AWS has introduced a complete, cloud-deployable solution for building voice AI agents using Amazon Nova Sonic, a unified speech-to-speech foundation model in Amazon Bedrock.
The open-source asset leverages AWS CDK to orchestrate a scalable stack, including WebSockets, Cognito authentication, ECS/Fargate compute, DynamoDB storage, and Bedrock Knowledge Bases, for managing conversational sessions. This architecture enables real-time, human-like voice conversations, context retention, function/tool integration via the Model Context Protocol, and knowledge-aware responses.
Ideal for use cases like AI call centers, this approach streamlines deployment without separate speech‑recognition or TTS components, reducing complexity while delivering low-latency, expressive, fully agentic voice experiences on AWS.