Food Scanner Production Flow

End-to-end architecture for static delivery, Cognito sign-in, protected Lambda APIs, OpenAI-powered analysis, and DynamoDB-backed scan history.

Delivery and Frontend

User BrowserLoads the app and submits meal photos

→

CloudFront CDNapp.joesparkman.com edge delivery

→

S3 Static HostingNext export assets and HTML pages

Auth and API Routing

Amazon CognitoHosted sign-in and JWT tokens

→

API Gateway/health, /config, /analyze, /scans

→

Lambda (Python)Routing, auth, analysis, and persistence

Analysis and Persistence

OpenAI APIImage analysis and recipe estimates

↔

Lambda RuntimePrompting, fallback logic, response shaping

→

DynamoDBPer-user scan history storage

Protected endpoints require Cognito bearer tokens. Cloud scan records are keyed by the signed-in user so scan history stays private.

Hosting: CloudFront serves the app globally while S3 stores the exported frontend assets.

Auth: Cognito handles sign-in and token issuance for protected save and list routes.

Compute: API Gateway routes requests to Lambda for analysis and persistence operations.

Data: DynamoDB stores per-user scan history and correction updates.

AI: OpenAI returns a meal estimate and tips, with a fallback when the model path is unavailable.

Security: Credentials stay server-side and the frontend only sends user tokens and photo data.