The business problem
A classification engine that works for one catalog is not yet a service that can operate safely across multiple stores. Each tenant needs independent authorization, configuration, work history, usage records, review state, and controlled write-back.
What DSWD delivered
DSWD built the service layer around the classification capability: Shopify application authorization, store onboarding, tenant and job records, background orchestration, usage tracking, review, and write-back through a stable application interface.
The operator experience
Each store is onboarded separately and retains its own configuration and job history. Work runs in the background, exposes progress and review states, and keeps approved write-back distinct from classification.
Safeguards and operating qualities
- Tenant data and credentials are isolated by design.
- Long-running work is asynchronous and observable.
- Classification and service responsibilities meet at a defined boundary.
- Usage and operating cost can be reviewed per tenant.
- Review and write-back remain explicit actions.
Project-specific technical choices
For this project, we used a multi-tenant service architecture and managed background processing because authorization, job state, and review need to remain dependable across stores while the underlying classification capability evolves independently.
Evidence and claim boundary
The service is deployed in production and used multiple times daily. This page describes the operating capabilities in use without publishing tenant information or claiming a packaged, self-service product offering.