Dev (Vite dev server / e2e mock mode): falls back to 'http://localhost:5179' so
Playwright route interception on http://localhost:5179/** keeps working.
Production (npm run build -> PROD=true): falls back to '/api' so nginx proxy
strips the prefix without any hostname baked in at build time.
VITE_API_BASE_URL (set by Aspire in dev, injected LAN IP) always wins over both.
Evidence: npm run e2e 48/48 PASS (mock mode, both viewports).
Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
- GerbilManagerWebAPI/Dockerfile: net7->net10 multi-stage; build context=repo
root (includes ServiceDefaults); port 8080 (.NET 10 container default)
- gerbil-manager-web/Dockerfile: drop VITE_API_BASE_URL build-arg; nginx proxies
/api/* to api:8080 (strip prefix), /scalar, /openapi/ -- one published port 80
- gerbil-manager-web/src/api/client.ts: default API_BASE_URL /api (relative),
no hostname coupling at build time; Aspire still injects absolute LAN URL in dev
- GerbilManagerWebAPI/Program.cs: run EF Migrate() at startup always (not Dev-only)
- GerbilManager.ServiceDefaults/Extensions.cs: expose /health + /alive in all
environments (compose healthcheck requires them in Production)
Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
- Remove gerbil-manager-nextjs (unused scaffold; recoverable from history)
- Scaffold gerbil-manager-web: Vite + React 19 + TypeScript
- German-only UI (lang=de, central strings in src/strings/de.ts)
- Mobile-first shell: bottom tab bar on phones, sidebar on >=768px
- react-router with skeleton routes (Start, Rennmaeuse, Wuerfe, Zuechter)
- API client (src/api/client.ts) pointing at GerbilManagerWebAPI,
configurable via VITE_API_BASE_URL (default http://localhost:5179)
- nginx-based Dockerfile with SPA fallback; docker-compose frontend
service now builds gerbil-manager-web
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
