web Later¶

Status¶

Accepted

Context¶

The target architecture eventually envisions universaladmin-core, universaladmin-api, universaladmin-sdk, and universaladmin-web as separate artifacts. But today there is no external extension and no web app that would compile against -api/-sdk, and no web process that would need -web.

Decision¶

A single Gradle project (universaladmin-core, the repo root) for now. No multi-project scaffold with empty modules. The future module boundary is instead marked in code:

The interfaces a future extension will need (Module, Action, GuiPage, PermissionRegistry, AuditService, ...) already live in their own, focused packages rather than being scattered - the later cut "these packages move to universaladmin-api" is thus a mechanical step, not a redesign.
Storage/threading/config are already cut so a web layer could call the same services without changing the core (see ../web-future.md).

When the actual multi-project split happens is in ROADMAP.md (Phase 4 for -api/-sdk, Phase 6 for -web) and will be decided at the point a module actually has real content.

Consequences¶

Less Gradle overhead and configuration now.
The split isn't "free" when it happens - it brings real work then (versioning, publishing, possibly separate repos for -sdk examples). That's deliberately accepted rather than front-loading that work today with no users yet.
Until the split, 0005-extension-ready-design.md still applies: the code has to be written so the split stays feasible.

Alternatives¶

Multi-project from the start (empty api/sdk/web modules): rejected - pure structure with no content makes the repository harder to navigate, with nothing using it today. Can be added at any time once a module actually gets real code.