Home Chef

The problem

Weeknight cooking for a household still means deciding what to eat, scaling recipes, building a grocery list, and shopping, usually across notes apps, chat threads, and store sites. I wanted one place that knew our preferences and recent meals and could help without becoming another SaaS I had to maintain for strangers.

The solution

I built a passcode-protected web app (and PWA) that plans meal batches, chats with an AI chef persona, tracks preferences and history, manages groceries, and shops Harris Teeter items via the Kroger API. It is a household kitchen tool, not a public meal-planning product.

Technical execution

• Next.js App Router panel with mobile-first navigation, session-gated access, and PWA install support for phone use in the kitchen.
• Vercel AI SDK (`ai`) with AI Gateway for meal planning and Chef Carmela chat, plus a cost view for monthly AI usage.
• Neon Postgres with Drizzle for meals, recipes, preferences, groceries, history, and feedback.
• Harris Teeter / Kroger OAuth and per-item product picker to add grocery lines to the cart (add-only; no store scraping).
• Built with modern AI-assisted development workflows while shipping a real AI-powered product for daily household use.

Why it matters

Shows how I use AI where it earns its keep (planning and conversation), wired into a practical everyday workflow with auth, data, and a real grocery integration. Also honest about scope: private household app, gated on purpose so costs and personal details stay contained.