Health Bud

• 

  User asking the great Health Bud how to improve their marathon speed

Inspiration

We wanted to create a health tool that felt personal and engaging rather than clinical. The idea of blending a retro RPG aesthetic with modern AI came from the question: What if your health assistant felt like an in-game companion instead of an app? For most of us, this was also our first hackathon, so we wanted to build something ambitious, fun, and polished within 24 hours.

What it does

Health Bud is a retro-styled health companion that:

Talks to the user through an RPG-style dialogue system

Analyzes personal stats like BMI, calories, water intake, and sleep recommendations

Generates personalized health advice using Gemini

Classifies user concerns (injury, heart issues, mental health, etc.)

Uses the Google Places API to recommend real-world locations such as gyms, clinics, therapists, and rehab centers

Displays everything in a pixel-art interface with animations, sound effects, and a typewriter text engine

How we built it

Frontend: Next.js + Tailwind, custom pixel art, audio cues, dialogue system, boot sequence, and a turn-based chat viewer.

Backend: Express.js with Gemini 2.5 Flash for health reasoning and natural language understanding.

APIs: Google Places for location recommendations tied to a custom health-need classifier.

Logic: We implemented BMI and other calculated metrics using LaTeX-friendly formulas for clarity and consistency.

Challenges we ran into

Tailwind’s class-purging system broke several dynamic color changes

ESM import issues in Node caused multiple backend crashes early on

The typewriter engine initially erased the first character due to hidden whitespace in Gemini responses

Integrating Gemini + Places under time pressure required debugging CORS, environment variables, and rate limits

Managing sprite states, audio timing, and chat flow simultaneously created complex UI behavior

Coordinating GitHub forks, merges, and conflicts as a first-time team took adjustment

Accomplishments that we're proud of

Building a surprisingly polished full-stack app in one weekend

Shipping a complete health assistant with real-world recommendations

Designing a unique retro aesthetic with sprites, sound, and animated dialogue

Successfully integrating multiple APIs and an AI model

Getting the entire team contributing across different parts of the project despite this being our first hackathon

What we learned

How to structure prompts for reliable, safe, and concise AI responses

How to combine multiple external APIs into a unified user experience

How to manage React state for animations, audio, and incremental text rendering

How to collaborate under real deadlines using Git, forks, and live debugging

How to balance creativity with technical constraints during rapid development

What's next for *****

Adding full conversation memory and long-term health goal tracking

Expanding the classifier to support more detailed medical categories

Introducing mini health “quests” that guide users toward habits or routines

Creating a mobile-friendly version

Allowing Bud to adapt its personality and sprite animations based on long-term user behavior

Integrating more data sources (fitness trackers, step counters, wearables)

Built With

• css
• express.js
• google-gemini
• google-places
• html
• next.js
• node.js
• pixelart
• react.js
• tailwind
• typescript