Member Junction
    Preparing search index...

    Interface RealtimeSessionTuningConfig

    Provider-session tuning for a Realtime co-agent — the config-cascade source for the driver feature keys (GetSessionTuningSettings projects these onto the flat bag keys the drivers consume). Every field is optional; absent fields contribute NOTHING to the bag.

    interface RealtimeSessionTuningConfig {
        effortLevel?: string | number;
        inputTranscriptionModel?: string;
        mcpTools?: JSONObjectLike[];
        parallelToolCalls?: boolean;
    }
    Index

    Properties

    effortLevel? inputTranscriptionModel? mcpTools? parallelToolCalls?

    Properties

    effortLevel?: string | number

    MJ-normalized effort level (the ChatParams.effortLevel vocabulary: numeric 1–100 or a named level). Providers with reasoning realtime models (OpenAI gpt-realtime-2.x) map it to their native literals; others scrub it.

    inputTranscriptionModel?: string

    Per-session input-transcription model override (OpenAI-protocol providers).

    mcpTools?: JSONObjectLike[]

    Remote MCP server tool declarations (provider-native { type:'mcp', server_label, server_url|connector_id, … } objects) for providers that support MCP in session.tools. Until an approval UX exists, declare servers with require_approval: 'never' — an approval request is auto-denied by the driver.

    parallelToolCalls?: boolean

    Whether the model may run tool calls in parallel (GA reasoning realtime models).

    Settings

    Member Visibility

    On This Page

    Properties