AI 原生开放教材

外观

OpenAI API 参考（字段级）官方文档·英文原文 ↗原件 ↓仓库 ↗

Fork WebSocket ​

For the complete documentation index, see llms.txt. Markdown versions of documentation pages are available by appending .md to the page URL.

Start a new Live session from stored conversation state. Send and receive audio and control events on the new WebSocket connection.

WS /v1/live/sessions/\{session_id\}/fork

Connection ​

wss://api.openai.com/v1/live/sessions/\{session_id\}/fork

Authenticate from your backend with your OpenAI API key in the Authorization: Bearer $OPENAI_API_KEY header. Keep the key on your server.

session_id (required path parameter): The ID of the stored source session to fork. The fork receives a new session ID.

No query parameters. After connecting, send session.start with a session overrides object. An empty object inherits the stored configuration. Do not supply a new model. Wait for session.started before sending audio or other commands.

Inputs ​

Required once after connecting. Omitted settings are inherited, including store. You may override Responses delegation settings, storage, and the new WebSocket audio format. Frontend client permissions apply only to WebRTC forks.

First message: session.start ​

json
{
  "type": "session.start",
  "session": {}
}

All client events

Outputs ​

The server confirms the new session is ready. Use its new ID for subsequent sideband connections and session controls.

Fork ready · excerpt: session.started ​

json
{
  "type": "session.started",
  "event_id": "event_started_1",
  "session": {
    "id": "live_fork_123",
    "expires_at": 1788307200,
    "status": "active",
    "model": "gpt-live-1"
  }
}

All server events

Learn how to store a session and fork its conversation.

Client events ​

session.start ​

Start a Live session after connecting to a stored session’s fork WebSocket. Send an empty session object to use the stored configuration.

本篇目录 ​