate=sample_rate,
frame_width=frame_width,
**kwargs),
],
)
return pipeline
def test_events():
stages = [
mock.MagicMock(),
mock.MagicMock(),
mock.MagicMock(),
]
pipeline = SpeechPipeline(mock.MagicMock(), stages=stages)
@pipeline.event
def on_speech(context):
context.transcript = "event triggered"
# test empty event
pipeline.event(name="dummy_event")
pipeline.context.event("speech")
assert pipeline.context.transcript == "event triggered"
def test_start_stop():
stages = [
mock.MagicMock(),
mock.MagicMock(),
mock.MagicMock(),
]
pipeline = SpeechPipeline(mock.MagicMock(), stages=stages)
pipeline.start()
assert pipeline.is_running
# test second call to start, which ignores if the pipeline is running
pipeline.start()
pipeline.step()
pipeline.close()
assert not pipeline.is_running
def test_dispatch():
stages = [
mock.MagicMock(),
mock.MagicMock(),
mock.MagicMock(),
]
pipeline = SpeechPipeline(mock.MagicMock(), stages=stages)
pipeline.start()
pipeline._dispatch()
pipeline.is_managed = True
pipeline._dispatch()
def test_activate_deactivate():
stages = [
mock.MagicMock(),
mock.MagicMock(),
mock.MagicMock(),
]
pipeline = SpeechPipeline(mock.MagicMock(), stages=stages)
pipeline.start()
pipeline.activate()
assert pipeline.context.is_active
pipeline.deactivate()
assert not pipeline.context.is_active
def test_cleanup():
stages = [
mock.MagicMock(),
mock.MagicMock(),
mock.MagicMock(),
]
pipeline = SpeechPipeline(mock.MagicMock(), stages=stages)
pipeline.start()
assert pipeline.is_running
pipeline.stop()
assert not pipeline.is_running
pipeline.cleanup()
assert not pipeline._stages
assert not pipeline._input_source
def test_cleanup():
stages = [
mock.MagicMock(),
mock.MagicMock(),
mock.MagicMock(),
]
pipeline = SpeechPipeline(mock.MagicMock(), stages=stages)
pipeline.start()
assert pipeline.is_running
pipeline.stop()
assert not pipeline.is_running
# run after stopped will trigger clean up
pipeline.run()
assert not pipeline._stages
assert not pipeline._input_source
def create(
spokestack_id: str,
spokestack_secret: str,
sample_rate: int = 16000,
frame_width: int = 20,
model_dir: str = "",
**kwargs: Any,
) -> SpeechPipeline:
"""Creates a speech pipeline instance from profile
Args:
spokestack_id (str): spokestack API id.
spokestack_secret (str): Spokestack API secret.
sample_rate (int): sample rate of the audio (Hz).
frame_width (int): width of the audio frame: 10, 20, or 30 (ms).
model_dir (str): Directory containing the tflite wakeword models.
Returns:
"""
pipeline = SpeechPipeline(
input_source=PyAudioInput(frame_width=frame_width,
sample_rate=sample_rate,
**kwargs),
stages=[
AutomaticGainControl(sample_rate=sample_rate,
frame_width=frame_width),
AutomaticNoiseSuppression(sample_rate=sample_rate),
VoiceActivityDetector(
frame_width=frame_width,
sample_rate=sample_rate,
**kwargs,
),
WakewordTrigger(model_dir=model_dir, **kwargs),
ActivationTimeout(frame_width=frame_width, **kwargs),
CloudSpeechRecognizer(
spokestack_secret=spokestack_secret,
spokestack_id=spokestack_id,
**kwargs,
),
],
)
return pipeline
def create(classes: List[str],
model_dir: str,
sample_rate: int = 16000,
frame_width: int = 20,
**kwargs: Any) -> SpeechPipeline:
"""Create a speech pipeline instance from profile.
Args:
model_dir (str): Directory containing the tflite keyword models.
classes: (List(str)): Classes for the keyword model to recognize
sample_rate (int): sample rate of the audio (Hz).
frame_width (int): width of the audio frame: 10, 20, or 30 (ms).
"""
pipeline = SpeechPipeline(
input_source=PyAudioInput(frame_width=frame_width,
sample_rate=sample_rate,
**kwargs),
stages=[
AutomaticGainControl(sample_rate=sample_rate,
frame_width=frame_width),
AutomaticNoiseSuppression(sample_rate=sample_rate),
VoiceActivityDetector(sample_rate=sample_rate,
frame_width=frame_width,
**kwargs),
VoiceActivityTrigger(),
KeywordRecognizer(
classes=classes,
model_dir=model_dir,
sample_rate=sample_rate,
**kwargs,
),
ActivationTimeout(frame_width=frame_width, **kwargs),
],
)
return pipeline
def main():
pipeline = SpeechPipeline(
PyAudioInput(frame_width=20,
sample_rate=16000,
exception_on_overflow=False),
[
VoiceActivityDetector(),
WakewordTrigger(pre_emphasis=0.97, model_dir="tflite"),
GoogleSpeechRecognizer(GOOGLE_CREDS),
ActivationTimeout(),
],
)
dialogue_manager = DialogueManager(
"tflite", "distilbert-base-cased-distilled-squad")
manager = TextToSpeechManager(
TextToSpeechClient(KEY_ID, KEY_SECRET),
PyAudioOutput(),
)
@pipeline.event
def on_activate(context):
print(context.is_active)
@pipeline.event
def on_recognize(context):
pipeline.pause()
answer = dialogue_manager(context.transcript)
manager.synthesize(answer, "text", "demo-male")
pipeline.resume()
@pipeline.event
def on_deactivate(context):
print(context.is_active)
manager.synthesize(dialogue_manager.greet(), "text", "demo-male")
pipeline.start()
pipeline.run()
def main():
pipeline = SpeechPipeline(
PyAudioInput(frame_width=20, sample_rate=16000, exception_on_overflow=False),
[
VoiceActivityDetector(),
WakewordTrigger(pre_emphasis=0.97, model_dir="tflite"),
CloudSpeechRecognizer(spokestack_id=KEY_ID, spokestack_secret=KEY_SECRET),
ActivationTimeout(),
],
)
nlu = TFLiteNLU("tflite")
dialogue_manager = DialogueManager()
manager = TextToSpeechManager(
TextToSpeechClient(KEY_ID, KEY_SECRET), PyAudioOutput(),
)
@pipeline.event
def on_activate(context):
print("active")
@pipeline.event
def on_recognize(context):
pipeline.pause()
results = nlu(context.transcript)
response = dialogue_manager(results)
if response:
manager.synthesize(response, "text", "demo-male")
pipeline.resume()
if results.intent == "AMAZON.StopIntent":
pipeline.stop()
manager.synthesize(Response.WELCOME.value, "text", "demo-male")
pipeline.start()
pipeline.run()
def test_pause_resume():
stages = [
mock.MagicMock(),
mock.MagicMock(),
mock.MagicMock(),
]
pipeline = SpeechPipeline(mock.MagicMock(), stages=stages)
pipeline.start()
assert pipeline.is_running
pipeline.step()
pipeline.pause()
pipeline._input_source.stop.assert_called()
# verify it does nothing
pipeline.step()
pipeline.resume()
pipeline._input_source.start.assert_called()
pipeline.close()
