包package agentscope.realtime

agentscope.realtime

module agentscope.realtime

agentscope.realtime 属于实时语音，低延迟语音与多模态会话。该包共收录 29 个公开类型。

The model side of a voice session: realtime models, their cards and events, transports and VAD. The agent that ties them together is :class:agentscope.agent.RealtimeAgent.

源码Source agentscope/realtime/__init__.py:1

类Classes

`AudioDeltaEvent` 类class	音频增量事件，描述执行过程中的事件。 A chunk of assistant speech.
`AudioFrame` 类class	一种分块的上游用户音频。 A chunk of upstream user audio.
`ControlFrame` 类class	Control Frame，处理低延迟的实时语音与多模态交互。 One upstream control event.
`ControlFrameType` 类class	Control Frame 类型，处理低延迟的实时语音与多模态交互。 The non-audio things a remote client can say.
`DashScopeAudioRealtimeModel` 类class	DashScope 音频实时模型，处理低延迟的实时语音与多模态交互。 The Qwen-Audio-Realtime API over WebSocket.
`DashScopeRealtimeModel` 类class	DashScope 实时模型，处理低延迟的实时语音与多模态交互。 The Qwen-Omni realtime API over WebSocket.
`GeminiRealtimeModel` 类class	Gemini 实时模型，处理低延迟的实时语音与多模态交互。 The Gemini Live API over WebSocket.
`InputTranscriptionEvent` 类class	输入 Transcription 事件，描述执行过程中的事件。 The settled transcript of one user turn.
`InputTranscriptionFailedEvent` 类class	输入 Transcription 失败事件，描述执行过程中的事件。 The provider could not transcribe one committed user turn.
`LocalAudioTransport` 类class	本地音频传输层，处理低延迟的实时语音与多模态交互。 Microphone in, speaker out, via ``sounddevice``.
`ModelDisconnectedError` 类class	模型已断开错误，处理低延迟的实时语音与多模态交互。 The provider closed the session; the next user audio reconnects.
`ModelErrorEvent` 类class	模型错误事件，描述执行过程中的事件。 An error reported by the provider.
`ModelEvent` 类class	模型事件，描述执行过程中的事件。 Common base class, for type hinting only.
`OpenAIRealtimeModel` 类class	OpenAI 实时模型，处理低延迟的实时语音与多模态交互。 The OpenAI Realtime API over WebSocket.
`PlayoutPosition` 类class	Playout Position，处理低延迟的实时语音与多模态交互。 How much assistant audio actually left the speaker — the only sound basis for truncating a barge-in.
`RealtimeModelBase` 类class	实时模型基类，处理低延迟的实时语音与多模态交互。 A bidirectional realtime model: audio in, audio and tool calls out.
`RealtimeModelCard` 类class	实时模型卡片，处理低延迟的实时语音与多模态交互。 Describes one realtime model. Everything here varies with the model name; what the adapter implements lives on :class:`RealtimeModelBase`.
`ResponseCreatedEvent` 类class	响应已创建事件，描述执行过程中的事件。 The provider started producing a reply.
`ResponseDoneEvent` 类class	响应 Done 事件，描述执行过程中的事件。 The provider finished producing a reply.
`SessionEndedEvent` 类class	会话 Ended 事件，描述执行过程中的事件。 The provider session is gone; no further events will arrive.
`SpeechEndedEvent` 类class	语音 Ended 事件，描述执行过程中的事件。 The provider's own VAD detected the user stopping.
`SpeechStartedEvent` 类class	语音已启动事件，描述执行过程中的事件。 The provider's own VAD detected the user starting to speak.
`SpeechTransition` 类class	语音 Transition，处理低延迟的实时语音与多模态交互。 A transition in voice activity, as detected by a VAD. Not an event class; the model-side ``SpeechStartedEvent`` is what a provider reports.
`ToolCallEvent` 类class	工具调用事件，描述执行过程中的事件。 A complete tool call requested by the provider.
`TranscriptDeltaEvent` 类class	Transcript 增量事件，描述执行过程中的事件。 A chunk of the assistant's spoken text.
`TransportBase` 类class	传输层基类，处理低延迟的实时语音与多模态交互。 Connects a :class:`RealtimeAgent` to where audio comes from and goes to.
`VADBase` 类class	语音活动检测基类，处理低延迟的实时语音与多模态交互。 Voice activity detection over a stream of PCM16 mono audio.
`XAIRealtimeModel` 类class	xAI 实时模型，处理低延迟的实时语音与多模态交互。 The Grok voice speech-to-speech API over WebSocket.

属性Attributes

TransportFrame

字段attribute

—

属性详情Attribute Details

TransportFrame

 = AudioFrame | ControlFrame

TransportFrame 字段。

The transport frame field.

源码Source agentscope/realtime/_transport/_base.py:45