Defined in: realtime/types.ts:30

Configuration for a realtime session

optional instructions: string;

Defined in: realtime/types.ts:36

System instructions for the assistant

optional maxOutputTokens: number | "inf";

Defined in: realtime/types.ts:48

Maximum number of tokens in a response

optional model: string;

Defined in: realtime/types.ts:32

Model to use for the session

optional outputModalities: ("text" | "audio")[];

Defined in: realtime/types.ts:44

Output modalities for responses (e.g., ['audio', 'text'], ['text'])

optional providerOptions: Record<string, any>;

Defined in: realtime/types.ts:52

Provider-specific options

optional semanticEagerness: "low" | "high" | "medium";

Defined in: realtime/types.ts:50

Eagerness level for semantic VAD ('low', 'medium', 'high')

optional temperature: number;

Defined in: realtime/types.ts:46

Temperature for generation (provider-specific range, e.g., 0.6-1.2 for OpenAI)

optional tools: RealtimeToolConfig[];

Defined in: realtime/types.ts:38

Tools available in the session

optional vadConfig: VADConfig;

Defined in: realtime/types.ts:42

VAD configuration

optional vadMode: "server" | "manual" | "semantic";

Defined in: realtime/types.ts:40

VAD mode

optional voice: string;

Defined in: realtime/types.ts:34

Voice to use for audio output