GrocerEase AI

GrocerEase AI - Your AI companion that simplifies SNAP/WIC benefits, food access and stretches every dollar.

Inspiration

We came to ShellHacks with a vision of building something that could truly help people. After hearing the sponsor talks and opening ceremony, we felt inspired to build an application that could make a real impact on communities, especially seniors, and SNAP/WIC users who face daily challenges accessing nutritious food. That inspiration shaped GrocerEase AI, our AI companion designed to simplify food access and stretch every dollar.

What it does

GrocerEase AI uses three intelligent agents that work together to make grocery shopping easier, healthier, and more affordable, they are:

Budget and Store Agent that scans stores like Walmart and Target, tracking prices and SNAP/WIC eligibility.
Nutrition Agent personalizes recommendations to maximize nutritional value on a budget and adapt to dietary needs.
Transit Agent maps all available routes to stores, factoring in public transit schedules and walking distance.

The agents coordinate using Google’s A2A protocol to produce one simple plan: where to shop, what to buy, how much it’ll cost, and how to get there safely.

How we built it

We created an AI-powered grocery shopping assistant that combines budget optimization with nutrition analysis for SNAP/WIC users. Here's what it does: Key Features:

Budget Agent Integration: Uses a LLM Agent - Google Gemini to find affordable grocery items within budget constraints.
Nutrition Agent: Provides health and nutrition guidance for food choices.
Store Comparison: Recommends between Walmart and Target based on pricing and quality.
Coordinated Response: Combines budget, nutrition, and store recommendations into one comprehensive shopping guide.

How It Works

User Input: Takes grocery requests via command line
Budget Analysis: First calls the budget agent to find items within SNAP/WIC budget
Nutrition Analysis: Then provides nutrition guidance for the selected foods
Combined Output: Delivers a unified response with:
Complete shopping list with prices
Store recommendations (Walmart vs Target)
Nutrition insights
Actionable shopping strategy

Technical Architecture

Built using Google's ADK (Agent Development Kit)
Uses Gemini 2.0 Flash as the coordinator model
Implements asyncio for concurrent agent communication
Features session management for interactive conversations

Challenges we ran into

Designing a consistent message schema so all agents could understand each other.
Balancing scope: focusing on 3 agents for MVP while planning future enhancements.
Debugging real-time communication between agents under hackathon time pressure.

Accomplishments that we're proud of

Learned and applied new tools like Google ADK and A2A within a short hackathon timeframe.
Developed an app concept that judges and users can immediately understand and connect with.
Built a multi-agent system where agents collaborate instead of working in isolation.
Created a solution that addresses real human problems.
Efficiently collaborating within the team and most importantly learning throughout the process from peers and having a fun time.

What we learned

How to design and structure AI agents with loops that adapt to user data.
The importance of standardized communication (JSON schemas) in multi-agent systems.
The potential of AI for social impact, building solutions not just for convenience, but for making a difference and helping others.
Collaboration skills: dividing tasks across agents and integrating them into one system.

What's next for GrocerEase AI

Scanning grocery stores in nearby and local areas and finding the best possible transit to that place.
Scaling our functionalities for real use.
Crime/safety data focusing on geographical location.