Centerbeam AI

Nova SDK

Voice Agent SDK for JavaScript, TypeScript, and SwiftUI

The Nova SDK abstracts real-time voice infrastructure so teams can ship hands-free experiences without managing WebSocket protocols, audio encoding, or provider APIs directly.

Quick start

01

Install

npm install @centerbeam/nova

Coming soon — the package is not yet published to npm. Request early access and we'll get you the SDK. Request access →

02

Initialize

const nova = new Nova({ apiKey: process.env.NOVA_API_KEY })
03

Mount and start

await nova.mount("#agent")
await nova.start()

SDK capabilities

WebSocket session management
Real-time audio capture and encoding
Turn-taking and barge-in interruption
Event-driven tool dispatch
Wake word and VAD configuration
Confirm-before-act pattern support
Offline and edge relay support
Multi-language (100 languages)

Frequently Asked Questions

What languages does the Nova SDK support?

Nova SDK supports JavaScript, TypeScript, and SwiftUI (iOS 17+ and macOS 14+). Both web and native Apple platform applications are supported from the same underlying runtime.

What AI providers does Nova SDK support?

Nova SDK supports OpenAI Realtime, Google Gemini Live, and xAI Grok Voice Agent as underlying AI providers. Providers are configured at session start without code changes to the application layer. Capabilities vary by provider — tool calling, file search, and RAG run on the OpenAI provider today. See each provider page for its capability set.

Does the Nova SDK support edge or offline deployment?

Yes. Nova supports edge deployment configurations. The relay server component can be deployed on-prem or at the edge for environments with restricted network access.

Ready to build hands-free voice into your next product?