Cerebras and LiveKit

Try Cerebras AI

Experience Cerebras's fast inference in a LiveKit-powered voice AI playground

Cerebras ecosystem support

Cerebras provides high-throughput, low-latency AI inference for open models like Llama and DeepSeek. LiveKit Agents has full support for Cerebras inference via the OpenAI plugin, as Cerebras is an OpenAI-compatible LLM provider.

Getting started

Use the Voice AI quickstart to build a voice AI app with Cerebras. Select an STT-LLM-TTS pipeline model type and add the following components to build on Cerebras.

Install the OpenAI plugin:

pip install "livekit-agents[openai]~=1.0rc"

Add your Cerebras API key to your .env file:

CEREBRAS_API_KEY=<your-cerebras-api-key>

Use the Cerebras LLM to initialize your AgentSession:

from livekit.plugins import openai

# ...

# in your entrypoint function
session = AgentSession(
    llm=openai.LLM.with_cerebras(
        model="llama-3.3-70b",
    ),
)

For a full list of supported models, including DeepSeek, see the Cerebras docs.

LiveKit Agents overview

LiveKit Agents is an open-source framework for building realtime AI apps using WebRTC transport to end-user devices and WebSockets or HTTPS for backend services.

• Agent workflows: Build complex voice AI apps with discrete stages and handoffs.
• Telephony: Inbound and outbound calling using SIP trunks.
• Frontend SDKs: Full-featured SDKs and UI components for JavaScript, Swift, Kotlin, Flutter, React Native, and Unity.
• Python and Node.js: Build voice AI apps in Python or Node.js.
• Dispatch and load balancing: Built-in support for request distribution and load balancing.
• LiveKit Cloud: Fully-managed LiveKit server with global scale and low latency (you can also self-host).

Further reading

More information about integrating Llama is available in the following article: