# Port Plan

## Goals

- Replace the custom FastAPI auth flow with Django sessions and `AUTH_USER_MODEL`
- Collapse the two legacy SQLite files into one relational schema owned by Django migrations
- Move domain logic out of route files into app models and services
- Replace the React monolith with a SvelteKit frontend structured by route and domain
- Prepare for `shadcn-svelte` instead of continuing with bespoke one-off UI patterns

## Target backend split

- `accounts`: user, role, permission, device, business-access concerns
- `core`: business, product, vendor, category master data
- `operations`: invoices, inventory, events, scheduling
- `reporting`: pre-aggregated dashboard tables and future ETL landing points
- `notifications`: reminders and inbox state
- `api`: thin JSON layer for the Svelte frontend

## Legacy migration strategy

1. Create Django migrations from the new model set.
2. Migrate the new database.
3. Run `import_legacy_data` to ingest the bundled legacy SQLite files in `data/legacy/`.
4. Rebuild auth credentials by forcing password resets for imported users.
5. Move ETL scripts into Django management commands later instead of repo-root scripts.

## Frontend direction

- SvelteKit routes mirror the existing user navigation
- app shell owns layout, not each page
- each large legacy page becomes a route plus smaller components
- fetch is centralized in `$lib/api/client.ts`
- visual primitives are compatible with `shadcn-svelte` conventions

## Known gaps

- Migrations were not generated, by design, because nothing heavy was run
- API coverage is partial; the high-risk domains are modeled first
- Scheduling and event recurrence still need dedicated service extraction
- ETL jobs have not been rewritten yet