Section of the AI Wrapper App
The first-time setup wizard for a Lumina-style AI wrapper app. A premium Apple-inspired 3-step flow with a persistent left branding panel (Lumina wordmark, tagline, dot step indicator) and a right content panel that changes per step. Step 1 connects AI models (Flux Pro, DALL-E 3, SDXL) with API key inputs. Step 2 picks a default style preset. Step 3 is a confirmation screen with a large icon and a "You are all set" message. Get started button in blue.
Section preview
Paste the prompt into your AI coding tool and it builds just this section. Each version is scoped to the Onboarding Section, part of the AI Wrapper App.
Build the Onboarding wizard for a Lumina AI image generation wrapper app in this repository. Design system: white background (#FFFFFF), black text, grey (#F5F5F5, #EBEBEB) surfaces, blue #2B65F6 only on the primary CTA button, sans-serif font, font-medium max weight, 12-14px body, 16-20px headings, rounded-xl / rounded-2xl cards, no all-caps text. Build: a 3-step premium first-run setup wizard. Layout: persistent left panel (320px, grey #F5F5F5 bg) with the Lumina L logo in a black rounded square, headline "Your AI creative studio." at 22px font-medium, a tagline, and a dot/dash step indicator at the bottom showing current step. Right content panel changes per step. Step 1 ("Connect your first model"): 3 model cards (Flux Pro, DALL-E 3, SDXL) each with a gradient icon, name, description, and a password API key input. Step 2 ("Choose your default style"): 2x2 grid of style cards (Cinematic, Photorealistic, Illustration, Watercolor) that toggle a black border on selection. Step 3 (confirmation screen): centered large gradient icon, "You are all set." at 22px, short message. Navigation: Skip / Back (text links) and Continue / Open Studio button in blue #2B65F6.. Use Next.js 14 App Router with Tailwind CSS. No motion libraries. Run npx tsc --noEmit when done.More sections you can vibe code alongside the Onboarding Section. Each ships with prompts for Claude Code, Cursor, Codex, Lovable, Replit, and Google Stitch.
Prompt History Section
AI wrapper app prompt history section for Claude Code and Cursor: searchable list of past generations with thumbnail, model badge, date, and action buttons.
Image Detail Section
AI wrapper app image detail section for Claude Code and Cursor: full-bleed image view with prompt, model metadata, download, variations, and share.
Model Settings Section
AI wrapper app model settings section for Claude Code and Cursor: per-model API key, quality, steps, guidance, negative prompt, and style preset configuration.
Copy the prompt into Claude Code, Cursor, Lovable, Codex, Replit, or Google Stitch. The generated page at app/onboarding/page.tsx is a 3-step wizard with local step state. Step 1 collects API keys for Flux, DALL-E, and SDXL. Step 2 picks a default generation style. Step 3 confirms and redirects to the Studio.
Validate each key server-side in a server action. For Replicate, make a GET request to https://api.replicate.com/v1/account with the key and check for a 200 response. For OpenAI, call openai.models.list() and catch 401 errors. Return success/error to the client and show inline feedback next to the input.
After Step 3, call a server action that sets a has_completed_onboarding boolean on the Supabase user profile. In the middleware (middleware.ts), check this flag for authenticated users and redirect away from /onboarding to /studio if true. The flag check is a single Supabase row read, so it is fast.
A vibe coding template is a full-app scaffold you describe to an AI coding tool instead of writing from scratch. You paste the prompt into Cursor, Claude Code, Lovable, Codex, Replit, or Google Stitch, and the AI generates the correct file structure, routes, components, and design system automatically. The difference from a regular template is that you own the output from the first line: you can immediately ask the AI to swap the auth provider, change the brand color, or add a new route — no boilerplate editing required.
Yes. Every template, section, code file, and included image is free to use in your website, app, client project, SaaS product, or open-source build with no attribution required. There is no license fee, no royalty, and no copyright restriction. The code is yours to keep, modify, and ship commercially the moment you copy it.
A live preview of the full template, the complete multi-file source code as a single copy-paste string with `// === filename ===` headers so Cursor and Claude Code split it into the correct files automatically, and the prompt that regenerates the template from scratch in any of the 6 AI coding tools. Every vibe coding template ships with the shared AI Library design system: white background, black on grey typography, brand blue reserved for primary CTAs, small type, minimal icons.
A component is a single UI unit (a chat interface, a bento grid, a scroll progress bar). A vibe coding template is a full app made up of many components: routes, layout, navigation, data flow, and interactions wired together. Both are in the AI Library. Templates live at /prompts/ai-library/vibe-coding-templates. Components live at /prompts/ai-library/website-sections. Every template page cross-links to the AI Library components used inside its sections.
Yes. Every template can be copied, adapted, sold, and shipped in commercial projects with no attribution required. This applies to the full-app parent code and every child section. The AI prompts are also unrestricted. Use them for client work, indie SaaS, agency deliverables, internal tools, and open-source projects.
Every template is built on Next.js 14+ App Router + Tailwind CSS v3 + Switzer font, written in TypeScript. Backend choices are yours: they work with Supabase, Clerk, or Auth.js for authentication, and with Postgres, Convex, PlanetScale, or Neon for data. Templates deploy on Vercel, Cloudflare Pages, or Netlify. No motion library required. All styling is Tailwind utility classes.
Open the template page, click Copy full code, and paste directly into a Cursor or Claude Code chat. Both tools detect the `// === filename ===` headers and split the paste into the correct files in your project. From there you can ask them to swap the auth provider, change the color palette, wire the data to your API, or generate missing routes. The prompt tab gives you the exact vibe coding instruction to hand off instead.
Yes. Auth is wrapped behind a `src/lib/auth.ts` helper so you can swap Supabase, Clerk, Auth.js, or your own SSO by changing one file. The same pattern applies to the database client: swap Supabase for Postgres, Convex, PlanetScale, or Neon without touching the UI. Every template ships a Supabase reference implementation by default because it is the most common choice for AI-native SaaS.
Yes. Every child section is a standalone page and standalone code file. Copy the Users Table section from the SaaS Dashboard template into an AI Agent template and it works with no modification — they share the same design system, typography, spacing scale, and layout primitives. The prompt tab for each section is scoped to that section, so you can vibe code just the piece you need.
Three reasons. First, the prompt regenerates the template matched to your own design tokens, brand color, file structure, and auth provider — faster than editing someone else's code. Second, it is a starting instruction you can extend ("build this dashboard but replace analytics with a Kanban board") instead of a fixed snapshot. Third, each prompt is written specifically for the tool it targets: Claude Code, Cursor, Codex, Lovable, Replit, or Google Stitch — the output is consistent because the design constraints are identical across all six.
All six produce the same visual output. Cursor and Claude Code are best for dropping the full multi-file code into an existing repo where you want the AI to manage imports, run type checks, and iterate. Lovable and Replit are best for a fresh working preview URL in seconds without local dev setup. Codex fits agent-driven vibe coding workflows where the AI works autonomously across multiple files. Google Stitch is best if you want to iterate on the design visually before dropping the React code into your project.
Yes. The design system is Tailwind-native: change the brand color in `tailwind.config.ts` and every primary CTA updates. Change the font in `next/font` and every heading and body copy updates. Change the border-radius in the shared component variants and every card, button, and modal updates. Because the templates lean minimal (white background, black text, blue only on CTAs) they adapt easily to any brand without visual conflict.