vllm.entrypoints.speech_to_text.realtime.protocol

Classes:

• ErrorEvent –

  Error notification.

• InputAudioBufferAppend –

  Append audio chunk to buffer.

• InputAudioBufferCommit –

  Process accumulated audio buffer.

• SessionCreated –

  Connection established notification.

• SessionUpdate –

  Configure session parameters.

• TranscriptionDelta –

  Incremental transcription text.

• TranscriptionDone –

  Final transcription with usage stats.

ErrorEvent

Bases: OpenAIBaseModel

Error notification.

Source code in vllm/entrypoints/speech_to_text/realtime/protocol.py

class ErrorEvent(OpenAIBaseModel):
    """Error notification."""

    type: Literal["error"] = "error"
    error: str
    code: str | None = None

InputAudioBufferAppend

Bases: OpenAIBaseModel

Append audio chunk to buffer.

Source code in vllm/entrypoints/speech_to_text/realtime/protocol.py

class InputAudioBufferAppend(OpenAIBaseModel):
    """Append audio chunk to buffer."""

    type: Literal["input_audio_buffer.append"] = "input_audio_buffer.append"
    audio: str  # base64-encoded PCM16 @ 16kHz

InputAudioBufferCommit

Bases: OpenAIBaseModel

Process accumulated audio buffer.

Source code in vllm/entrypoints/speech_to_text/realtime/protocol.py

class InputAudioBufferCommit(OpenAIBaseModel):
    """Process accumulated audio buffer."""

    type: Literal["input_audio_buffer.commit"] = "input_audio_buffer.commit"
    final: bool = False

SessionCreated

Bases: OpenAIBaseModel

Connection established notification.

Source code in vllm/entrypoints/speech_to_text/realtime/protocol.py

class SessionCreated(OpenAIBaseModel):
    """Connection established notification."""

    type: Literal["session.created"] = "session.created"
    id: str = Field(default_factory=lambda: f"sess-{random_uuid()}")
    created: int = Field(default_factory=lambda: int(time.time()))

SessionUpdate

Bases: OpenAIBaseModel

Configure session parameters.

Source code in vllm/entrypoints/speech_to_text/realtime/protocol.py

class SessionUpdate(OpenAIBaseModel):
    """Configure session parameters."""

    type: Literal["session.update"] = "session.update"
    model: str | None = None

TranscriptionDelta

Bases: OpenAIBaseModel

Incremental transcription text.

Source code in vllm/entrypoints/speech_to_text/realtime/protocol.py

class TranscriptionDelta(OpenAIBaseModel):
    """Incremental transcription text."""

    type: Literal["transcription.delta"] = "transcription.delta"
    delta: str  # Incremental text

TranscriptionDone

Bases: OpenAIBaseModel

Final transcription with usage stats.

Source code in vllm/entrypoints/speech_to_text/realtime/protocol.py

class TranscriptionDone(OpenAIBaseModel):
    """Final transcription with usage stats."""

    type: Literal["transcription.done"] = "transcription.done"
    text: str  # Complete transcription
    usage: UsageInfo | None = None