• 

  Third place!

FreshFridge 🧊

A retrofit kit that makes any dumb fridge smart — powered by a multi-agent AI swarm built on Google ADK and the A2A protocol.

What It Does

FreshFridge turns any refrigerator into an intelligent inventory management system:

1. Sees what's inside using camera + sensors (Raspberry Pi + ESP32-CAM)
2. Knows what's running low, missing, or going stale (Gemini Vision AI)
3. Shops for the best prices across multiple retailers (Agent Swarm)
4. Saves you money with optimized cross-store shopping lists

Architecture

Camera/Sensors → Fridge Agent → Shopper Agent → Price Scouts → Retailer Agents
     (Edge)        (Cloud)        (Cloud)         (Cloud)         (Cloud)

Four tiers of AI agents communicating via Google's A2A protocol:

• Fridge Agent — Analyzes fridge contents with Gemini Vision
• Shopper Agent — Builds optimized shopping lists from needs
• Price Scout Agents — Survey retailers for best prices (one per store)
• Retailer Inventory Agents — Serve store inventory and pricing data

Tech Stack

Project Structure

freshfridge/
├── agents/              ← ADK agents (Cloud Run)
│   ├── fridge_agent/
│   ├── shopper_agent/
│   ├── price_scout_agent/
│   └── retailer_agent/
├── a2a/agent_cards/     ← A2A protocol agent cards
├── edge/                ← Hardware code
│   ├── pi/              ← Raspberry Pi scripts
│   └── esp32/           ← ESP32-CAM MicroPython
├── dashboard/           ← Web dashboard
├── deploy/              ← GCP deployment scripts
└── mock_data/           ← Sample data for testing

Quick Start

# Clone
git clone https://github.com/jemsbhai/freshfridge.git
cd freshfridge

# Set up environment
cp .env.example .env
# Edit .env with your Google Cloud credentials

# Install dependencies
pip install -r requirements.txt

# Run locally
python -m agents.fridge_agent.agent

Team

Built for Google Agent Hackathon 2026.

License

MIT

Built With

• dockerfile
• python
• shell