nemo_voice_agent.pipecat.services.nemo.stt

Module Contents

Classes

Name	Description
`NeMoSTTInputParams`	Input parameters for NeMo STT service.
`NemoSTTService`	NeMo Speech-to-Text service for Pipecat integration.

Functions

Name	Description
`get_stt_service_from_config`	Get the STT service from the config.

Data

ASR_EOU_MODELS

level

API

class nemo_voice_agent.pipecat.services.nemo.stt.NeMoSTTInputParams()

Bases: BaseModel

Input parameters for NeMo STT service.

att_context_size

Optional[List] = [70, 1]

buffer_size

int = 5

config_path

Optional[str] = None

frame_len_in_secs

Optional[float] = 0.08

language

Optional[Language] = Language.EN_US

raw_audio_frame_len_in_secs

Optional[float] = 0.016

class nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService(
    model: typing.Optional[str] = 'nnvidia/parakeet_realtime_...,
    device: typing.Optional[str] = 'cuda:0',
    sample_rate: typing.Optional[int] = 16000,
    params: typing.Optional[nemo_voice_agent.pipecat.services.nemo.stt.NeMoSTTInputParams] = None,
    has_turn_taking: typing.Optional[bool] = None,
    backend: typing.Optional[str] = 'legacy',
    decoder_type: typing.Optional[str] = 'rnnt',
    audio_logger: typing.Optional[nemo_voice_agent.pipecat.services.nemo.audio_logger.AudioLogger] = None,
    ignore_eou_eob: typing.Optional[bool] = False,
    ttfs_p99_latency: typing.Optional[float] = None,
    kwargs = {}
)

Bases: STTService

NeMo Speech-to-Text service for Pipecat integration.

_audio_buffer

= bytearray()

_audio_timestamps

= []

_bytes_per_buffer

_has_turn_taking

= has_turn_taking and not self._ignore_eou_eob

_queue

= asyncio.Queue()

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService._load_model()

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService._reset_stt_state()

Reset the state of the STT service.

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService.can_generate_metrics() -> bool

Only report initial metrics, no need to spam metrics every 80ms

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService.cancel(
    frame: pipecat.frames.frames.CancelFrame
)

async

Handle service cancellation.

Parameters:

frame

CancelFrame

CancelFrame that triggered this method

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService.process_audio_frame(
    frame: pipecat.frames.frames.AudioRawFrame,
    direction: pipecat.processors.frame_processor.FrameDirection
)

async

Process an audio frame for speech recognition.

If the service is muted, this method does nothing. Otherwise, it processes the audio frame and runs speech-to-text on it, yielding transcription results. If the frame has a user_id, it is stored for later use in transcription.

Parameters:

frame

AudioRawFrame

The audio frame to process.

direction

FrameDirection

The direction of frame processing.

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService.process_frame(
    frame: pipecat.frames.frames.Frame,
    direction: pipecat.processors.frame_processor.FrameDirection
)

async

Process incoming frames and handle VAD events.

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService.reset()

Reset the state of the STT service.

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService.run_stt(
    audio: bytes,
    audio_frame: typing.Optional[pipecat.frames.frames.AudioRawFrame] = None
) -> typing.AsyncGenerator[pipecat.frames.frames.Frame, None]

async

Process audio data and generate transcription frames.

Parameters:

audio

bytes

Raw audio bytes to transcribe

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService.set_language(
    language: pipecat.transcriptions.language.Language
)

async

Update the service’s recognition language.

Parameters:

language

Language

New language for recognition

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService.set_model(
    model: str
)

async

Update the service’s model.

Parameters:

model

str

New model name/path to use

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService.start(
    frame: pipecat.frames.frames.StartFrame
)

async

Handle service start.

Parameters:

frame

StartFrame

StartFrame containing initial configuration

nemo_voice_agent.pipecat.services.nemo.stt.NemoSTTService.stop(
    frame: pipecat.frames.frames.EndFrame
)

async

Handle service stop.

Parameters:

frame

EndFrame

EndFrame that triggered this method

nemo_voice_agent.pipecat.services.nemo.stt.get_stt_service_from_config(
    config: omegaconf.DictConfig,
    audio_logger: typing.Optional[nemo_voice_agent.pipecat.services.nemo.audio_logger.AudioLogger] = None
) -> pipecat.services.stt_service.STTService

Get the STT service from the config.

nemo_voice_agent.pipecat.services.nemo.stt.ASR_EOU_MODELS = ['nvidia/parakeet_realtime_eou_120m-v1']

nemo_voice_agent.pipecat.services.nemo.stt.level = logging.getEffectiveLevel()