_Hackathons always scared me a bit.
Too many ideas, too little time, and lots of “pro” developers showing crazy projects.
PromptWars changed that.
This time, I decided to keep it simple and actually finish something. I picked the Physical Event Experience challenge and built a tiny web app called Event Buddy AI.
The problem I wanted to solve
At any physical event, people are always confused:
“Where is Hall A or Hall B?”
“Where’s the help desk or food court?”
“What can I attend next that actually matches my interests?”
“I only have 30 minutes. What should I do?”
Most event apps are either too heavy or don’t give quick answers. I wanted a lightweight assistant that feels like asking a friend who knows the venue and schedule.
What I built – Event Buddy AI
Event Buddy AI is a simple web app where an attendee can:
Ask venue questions (halls, food, washrooms, help desk).
Get a suggestion for the next session based on their interests (AI, design, startup, etc.).
Use a few sample questions if they don’t know what to type.
See small info cards for halls and key locations.
It’s not a big enterprise product – it’s a focused helper that answers “What do I do now and where do I go?”
Tech stack
I wanted to lean on AI tools instead of overcomplicating things:
Google Antigravity for planning and generating most of the code.
Python + Flask for the backend.
HTML, CSS, JS for a simple, responsive UI.
Local JSON file to store sample event data (halls, sessions, locations).
Google Cloud Run to deploy the app and get a public URL.
Antigravity helped me quickly scaffold the project structure, generate the Flask app, and even prepare it for Cloud Run. I focused more on the logic and UX instead of wiring everything from scratch.
How it works
The app loads event data from a JSON file (halls, times, tags like AI/design/startup).
The user types a question or clicks a sample question.
The backend does some simple intent + keyword matching on the question.
It returns:
A direct answer (e.g., “Hall A is on Level 1 near the main entrance”), or
A recommended session (e.g., “You like AI? Attend ‘Intro to GenAI’ at 3:00 PM in Hall B.”)
The UI shows the answer clearly, with an option to ask another question.
No heavy databases, no complex auth – just a fast conversational layer over structured event data.
Challenges I hit
Folder navigation & Cloud Run commands:
I messed up basic things like cd paths and service names with spaces. Small details, but they break commands quickly.
Cloud Run permissions:
My first deploy failed due to missing IAM permissions for the Compute Engine default service account to read from Cloud Storage. Fixing that taught me more about roles/storage.objectViewer than any tutorial.
Time management:
The biggest win was resisting the urge to add 10 extra features. I kept reminding myself: working and simple is better than big and broken.
What I learned
You don’t have to be a pro to ship something useful in a hackathon.
AI tools like Google Antigravity are powerful if you give them a clear prompt and a realistic scope.
Cloud Run makes it easy to go from “local Flask app” to a URL you can share with anyone.
A small, well-explained project with clean UX and a live demo is much better than an unfinished “big idea.”
Links
GitHub repo: https://github.com/Varsha-678/Event-Buddy-AI-Promptwars.git
Live demo on Cloud Run: https://event-buddy-ai-14332155047.us-central1.run.app
Top comments (0)