Senior Helper

Inspiration

Our inspiration came from a deeply personal place that resonated with our entire team. One of our team member's fathers is 80 years old, and he shared the story of his dad's struggle with remembering the small but crucial details of his day, like medication times. This story was immediately familiar to all of us, as we've all seen our own parents or grandparents navigate similar challenges. We united around a single, powerful mission: to build a tool that we would genuinely want our own families to use.

What it does

Senior Helper is a conversational web application designed to be a friendly digital assistant for seniors. It allows users to interact naturally with an AI agent . Specifically, it: Provides Daily Reminders for medications and appointments by understanding natural language queries. Offers Live Weather Updates for planning daily activities. Acts as a One-Command Emergency Alert system, instantly sending an SMS to a pre-defined contact via Twilio if the user expresses distress.

How we built it

We engineered a full-stack website within a 36-hour timeframe. Frontend: The user interface was crafted using HTML, Tailwind CSS, and vanilla JavaScript. This static site is hosted on Netlify. Backend: The intelligence is powered by a Python Flask server, deployed on Render. This server exposes a REST API that the frontend communicates with. AI Agent: The core of the backend is Google's Gemini model, configured with a custom persona for seniors and empowered with function calling. The agent autonomously decides when to use its defined tools (Twilio for SMS, OpenWeatherMap for weather) based on the user's conversational intent. We used Git and GitHub for version control, enabling collaborative development and ensuring a smooth deployment pipeline.

Challenges we ran into

API Configuration & Environment Variables: Securely managing and correctly configuring multiple API keys (Gemini, Twilio, OpenWeatherMap) across local and cloud environments (Render) required meticulous attention to detail. Gemini Function Calling Nuances: Integrating Google's Gemini agent with custom tools and ensuring the correct format for function_response involved troubleshooting specific library behaviors and API structures. Deployment & Pathing: Resolving ModuleNotFoundError issues on Render, even with a blank root directory, required diagnostic steps to understand the server's exact file perception and correctly configure Gunicorn's start command. Also we had issues launching the static site, at first we tried porkbun, but it was not a good option and we were unable to make it work and we got stuck for hours, so we switched to netlify, which gave us a functioning website in minutes.

Accomplishments that we're proud of

As first-time hackers, we're extremely proud to have built a project that successfully deploys and works as we intended, especially with the Agent being able to utilize multiple tools.

What we learned

We learned deployment skills using Netlify and Render, along with HTML and JavaScript adjustment techniques for the frontend. We also were able to use the Gemini API through Python.