Principle:Openai Openai node Realtime Session Configuration
| Knowledge Sources | |
|---|---|
| Domains | Realtime, Configuration |
| Last Updated | 2026-02-15 00:00 GMT |
Overview
A principle for configuring a Realtime API session with modalities, voice settings, instructions, and tool definitions after establishing a WebSocket connection.
Description
Realtime Session Configuration customizes the behavior of an active Realtime session. After connection, the session has default settings. The client sends a session.update event to configure: supported modalities (text, audio, or both), voice selection, system instructions, input/output audio formats, turn detection settings, and available tools.
The server acknowledges configuration with a session.updated event. Configuration can be updated at any time during the session.
Usage
Use this principle immediately after establishing a Realtime connection to set session parameters. Can also be used mid-session to change behavior (e.g., switching voice or adding tools).
Theoretical Basis
Session configuration follows a Configure After Connect pattern:
// 1. Connection established (session.created event received)
// 2. Send session configuration
rt.send({
type: 'session.update',
session: {
modalities: ['text', 'audio'],
instructions: 'You are a helpful assistant.',
voice: 'alloy',
input_audio_format: 'pcm16',
output_audio_format: 'pcm16',
turn_detection: { type: 'server_vad' },
tools: [...],
},
})
// 3. Server confirms: session.updated event