Under the Hood of
MealHouse.app
A multi-tenant, AI-powered meal planning PWA leveraging strict data isolation algorithms and edge-level personalization for distributed families.
The Purpose | Why MealHouse Exists
The Problem: Cognitive Load of Dinner Deciding what to eat, manually scanning scattered recipes, cross-referencing ingredients, and building a shopping list is an exhausting weekly chore. Generic meal planning apps feel impersonal because they constantly suggest trendy recipes you don't actually intend to cook.
The Mission: Algorithm-Driven Organization MealHouse flips the model. Instead of suggesting generic meals, it ingests your family's actual recipes (via four distinct import engines) and creates an autonomous, personalized ecosystem. It generates a 7-day meal plan exclusively from your database, cross-referencing a 30-day historical cache to mathematically ensure meals never repeat too frequently.
The Impact: Accessible Multi-Tenancy
Priced at an aggressive $2.99/month, the SaaS democratizes advanced AI organization. Additionally, it isn't deployed as a shared global dashboard. Through complex wildcard routing, each family receives an isolated, secure subdomain (e.g., burchfamily.mealhouse.app) giving them a deeply personalized application environment.
The Application | Engineered for Daily Friction
Dynamic PWA Installation This isn't a clunky mobile website; it is a true Progressive Web App (PWA). Users download it directly to their iOS or Android Home Screen, fully bypassing App Store friction. The home screen icon is completely customizable, featuring the translucent MealHouse logo overlaid with a family-selected emoji or icon.
Algorithmic Grocery Engine The 7-day AI plan isn't just a static calendar—it's a computational engine. The moment the AI locks in the week's meals, the app parses every selected recipe and automatically compiles a deduplicated, reactive ingredient checklist. Because the database is highly synchronized, checking off an ingredient instantly updates the list on every family member's device.
Data Import Infrastructure Users aren't locked into manual data entry. The platform supports four distinct ingestion methods—processing complex web URL scraping, unstructured text dumps, or manual inputs, and algorithmically converting them into mathematically structured JSON recipe cards.
System Architecture
Multi-Tenant Edge Routing Engineered custom wildcard DNS routing and Next.js Edge Middleware to seamlessly intercept subdomains. It verifies the tenant identifier, authenticates the session, and restricts access solely to that specific family's database partition to guarantee strict data isolation.
Serverless AI Context Pipeline Built a custom algorithm that fetches the family's 30-day recipe history from the database, maps it alongside their complete recipe catalog, and injects it securely into the LLM context window. The agent processes this data against strict repetition constraints to return varying, perfectly formatted meal objects.
Event-Driven Subscriptions Utilized Stripe webhooks for autonomous subscription lifecycle management (handling the $2.99/mo structure), automatically provisioning the subdomain and creating the initial database shards via Firebase Admin SDK upon successful checkout.
Core Stack:
Next.js App Router • Vercel Edge Middleware • Progressive Web App (PWA) • Google Generative AI (Gemini) • Firebase Auth / Firestore • Stripe API • Tailwind CSS