WasteNot

Inspiration

Every day, households throw away perfectly good food because they lack the time to plan recipes or the network to share surplus. This repetitive waste harms the environment and our wallets. We wanted to build an "Agent for Humans" that handles this repetitive cognitive load. WasteNot deploys a coordinated swarm of AI agents that not only save your food but also connect you with your community.

What it does

When a user inputs expiring ingredients (e.g., "2 bananas, milk, bread"), WasteNot triggers a Multi-Agent Router:

Culinary Agent: Analyzes the ingredients and generates a zero-waste recipe. Eco Agent: Calculates the precise CO2 and virtual water saved by not throwing the food away, providing instant gratification. Community Agent: If there's excess, drafts a polite, ready-to-send message for local neighborhood groups (WhatsApp/Nextdoor) to offer the food.

How we built it

AI Orchestration: AWS Strands Agents SDK (Python) used to define a multi-agent system with a Router pattern. Agent Tools: Each agent is equipped with specific @tool functions (recipe generation, carbon footprint calculation, NLP message drafting). Frontend: React 18 + Vite + Tailwind CSS + Framer Motion to visualize the real-time agent collaboration.

Challenges we ran into

Designing the inter-agent communication. The Culinary Agent's output (what you will cook) needed to dictate whether the Community Agent should trigger (if there's still excess left). We solved this by implementing a stateful router in the Strands SDK.

What we learned

We learned that "Good Neighbor" agents need to balance utility with social etiquette. Drafting a message to neighbors required careful prompt engineering to sound human, not robotic.