VVyshyvka
studio

Case study

Modern Recipes

An authenticated recipe platform for coordinating recipes, pantry inventory, USDA-backed ingredient nutrition, meal calendars, shopping lists, and AI-assisted recipe and media workflows.

Capabilities
Design, Frontend, Backend, DBA, Data Modeling, AI Workflows
Type
Full-stack product system
Stack
Next.js, React, TypeScript, Drizzle ORM, Cloudflare
Scope
Meals, Pantry, Nutrition, Shopping, AI workflows
Year
2025

Product inventory

Repository evidence: routes, data model, interface screens, and workflow code.

Page routes
57
Drizzle tables
92
UI components
459
Pipeline files
171
API routes
4

Project overview

An authenticated recipe platform for coordinating recipes, pantry inventory, USDA-backed ingredient nutrition, meal calendars, shopping lists, and AI-assisted recipe and media workflows.

Household meal planning breaks down when recipes, pantry inventory, ingredient nutrition, calendar slots, shopping lists, and media workflows live in separate tools. Tracker models those pieces as connected operational records so planning, cooking, stocking, and review can happen in one place.

What was built

A broad household domain turned into product software: large nutrition tables, catalog cleanup, recipe generation, and media pipelines in one application.

Recipe library

Structured recipes with ingredients, methods, imagery, filters, and detail views designed for reuse in planning and shopping.

Pantry and ingredients

Ingredient and pantry records connect household inventory, food imagery, source names, aliases, and usage context.

Nutrition data layer

FoodData Central Foundation and SR Legacy records are cleaned, deduplicated, normalized into hundreds of thousands of nutrient rows, and linked to ingredient profiles with source provenance.

Meal calendar

Weekly meal slots, recipe selection, and planning state connect the recipe catalog to repeated household routines.

Shopping aggregation

Shopping views derive categorized purchasing work from planned meals, pantry review, and ingredient records.

AI and media workflows

Natural-language recipe generation turns prompts into structured recipes using nutrient targets, pantry/equipment preferences, dietary constraints, ingredient matching, and queued image jobs.

Selected details

Implementation details from the product, data, workflow, and infrastructure layers.

  • 01

    Nutrition profiles, nutrients, ingredient links, and unit conversions are modeled as structured data with provenance rather than medical guidance.

  • 02

    USDA import and repair scripts processed FoodData Central Foundation and SR Legacy records into normalized profile, nutrient, ingredient, and unit-conversion tables.

  • 03

    The local dataset includes thousands of ingredient/profile links and hundreds of thousands of normalized nutrient rows, giving recipe generation real nutrition context.

  • 04

    Dietary compatibility evaluation produces blocks, warnings, ranking reasons, and unknown-fact states for planning support.

  • 05

    Recipe generation accepts instructions such as nutrient emphasis, cuisine, dietary restrictions, preferred or avoided ingredients, equipment, pantry mode, and time constraints.

  • 06

    Generated recipe ingredients are post-processed against catalog candidates before becoming linked records.

  • 07

    Recipe, ingredient, and equipment image generation moves saved jobs through database rows, Cloudflare queue infrastructure, WebP optimization, R2 storage, and retry/status tracking.

  • 08

    Upload and file-serving routes support authenticated ownership checks, WebP conversion, local storage, and R2-backed deployment.

Systems overview

How the public pages, staff workflows, data, review steps, and infrastructure connect.

Recipes

Structured ingredients, steps, assets.

Nutrition

Profiles, calculations, provenance.

Meal plans

Calendar slots, templates, recurrence.

Shopping

Aggregated needs and pantry review.

Pantry

Catalog, inventory, household context.

Technology

The platform stack behind the build.

Next.js
React
TypeScript
Drizzle ORM
SQLite
Cloudflare D1
Cloudflare R2
NextAuth
Zod

Design language

Warm editorial minimalism for household operations: parchment-white surfaces, thin structural rules, compact type, restrained teal accents, and food imagery used as record evidence.

Philosophy

Operational calm for repeated household work

Tracker treats food records as household operations data: searchable, inspectable, and connected to weekly planning.

Record gridsInspector drawersMeal calendarsGrouped shopping lists
Aa
Moderustic, Cormorant, Fira CodeUI, editorial display, numeric utility
Ink#1a1a1a
Sand#faf8f4
Rule#efe9e0
Teal#619394
Spice#c99950

Selected screens

Additional screens from the case study.