Interface GPTLiveModelOptions

Options for the OpenAI GPT-Live full-duplex voice model.

interface GPTLiveModelOptions {
    apiKey?: string;
    baseURL?: string;
    connOptions?: APIConnectOptions;
    delegation?: DelegationTarget;
    maxSessionDuration?: null | number;
    model?: string;
    responsesOptions?: ResponsesDelegationOptions;
    voice?: Record<string, unknown> | GPTLiveVoices | string & {};
}

Properties

apiKey? baseURL? connOptions? delegation? maxSessionDuration? model? responsesOptions? voice?

Properties

Optional apiKey

apiKey?: string

Falls back to OPENAI_API_KEY.

Optional baseURL

baseURL?: string

Falls back to OPENAI_BASE_URL, then https://api.openai.com/v1.

Optional connOptions

connOptions?: APIConnectOptions

Connection/startup timeout and retry limits. Defaults to DEFAULT_API_CONNECT_OPTIONS.

Optional delegation

delegation?: DelegationTarget

Fixed at startup. Client delegation requires an agent with no tools.

Optional maxSessionDuration

maxSessionDuration?: null | number

Recycle the connection after this many milliseconds. Null (the default) disables the timer.

Optional model

model?: string

Voice model identifier. Defaults to gpt-live-1.

Optional responsesOptions

responsesOptions?: ResponsesDelegationOptions

Backend configuration when delegation is responses (the default).

Optional voice

voice?: Record<string, unknown> | GPTLiveVoices | string & {}

A named voice (marin by default), or an authorized custom voice object. Fixed at startup. Names pass through unchanged.

Settings

Member Visibility

Theme

On This Page

apiKeybaseURLconnOptionsdelegationmaxSessionDurationmodelresponsesOptionsvoice