Introduction
As a basketball captain and a Computer Science student, I’ve spent a lot of time in stadiums. While the game on the court is thrilling, the "game" of navigating the venue—finding your gate, waiting for overpriced nachos, and deciphering confusing signage—is usually a loss.
For the Hack2Skill x Google for Developers PromptWar, I decided to solve this using Agentic AI. Introducing VibeFlow AI: a multimodal event concierge built with Gemini 1.5 Pro and Google Antigravity.
The Challenge
Large-scale venues face three major friction points:
Navigation Fatigue: Physical signs are hard to read in crowds.
Crowd Bottlenecks: Gates and concession stands get overwhelmed.
Accessibility Gaps: Real-time information is rarely inclusive.
The Solution: VibeFlow AI
VibeFlow isn't just a static app; it's a system of coordinated agents that act as the stadium's digital nervous system.
🧠 Multimodal Vision Assistant
Using Gemini 1.5 Pro, attendees can snap a photo of their physical ticket. The agent parses the sector, row, and seat, then generates a natural-language "walkthrough" to their location. No more squinting at maps.
🗺️ Predictive Crowd Routing
We integrated the Google Maps Platform with simulated gate density data. If Gate 4 is backed up, the agent proactively reroutes fans to Gate 2, ensuring a steady flow and shorter wait times.
⚡ The Tech Stack
Orchestration: Google Antigravity (Agentic IDE)
LLM: Gemini 1.5 Pro (Multimodal)
Frontend: Streamlit
Data State: Firebase Firestore
Navigation: Google Maps Directions API
Technical Deep Dive: The Agentic Workflow
One of the most unique aspects of this build was using Google Antigravity. Instead of writing every line of boilerplate, I acted as the Architect, defining Standard Operating Procedures (SOPs) in an AGENTS.md file.
Markdown
AGENTS.md Excerpt
- Security: Sanitize all Vision inputs.
- Efficiency: Cache Gemini API responses using @st.cache_data.
- Accessibility: Mandatory ARIA labels and alt-text generation. This allowed the AI agents to "self-police" the code quality while I focused on the logic for the Ekana Stadium (Lucknow) case study.
Why Accessibility Matters
We built VibeFlow with an "Inclusion-First" mindset. By leveraging Gemini's ability to describe surroundings, a visually impaired fan can point their camera at the field, and the AI will describe the current score, the crowd's energy, or the nearest exit path.
Deployment & Links
This project is live and ready for the PromptWar evaluation!
🚀 Live Demo: https://vibeflow-ai-154600966753.us-central1.run.app
💻 GitHub Repo: https://github.com/dev-vrats/VibeFlow-AI.git
Closing Thoughts
Building this during the PromptWar taught me that the future of development isn't just writing code—it's orchestrating intelligence. Huge thanks to Hack2Skill and Google for Developers for the opportunity to build in this new "Agentic" era.
What do you think is the biggest pain point at sporting events? Let’s discuss in the comments!
Top comments (0)