Built “Event Buddy AI” in a day for PromptWars (Google Antigravity + Cloud Run)

_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