Eddy: Everyday AI Agent

Inspiration

As a Computer Science student from India studying in the UK, I make everyday decisions that are small individually but add up: what to cook, how to travel, how to organise my day, and how to remember everything I need to do.

I care about environmental impact, but researching every choice or going out of my way to make a more sustainable one can be difficult when life is already busy.

That inspired Eddy. Its guiding idea is simple:

You shouldn’t have to reorganise your life around sustainability. Your everyday assistant should help make thoughtful choices easier by default.

What it does

Eddy is an AI agent for everyday life. It brings together personal context such as food, pantry items, reminders, calendar events, and preferences to help with tasks like:

• Meal planning: Suggests meals using ingredients I have and adapts them to my preferences.
• Travel decisions: Helps compare options while considering factors such as time, cost, and environmental impact.
• Daily planning: Uses available context to help organise my day.
• Reminders: Helps me keep track of everyday tasks.
• Memory: Remembers preferences I choose to share.

Environmental impact is one consideration, not the only one. Eddy is designed to help users understand trade-offs without guilt or judgement. Environmental figures are estimates, not exact measurements.

How I built it

I used Flutter and Dart for the mobile frontend, Python and FastAPI for the backend, Strands Agents for the agent, and SQLite for persistent data.

The Flutter app communicates with the backend over HTTP. The agent uses tools to work with application data and carry out supported tasks. This allows Eddy to do more than generate a conversational answer: it can use context, call tools, and return results to the interface.

For the hackathon MVP, I focused on a small set of connected everyday workflows rather than trying to build every possible assistant feature. Some integrations and environmental calculations use simplified or demo data, so the project can demonstrate the experience without claiming complete real-world data coverage.

Challenges I ran into

One of my main challenges was connecting the agent to real application behaviour. A useful agent needs more than a good prompt: its tools must work with the right data, handle user requests reliably, and avoid claiming that an action happened when it did not.

I also had to keep the agent’s capabilities and the interface aligned. A polished chat experience is not enough on its own; users need to understand what the agent is doing, what it changed, and when their confirmation is needed.

Finally, I had to keep the scope realistic for a hackathon. I chose to prioritise a few connected workflows over adding lots of features that would be difficult to finish and test properly.

Accomplishments that I'm proud of

I’m proud to have built a working mobile app connected to a backend agent that can use tools and interact with persistent application data.

I’m especially proud of connecting everyday tasks that are often handled separately. Eddy can use pantry context to help with meal decisions, remember user-provided preferences, handle reminders, and support planning and travel decisions.

Most importantly, I built around the idea that environmental impact should fit into the user’s existing life, rather than becoming another task they have to manage.

What I learned

Building Eddy taught me that creating an agentic product involves much more than connecting a language model to a chat interface.

The agent needs useful tools, reliable data, clear boundaries, and a user interface that makes its actions understandable. I also learned to be careful about environmental claims: estimates should be transparent, and the system should explain trade-offs rather than presenting one option as universally best.

The project gave me practical experience bringing together an agent framework, backend services, persistent data, and a mobile interface into one product.

What's next for Eddy: Everyday AI Agent

Next, I’d like to improve the reliability and depth of the daily planning workflow, connect more real-world data sources, and give users clearer control over their stored preferences and permissions.

I’d also like to make environmental estimates more robust by using better-supported data sources and clearly documenting the methodology.

My long-term goal remains:

An everyday assistant that makes life easier while helping better-informed choices fit naturally into it.

Built With

• ai-agents
• artificial-intelligence
• dart
• fastapi
• flutter
• mobile-app
• python
• sqlalchemy
• sqlite
• strands
• strands-agents
• sustainability