Migrate from the OpenAI SDK

If your app already uses the OpenAI Realtime API, you change three values and keep the rest of your code. 60db accepts the OpenAI GA protocol, so the same events, session fields and client libraries work.

What changes

Setting	OpenAI	60db
Base URL	`https://api.openai.com/v1` or `wss://api.openai.com/v1/realtime`	`https://api.60db.ai/v1` or `wss://api.60db.ai/v1/realtime`
Model	`gpt-realtime`	`60db-realtime-v1-mini` (`gpt-realtime` also maps to it)
Key	`OPENAI_API_KEY` (`sk-...`)	60db API key (`sk_...`)
Voice	OpenAI voices (`alloy`, `marin`, …)	60db voice IDs. OpenAI voice names map to 60db stock voices

Only the three values above change. Event names, session shape, tool definitions and response.done usage follow the OpenAI GA spec.

Migrate

Replace the key and base URL

export SIXTYDB_API_KEY=sk_your_api_key

Use SIXTYDB_API_KEY in place of OPENAI_API_KEY, and set the base URL below in your client.

Change the model and voice

Set the model to 60db-realtime-v1-mini. Set audio.output.voice to a 60db voice ID. Use List voices to find one.

Test a session

Run one text turn and one audio turn. Check that response.done returns usage and that audio plays.

Client examples

@openai/agents-realtime
openai (Node)
openai (Python)

import {
  RealtimeAgent,
  RealtimeSession,
  OpenAIRealtimeWebSocket,
} from '@openai/agents-realtime';

const agent = new RealtimeAgent({
  name: 'Assistant',
  instructions: 'You are a friendly voice assistant.',
});

const transport = new OpenAIRealtimeWebSocket({
  url: 'wss://api.60db.ai/v1/realtime',
});

const session = new RealtimeSession(agent, {
  transport,
  model: '60db-realtime-v1-mini',
  config: {
    outputModalities: ['audio'],
    audio: { output: { voice: 'YOUR_VOICE_ID' } },
  },
});

// Server-side: use an sk_ key. In the browser, use an ek_ key (see below).
await session.connect({ apiKey: process.env.SIXTYDB_API_KEY });

import OpenAI from 'openai';
import { OpenAIRealtimeWS } from 'openai/realtime/ws';

const client = new OpenAI({
  apiKey: process.env.SIXTYDB_API_KEY,
  baseURL: 'https://api.60db.ai/v1',
});

const rt = await OpenAIRealtimeWS.create(client, { model: '60db-realtime-v1-mini' });
rt.send({
  type: 'session.update',
  session: {
    type: 'realtime',
    audio: { output: { voice: 'YOUR_VOICE_ID' } },
  },
});

import os
from openai import AsyncOpenAI

client = AsyncOpenAI(
    api_key=os.environ["SIXTYDB_API_KEY"],
    base_url="https://api.60db.ai/v1",
)

async with client.realtime.connect(model="60db-realtime-v1-mini") as conn:
    await conn.session.update(session={
        "type": "realtime",
        "audio": {"output": {"voice": "YOUR_VOICE_ID"}},
    })

Not supported yet

Code that depends on these features needs a change or must wait:

WebRTC transport
SIP calling (call_id)
Out-of-band responses (conversation: "none")
Image input
MCP tools inside a session

See the full list on the Realtime overview.

60db extensions

The memory, knowledge and intent session fields are 60db extensions. OpenAI ignores them, so they are safe to send to either service. Add them when you want Smart Memory, Collections or filler speech. See 60db extensions.

Browser ephemeral keys for web apps
Function tools work unchanged
Quickstart

Get Started

Guides

Reference

Migrate from the OpenAI SDK

What changes

Migrate

Client examples

Not supported yet

60db extensions

​What changes

​Migrate

​Client examples

​Not supported yet

​60db extensions

​Related

What changes

Migrate

Client examples

Not supported yet

60db extensions

Related