Varivo

Varivo starts with a simple promise: type in what you have, and it turns it into dinner — recipe, photo, and all. The interesting part is everything around that core moment — guests can try the product once without registering, then the app converts them into full users with credits, personalization, favorites, and a richer dashboard experience. The name itself is part of the pitch: *varivo* is a Balkan stew made from whatever vegetables are on hand — no recipe, no shopping list, just what's already in the kitchen turned into dinner. That's exactly what the app does, and no competitor in the space (SuperCook, ChefGPT, CuliNova) has a name that carries meaning like that.

The problem

A recipe generator alone isn't enough to feel production-ready. A real product needs a fast public entry point, guest-to-user conversion logic, personal recipe history, favorites, a way to plan meals beyond one-off generations, and billing and credit control for AI usage.

What I built

• Public landing page with hero CTA, recipe form, and generated recipe result card
• Guest mode with one free recipe generation before an account is required
• Full dashboard with recipe generation, my recipes, favorites, meal planner, and settings
• Personalized recipe generation using ingredients, cuisine, time, difficulty, and servings
• Background AI image generation for each recipe after text generation succeeds
• Favorites and a saved-recipe cookbook flow
• Meal planner supporting manual recipe assignment plus AI-generated full-day meal plans
• Credit system with daily bonus logic, wallet balance, and one-time credit packages (no subscription)
• PayPal checkout plus a Paddle overlay checkout for card / Apple Pay / Google Pay, both feeding the same credit wallet via webhook-verified fulfillment
• Recipe and cookbook PDF export

Design decisions

The product is split into two very different modes on purpose. The landing page is conversion-first: warm food imagery, a fast form, and a low-friction guest tryout. The dashboard is utility-first: denser navigation, credit visibility, and saved-recipe organization for returning users. The AI output is also treated like product data, not just a chat response. Recipes are structured, stored, rendered as cards, turned into images, and reused again in favorites, meal planning, and PDF export. Payments follow the same "treat it like real product infrastructure" approach: PayPal and Paddle run in parallel as two independent rails into one credit wallet, each with its own idempotent webhook path rather than a client-trusted "add credits" call — the frontend never mints credits itself.

Under the hood

The frontend is a React application styled with Tailwind CSS and shadcn/ui. Supabase provides auth, Postgres, storage, RLS-backed data access, and edge functions. OpenAI handles recipe generation and recipe image generation, and PayPal and Paddle are both used for credit purchases, with signature-verified webhooks as the single source of truth for fulfillment.

Why it matters

This project signals:

• Building an actual AI product instead of a single AI demo screen
• Connecting frontend UX with billing, persistence, and backend automation
• Designing both acquisition UX and retained-user workflows
• Handling structured AI output as reusable product data
• Running two payment providers side by side without regressing the one that already works

Final positioning line

A full-stack AI cooking platform with guest onboarding, structured recipe generation, dashboard workflows, meal planning, and credit billing behind the scenes.

</content>