Backend (api/routes.py) — replace sample data with live reads/writes:
- GET /overview reads production.temper_tracking (done carts, plant-scoped,
windowed, newest-first, one row per ttId); each order's 5-station belt is
derived from planned timestamps vs NOW(), faithful to fastpress
_temper_aggregates_all. sample:false.
- GET /reasons reads settings.reject_reasons (active=1).
- POST /reject WRITES production.rejects keyed by the ORDER (rejOrdernumber +
rejPress, press taken from the body) — an un-scoped port of PressV
send_reject; no rejTtId column needed (ttId is context-only). Failures
return HTTP 200 + {success:false}, NOT 4xx/5xx.
Belt: the single pre-oven buffer (Wagenzeit) shows under Unterwegs; Wagen=0.
Frontend — never report an unconfirmed booking as success:
- store.submitReject returns the gateway's own {success} and reports failure on
a transport error instead of faking success.
- store.refresh keeps the last REAL data on a transient poll error (only a cold
start falls back to demo carts) so a live terminal never shows fabricated carts.
- the failure toast is now red (was always green).
Verified live against cdbserv15 (12 carts, 25 reasons, INSERT round-trip).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
