Skip to content

AI-Driven Development Model

An AI-driven full-stack development model for building modern enterprise web applications: React + shadcn/ui on the frontend, SAP CAP as the Domain-Driven backend — orchestrated through an agentic, specification-driven AI workflow.

AI-Driven Development Model

Core idea

Technology does not drive the approach — the domain and an AI-assisted workflow do. SAP CAP is the central abstraction for domain model, services, persistence, and authorization. Backend and frontend are the materialization of a single, clean domain model.

Rendering technology — SSR out of scope

Alongside the frontend, we also evaluated whether rendering technologies such as Next.js for server-side rendering (SSR) would provide an advantage. Both variants were implemented, but the runtime cost (a Node.js server vs. a static file system) is out of all proportion to the benefit — only AI-generative renderings could have profited from it.

Since UDINA relies on SAP BTP for operations, our technology decision focuses on that platform. Deployment platforms like Vercel are therefore out of scope.

Principles

  • Domain first — Entities, Aggregates, and Bounded Contexts modeled via CDS namespaces before a single line of UI.
  • AI-driven — agents assist from specification through domain model, UI, tests, and deployment.
  • Type-safe end to end — continuous TypeScript type safety from the CAP service (cds-typer) into the React component.
  • i18n from day one — every app is localizable.
  • Guardrails, not free flight — a single source of truth for stack decisions keeps AI-generated code maintainable and reviewable.

Single source of truth

The binding stack decisions for all new projects live in the repository's ARCHITECTURE.md. If reality diverges (newer majors, deprecations), it is flagged there — agents and developers must not silently deviate.

Repository Layout

An npm workspaces monorepo:

packages/*     — frontend apps / shared libraries
srv/           — CAP service layer
db/            — CDS domain model & persistence
main.tsx       — app entry (root)

Building Blocks

Building blockContent
ArchitectureLayers, repository layout, runtime, dependency policy, deployment
AI-Driven WorkflowAI layer, agents, spec-driven development, guardrails
Hyper AbstractionOne domain model, one annotation layer, many rendering targets (Fiori, Sales & Service Cloud v2, Salesforce)
BackendDomain-Driven CAP and AI-driven ABAP (RAP via ARC-1)
FrontendReact & shadcn/ui, or UI5 Web Components for Fiori compliance
UDINA SpacesIntent-driven, metadata-driven portal: spaces compose intents, generic React pages render from CAP annotations + UDINA metadata, driven by click, AI chat, or CAP MCP