Build a voice travel concierge with Amazon Bedrock AgentCore, Managed Knowledge Base and Nova Sonic

AWS has published a walkthrough for adding a voice travel concierge to an airline application, built on Amazon Bedrock AgentCore, Amazon Nova 2.5 Sonic, and Amazon Bedrock Knowledge Bases. The concierge is designed to run alongside existing app screens rather than replace them, letting travelers move between tapping and talking in the same session, according to the post.

The solution separates the front end, AI agent, and backend into distinct layers. Amazon Cognito handles authentication and issues temporary AWS credentials; Amazon Bedrock AgentCore runtime hosts the agent with microVM isolation per session; AgentCore Gateway exposes backend endpoints as discoverable MCP tools; API Gateway publishes IAM-authorized REST endpoints; Lambda runs business logic for itineraries, seat maps, passenger updates, flight status, loyalty, policy lookups, and escalation; DynamoDB stores profiles, bookings, passengers, seat maps, purchase history, preferences, conversations, and flight data. Amazon SES sends email notifications, and AWS Amplify hosts the React front end.

Deployment uses AWS CDK across multiple stacks. The build uploads agent source to Amazon S3, triggers AWS CodeBuild to produce an ARM64 Docker image stored in Amazon ECR, and runs on AgentCore runtime with WebSocket support using the Strands Agents framework and Nova 2.5 Sonic.

For policy questions, the Gateway queries Amazon Bedrock Knowledge Bases through an MCP connector; the knowledge base returns policy passages as citations. The repository ships sample policy documents covering baggage, cancellation and refunds, change and rebooking, fare class rules, loyalty terms, pet travel, special assistance, and upgrades. The agent answers using grounded retrieval through Amazon Bedrock Knowledge Bases.

For live agent escalation, Lambda logs the escalation in DynamoDB and returns a reference number with an estimated wait time; the Amplify front end places the call. The post notes the sample does not include a contact center application, so connecting to a person depends on the configured support number.

The post states Nova 2.5 Sonic provides speech recognition across accents, background-noise robustness, bidirectional streaming, asynchronous tool calling, latency masking, barge-in, and multi-turn context. It also notes the solution's confirm-before-write pattern and knowledge base citations, and recommends Amazon Bedrock Guardrails for production deployments to filter prompt-injection attempts and validate response grounding.