Playroom Diet

Playroom Diet is an AI-powered child development tool that transforms chaotic toy piles into science-backed growth plans. Upload a photo of your child's playroom, and a multi-agent AI system analyzes the toys, maps them to professional skill frameworks, and creates a six-month development roadmap. It also delivers play quests that encourage creative use of the toys you already own. Think of it as a nutritionist for your child’s play, ensuring a balanced diet of developmental stimulation.

This project demonstrates a multi-agent AI pipeline using Gemini 3 and Apache Airflow hosted on Astronomer Astro.

1. Agent 1: Vision agent extracts toy inventory with bounding box coordinates for visualization.

2. Agent 2: Maps toys to the O*NET professional skills framework to identify which cognitive and physical abilities the toys develop and which require further development. Create recommendations of toys to add. Uses a custom tool function to access an O*NET abilities and occupations database to inform parents about which abilities are important for specific careers.

3. Agent 3: Safety-checks recommendations against CPSC guidelines and generates search queries to find the suggested toys on platforms like Amazon. Uses a tool function to retrieve latest product data.

4. Agent 4: Generates a Play Quest using existing toys to learn and enjoy until new toys arrive.

🧸 The result: Open Playroom Diet on any device, snap a photo of the chaos, and instantly get a science-backed development roadmap, shopping links, and a Play Quest you can start right now with the toys you already own.

Backend

• Python 3.12 + uv

• FastAPI + Uvicorn

• Supabase

• Pydantic

Frontend

• Vue.js 3 + Vite

• Tailwind CSS + daisyUI 5

• Vue Router 4

Airflow (AI pipeline)

• Apache Airflow 3.2

• Airflow Common AI provider

• Gemini API with Gemini 3

Infrastructure

• Backend: Google Cloud Run

• Frontend: Firebase Hosting

• Airflow: Astronomer Astro

• Database / Storage: Supabase