Installation

Prerequisites

• Node.js ≥ 20
• PostgreSQL database (local or hosted)
• Google credentials — either a Vertex AI service account or a Gemini API key (see Gemini auth modes)

Quick Start

cd next-app
cp .env.example .env      # fill in values (see below)
npm install
npm run prisma:deploy     # create/migrate database schema
npm run prisma:seed       # load demo user + example pipeline workflow
npm run dev -- --port 3006

App runs at http://localhost:3006. The seed creates a login: demo@tailormade.local / demo1234.

Environment Variables

All variables live in next-app/.env (gitignored). Copy from next-app/.env.example.

Database

Variable	Example	Purpose
DATABASE_URL	postgresql://postgres:password@localhost:5432/tailormade	Prisma connection string

Auth

Variable	Example	Purpose
JWT_SECRET_KEY	(random string)	HS256 signing key for access + refresh tokens
APP_URL	https://tailormade.global-invent-ai.com	Public base URL — used to build the Entra callback URI
ENTRA_TENANT_ID	(Azure AD tenant ID)	Microsoft Entra directory ID
ENTRA_CLIENT_ID	(app registration client ID)	Entra OAuth2 client
ENTRA_CLIENT_SECRET	(app registration secret)	Entra OAuth2 secret
ALLOW_USER_REGISTRATION	true	Enable password-based sign-up
ALLOWED_REGISTRATION_EMAIL_DOMAINS	capgemini.com	Comma-separated domain allowlist (empty = allow all)

Gemini Auth Modes

One switch controls which Google auth path the app uses:

GOOGLE_GENAI_USE_VERTEXAI=true    # canonical — Vertex AI via ADC / service account
# or
GOOGLE_GENAI_USE_VERTEXAI=false   # alternative — public Gemini API key
GEMINI_API_KEY=AIza...

Vertex AI (canonical) — requires a Google Cloud project with roles/aiplatform.user:

GOOGLE_GENAI_USE_VERTEXAI=true
GOOGLE_CLOUD_PROJECT=premi1078276-inventglobalai
GOOGLE_CLOUD_LOCATION=global

Local dev authenticates via gcloud auth application-default login. Production uses a service account key passed as GOOGLE_APPLICATION_CREDENTIALS_B64.

API key (alternative) — no Google Cloud project needed, but Veo video and Nano Banana Pro image model are unavailable:

GOOGLE_GENAI_USE_VERTEXAI=false
GEMINI_API_KEY=AIza...
GEMINI_IMAGE_MODEL=gemini-2.5-flash-image

AI Models

Variable	Default	Purpose
GEMINI_TEXT_MODEL	gemini-2.5-flash	Text generation, copy, alt-text
GEMINI_IMAGE_MODEL	gemini-3-pro-image-preview	Image generation (Nano Banana Pro)
PIPELINE_IMAGE_MODEL	(inherits GEMINI_IMAGE_MODEL)	Image Pipeline override
VEO_MODEL	veo-3.1-fast-generate-001	Video generation
VEO_LOCATION	us-central1	Veo region (always US)

Storage & CORS

Variable	Default	Purpose
UPLOAD_DIR	/uploads	Filesystem path for uploaded and generated images
CORS_ALLOWED_ORIGINS	(empty)	Comma-separated extra CORS origins

Database Setup

cd next-app
npm run prisma:deploy   # apply migrations (safe to re-run)
npm run prisma:seed     # demo superuser + example pipeline workflow

The seed is idempotent — re-running it is safe.

Local Dev — Vertex AI Auth

# one-time: write Application Default Credentials
gcloud auth application-default login
gcloud auth application-default set-quota-project premi1078276-inventglobalai

Tooling

cd next-app
npm run dev -- --port 3006   # dev server
npm run build                 # production build (lint skipped by design)
npm run typecheck             # tsc --noEmit
npm run test                  # vitest (141 tests)
npm run lint                  # eslint (0 errors, ~146 legacy warnings)