# Platform Vehicle Data Loader Refresh

## Context
- Reintroduced the normalized `vehicles` schema (make/model/model_year/trim/engine + bridges) with a new migration under `backend/src/features/platform/migrations/001_create_vehicle_lookup_schema.sql`.
- Added `scripts/load_vehicle_data.py`, a stand-alone Python loader that hydrates the schema from `data/make-models/*.json`.
- Loader truncates lookup tables, seeds default transmissions (Automatic, Manual), and inserts deduplicated year → make → model → trim → engine combinations.

## Follow-up Tasks (Backend API Team)
1. **Wire dropdown API to refreshed data**
   - Run `make migrate` (or `npm run migrate:all` inside backend container) to ensure the new schema exists.
   - Execute the loader (see command below) so Postgres has the latest lookup entries.
   - Verify both the platform dropdown queries and the admin catalog APIs surface the same data set (admin APIs now read directly from `vehicles.*` tables).
2. **Add Makefile wrapper**
   - Create a `make load-vehicle-data` task that shells into the backend container, installs `psycopg` if needed, and invokes `python3 scripts/load_vehicle_data.py` with the correct DB credentials and data directory.

### Loader Command Reference
```
PGPASSWORD=$(cat secrets/app/postgres-password.txt) \
python3 scripts/load_vehicle_data.py \
  --db-host 127.0.0.1 \
  --db-port 5432 \
  --db-user postgres \
  --db-name motovaultpro \
  --data-dir data/make-models
```

> Run the command from the repository root (outside of containers) while `mvp-postgres` is up. Adjust host/port if executing inside a container.

## 2025-11-07 Update
- Admin catalog CRUD endpoints have been refactored to read/write the normalized `vehicles.*` tables instead of the legacy `vehicle_dropdown_cache`.
- Platform cache invalidation now happens immediately after each catalog mutation so both the admin UI and the user dropdown APIs stay in sync.