transport

The bot’s eval transport: a WebSocket server speaking RTVI, plus the eval’s own behavior.

The harness sets per-connection query flags. skip_tts silences the bot’s speech for the session (text mode), applied before on_client_connected so a greeting made there is silent too. capture_bot_audio forwards the bot’s synthesized audio to the harness, for transcription and the recording. capture_bot_images reports each image the bot outputs, by size and format. trigger_disconnect fires the bot’s on_client_disconnected when the connection ends; it is off by default, since bots often cancel their pipeline there and the server serves several scenarios in a row.

The input transport also serves the harness’s image to a vision bot, which has no camera under eval. The user’s audio arrives as a continuous stream, so nothing else is special on the way in.

class pipecat.evals.transport.EvalTransportParams(*, audio_out_enabled: bool = False, audio_out_sample_rate: int | None = None, audio_out_channels: int = 1, audio_out_bitrate: int = 96000, audio_out_10ms_chunks: int = 4, audio_out_filter: BaseAudioFilter | None = None, audio_out_mixer: Mapping[str | None, ~pipecat.audio.mixers.base_audio_mixer.BaseAudioMixer] | None=None, audio_out_destinations: list[str] = <factory>, audio_out_end_silence_secs: int = 2, audio_out_auto_silence: bool = True, audio_out_write_timeout_secs: float = 10.0, audio_in_enabled: bool = False, audio_in_sample_rate: int | None = None, audio_in_channels: int = 1, audio_in_filter: BaseAudioFilter | None = None, audio_in_stream_on_start: bool = True, audio_in_passthrough: bool = True, video_in_enabled: bool = False, video_out_enabled: bool = False, video_out_is_live: bool = False, video_out_width: int = 1024, video_out_height: int = 768, video_out_bitrate: int | None = None, video_out_framerate: int = 30, video_out_color_format: str = 'RGB', video_out_codec: str | None = None, video_out_destinations: list[str] = <factory>, add_wav_header: bool = False, serializer: FrameSerializer | None = None, session_timeout: int | None = None, allowed_origins: list[str] = <factory>)

Bases: SingleClientWebsocketServerParams

Parameters of the eval transport, so a bot’s transport_params names it as such.

class pipecat.evals.transport.EvalInputTransport(transport: SingleClientWebsocketServerTransport, host: str, port: int, params: SingleClientWebsocketServerParams, callbacks: SingleClientWebsocketServerCallbacks, **kwargs)

Bases: SingleClientWebsocketServerInputTransport

Input transport that serves the harness’s image.

A vision bot asks for the user’s camera image; under eval there is no camera, so the image the harness registered for the turn is served instead.

async process_frame(frame: Frame, direction: FrameDirection)

Serve image requests; otherwise behave like the base input transport.

class pipecat.evals.transport.EvalOutputTransport(transport: SingleClientWebsocketServerTransport, params: SingleClientWebsocketServerParams, **kwargs)

Bases: SingleClientWebsocketServerOutputTransport

Output transport that reports the bot’s images to the harness.

The WebSocket server output has no video, so an image the bot outputs is handed to the serializer as it arrives, whether or not the bot enabled video output; the serializer sends it only when the harness asked.

async process_frame(frame: Frame, direction: FrameDirection)

Report output images; otherwise behave like the base output transport.

class pipecat.evals.transport.EvalTransport(params: SingleClientWebsocketServerParams, host: str = 'localhost', port: int = 8765, input_name: str | None = None, output_name: str | None = None)

Bases: SingleClientWebsocketServerTransport

WebSocket server transport used by the eval harness (see the module docstring).

input() → SingleClientWebsocketServerInputTransport

Return an input transport that can serve harness-provided images.

output() → SingleClientWebsocketServerOutputTransport

Return the eval output transport.