SmartRealtimeAdapter

Processes streaming, real-time data and conversation through the OpenAI realtime endpoint.

Attaches to Realtime stream

Built for the OpenAIEndpointRealtime endpoint, this adapter handles low-latency, streaming interaction — live audio, continuous input, immediate responses — rather than the request-and-wait pattern of the other adapters.

It underpins voice-first and live-assistant experiences where waiting for a full response would break the interaction.

Sessions are stateful and bidirectional: the adapter keeps the stream open, forwards input as it arrives and surfaces model output the moment tokens are available.

What it gives you

Streaming, low-latency processing.
Designed for the OpenAIEndpointRealtime endpoint.
Enables live voice and continuous interaction.

Use it for

Live voice assistants
Continuous transcription displays
Latency-sensitive conversational UI

Like every adapter, it runs through a SmartHub and inherits its SmartEndpoint — switching the model or provider never changes this code.

csharp

var hub = new SmartHub(this) {
    Endpoint = new OpenAIEndpointRealtime()
};

var realtime = new SmartRealtimeAdapter();
await realtime.StartAsync();   // begin a live session

Read the SmartRealtimeAdapter docs ↗