AI Inventory Pro: The Autonomous Inventory & Kitchen Agent

An agent that visually audits your inventory & kitchen. It cross-references receipts with current stock to predict waste, and autonomously plans meals using Gemini 3's reasoning capabilities.

Comment 1

💡 Inspiration

I noticed a "bug" in the modern AI landscape: most tools are built for tech-savvy youth, leaving the older generation behind. I wanted to build an AI that my grandmother could use—no typing, no complex prompts, just a camera. The goal was to turn the tedious chore of inventory & kitchen management into a seamless, "magic" experience for all ages.

🤖 What it does

AI Inventory Pro is a two-stage autonomous agent:

  1. The Auditor (Vision): Users snap photos of messy fridges or crinkled receipts. The agent uses Gemini 3.0 Flash to identify items, brand names, and quantities, structuring them into a digital inventory instantly.
  2. The Chef (Reasoning): The agent doesn't just list items; it understands them. It cross-references your current inventory to generate meal plan designed to minimize food waste through intelligent recipe selection.

⚙️ How I built it

  1. Brain: Google Gemini 3.0 Flash for high-speed reasoning.
  2. Frontend: Next.js with Tailwind CSS for a mobile-first "Grandma-friendly" UI.
  3. Backend: Supabase for persistent user credits. (Row level security)
  4. Auth: Clerk for secure, passwordless login options.
  5. Deployment: Vercel. I rapidly evolved the project from an initial HTML-only prototype to a full-stack Next.js application within a single 7-day sprint. (13 Dec 2025 to 20 Dec 2025)

🧠 Challenges I ran into

  • Visual Overlap: Distinguishing between stacked items in a crowded fridge. I optimized Gemini's spatial reasoning in the prompt to better segment objects.
  • Accuracy: To prevent "AI Hallucinations," I implemented a verification loop in the system prompt to ground answers strictly in the provided image data.

🏆 Accomplishments that I'm proud of

  • Truly "No-Type" Interface
  • Custom Backend: A custom proxy backend allowing global users to access Gemini 3.0 power seamlessly.
  • Traction: Reached 35,000 users on Threads in just one week, proving that a simplified UI can empower everyone to benefit from Generative AI.

🚀 What's next

Hands-Free Interaction: Integrating Gemini Live for a full Voice Mode. This will allow users to query their inventory or follow cooking steps via natural conversation while their hands are busy in the kitchen.

Industry Expansion (e.g.Real Estate): Scaling the "Autonomous Auditor" engine for the property market. I plan to adapt the vision models to perform Automated Property Inspections, allowing agents and landlords to snap photos of room conditions and appliances to generate structured maintenance logs and inventory reports instantly.

Built With

Share this project:

Updates