Realtime audio - Infistar Docs

Realtime audio uses a WebSocket connection for low-latency audio input and output. Confirm that your key and group have access to a Realtime model before connecting. Use wss://infistar.ai/v1/realtime?model=MODEL_ID internationally or wss://infistar.cc/v1/realtime?model=MODEL_ID in mainland China. Replace MODEL_ID with the exact available model ID.

import WebSocket from 'ws';

const socket = new WebSocket(
  'wss://infistar.ai/v1/realtime?model=MODEL_ID',
  { headers: { Authorization: `Bearer ${process.env.INFISTAR_API_KEY}` } },
);
socket.on('open', () => {
  socket.send(JSON.stringify({
    type: 'session.update',
    session: { modalities: ['text', 'audio'] },
  }));
});
socket.on('message', message => console.log(JSON.parse(message.toString())));

This Node.js example uses ws to set the authorization header. A browser’s native WebSocket constructor cannot directly set that header. Browser applications should authenticate and relay through their own server; do not expose the key in frontend code or an application package. For 401, check the key and Bearer header. For an unavailable model, recheck the group and model list. For connection failures, check wss:// and ensure /v1 appears only once.