@openai/agents-realtime/dist/realtimeSessionEvents.d.ts

The OpenAI Agents SDK is a lightweight yet powerful framework for building multi-agent workflows. This package contains the logic for building realtime voice agents on the server or in the browser.

import { type FunctionTool, OutputGuardrailTripwireTriggered, RunContext, RunToolApprovalItem } from '@openai/agents-core';
import { RealtimeGuardrailMetadata } from './guardrail';
import { RealtimeItem } from './items';
import { RealtimeAgent } from './realtimeAgent';
import { TransportEvent, TransportLayerAudio } from './transportLayerEvents';
import { RealtimeContextData } from './realtimeSession';
import { protocol } from '@openai/agents-core';
type AgentWithOrWithoutHistory<TContext> = RealtimeAgent<TContext> | RealtimeAgent<RealtimeContextData<TContext>>;
export type RealtimeToolApprovalRequest = {
    type: 'function_approval';
    tool: FunctionTool<RealtimeContextData<any>>;
    approvalItem: RunToolApprovalItem;
};
export type RealtimeSessionError = {
    type: 'error';
    error: unknown;
};
export type RealtimeSessionEventTypes<TContext = unknown> = {
    /**
     * Triggered when an agent starts its work on a response.
     */
    agent_start: [
        context: RunContext<RealtimeContextData<TContext>>,
        agent: AgentWithOrWithoutHistory<TContext>
    ];
    /**
     * Triggered when an agent ends its work on a response.
     */
    agent_end: [
        context: RunContext<RealtimeContextData<TContext>>,
        agent: AgentWithOrWithoutHistory<TContext>,
        output: string
    ];
    /**
     * Triggered when an agent hands off to another agent.
     */
    agent_handoff: [
        context: RunContext<RealtimeContextData<TContext>>,
        fromAgent: AgentWithOrWithoutHistory<TContext>,
        toAgent: AgentWithOrWithoutHistory<TContext>
    ];
    /**
     * Triggered when an agent starts a tool call.
     */
    agent_tool_start: [
        context: RunContext<RealtimeContextData<TContext>>,
        agent: AgentWithOrWithoutHistory<TContext>,
        tool: FunctionTool<RealtimeContextData<TContext>>,
        details: {
            toolCall: protocol.ToolCallItem;
        }
    ];
    /**
     * Triggered when an agent ends a tool call.
     */
    agent_tool_end: [
        context: RunContext<RealtimeContextData<TContext>>,
        agent: AgentWithOrWithoutHistory<TContext>,
        tool: FunctionTool<RealtimeContextData<TContext>>,
        result: string,
        details: {
            toolCall: protocol.ToolCallItem;
        }
    ];
    /**
     * Emits all the raw events from the transport layer.
     */
    transport_event: [event: TransportEvent];
    /**
     * Triggered when the agent starts generating audio.
     */
    audio_start: [
        context: RunContext<RealtimeContextData<TContext>>,
        agent: AgentWithOrWithoutHistory<TContext>
    ];
    /**
     * Triggered when the agent stops generating audio.
     */
    audio_stopped: [
        context: RunContext<RealtimeContextData<TContext>>,
        agent: AgentWithOrWithoutHistory<TContext>
    ];
    /**
     * Triggered when there is new audio data available for playing to the user.
     */
    audio: [event: TransportLayerAudio];
    /**
     * Triggered when the agent is interrupted. Can be listened to by the user to stop audio playback
     * or give visual indicators to the user.
     */
    audio_interrupted: [
        context: RunContext<RealtimeContextData<TContext>>,
        agent: AgentWithOrWithoutHistory<TContext>
    ];
    /**
     * Triggered when an output guardrail is tripped.
     */
    guardrail_tripped: [
        context: RunContext<RealtimeContextData<TContext>>,
        agent: AgentWithOrWithoutHistory<TContext>,
        error: OutputGuardrailTripwireTriggered<RealtimeGuardrailMetadata>,
        details: {
            itemId: string;
        }
    ];
    /**
     * Triggered when the history got updated. Contains the full history of the conversation.
     */
    history_updated: [history: RealtimeItem[]];
    /**
     * Triggered when a new item is added to the history. At this point the transcript/response
     * might still be in progress.
     */
    history_added: [item: RealtimeItem];
    /**
     * Triggered when an error occurs.
     */
    error: [error: RealtimeSessionError];
    /**
     * Triggered when a tool approval is requested.
     */
    tool_approval_requested: [
        context: RunContext<RealtimeContextData<TContext>>,
        agent: AgentWithOrWithoutHistory<TContext>,
        approvalRequest: RealtimeToolApprovalRequest
    ];
};
export {};