RRamainDocs
Infrastructure

Data Flow

How a workflow run moves through the system.

Workflow creation:

  1. User opens app from CloudFront, authenticates with Cognito (JWT).
  2. Opens WebSocket to API via ALB.
  3. Backend verifies JWT, loads user context from RDS, accepts connection.
  4. Backend requests Bedrock AgentCore Browser session, connects via CDP.
  5. User chats → Claude on Bedrock executes browser actions → results stream to user.
  6. All actions persisted to RDS.
  7. User saves → Claude compiles execution trace into reusable workflow graph → stored in RDS.

Workflow replay:

Saved workflows replay the same path with parameterized inputs.

Network: Backend → AWS APIs (Bedrock, AgentCore, S3, Secrets, CloudWatch) via VPC endpoints. External calls via NAT Gateway.

Components

Every AWS service used in the deployment, organized by layer, with the configured specification for each.

Security Posture

Encryption, network isolation, IAM, auditability, and data residency for the deployment.