AWS AI and LiveKit

AWS AI ecosystem support

Amazon's AWS AI is a comprehensive collection of production-ready AI services, which integrate with LiveKit in the following ways:

• Amazon Bedrock: Access to foundation models from leading AI companies.
• Amazon Polly: Text-to-speech service with lifelike voices.
• Amazon Transcribe: Speech-to-text service with high accuracy.
• Amazon Nova Sonic: Realtime, speech-to-speech model that uses a bidirectional streaming API for streaming events.

The LiveKit Agents AWS plugin supports these services for building voice AI applications.

Getting started

Use the voice AI quickstart to build a voice AI app with AWS services. Select a pipeline model type and add the following components to use AWS AI services:

Install the AWS plugin:

pip install "livekit-agents[aws]~=1.2"

Add your AWS credentials to your .env file:

AWS_ACCESS_KEY_ID=<your-aws-access-key-id>
AWS_SECRET_ACCESS_KEY=<your-aws-secret-access-key>
AWS_REGION=<your-aws-region>

Use the AWS services in your application:

from livekit.plugins import aws

# ...

# in your entrypoint function
session = AgentSession(
    llm=aws.LLM(
        model="anthropic.claude-3-5-sonnet-20240620-v1:0",
    ),
    tts=aws.TTS(
        voice="Ruth",
        speech_engine="generative",
        language="en-US",
    ),
    stt=aws.STT(
        session_id="my-session-id",
        language="en-US",
    ),
    # ... vad, turn_detection, etc.
)

Or use Amazon Nova Sonic, a state of the art speech-to-speech model:

pip install "livekit-agents-aws[realtime]~=1.2"

from livekit.plugins import aws

# ...

# in your entrypoint function
session = AgentSession(
    llm=aws.realtime.RealtimeModel()
)

AWS plugin documentation