🏟️ The Problem: The "Post-Game Bottleneck"
We've all been there: the final whistle blows, and 50,000 people try to squeeze through the same two exits. Large-scale venues face a massive challenge in real-time coordination. Static signage can't account for a sudden surge at Gate A or a 20-minute line for water at Section 102. I built VibeFlow to turn "crowd chaos" into a seamless experience.
💡 The Solution: VibeFlow
VibeFlow is a real-time, AI-powered stadium concierge. It uses Gemini 1.5 Flash to analyze live (simulated) IoT data from gates and concession stands to provide attendees with personalized, "low-friction" navigation advice. It doesn't just show you a map; it understands the current "vibe" of the crowd and tells you exactly where to go.
🛠️ Technical Architecture
I embraced the "Vibe Coding" philosophy, using Google Antigravity to iterate rapidly and deploy to production-grade infrastructure:
Frontend: Streamlit (Python) for a high-performance, real-time dashboard.
AI Engine: Gemini 2.5 Flash (for lightning-fast routing logic) and Gemini 2.5 Pro (for deep analytical "Vibe Checks").
Infrastructure: Deployed on Google Cloud Run with autoscaling enabled (Min instances: 1) to eliminate cold starts and handle stadium-sized traffic.
🧠 The Prompt Gallery: How I Built VibeFlow
To build VibeFlow, I used a series of targeted prompts to handle specific logic blocks. This allowed me to focus on the experience while the AI handled the heavy lifting.
The Architectural Prompt (The Foundation)
"Act as a Full-Stack Developer and Event Logistics Expert. Build a Streamlit app called 'VibeFlow'.
Logic: Create a simulated IoT data engine that generates random congestion levels (Low, Medium, High) for 4 stadium gates and 3 food zones.
UI: Use a dark-mode theme with neon green/red status indicators.
AI Integration: Connect Gemini-1.5-Flash to a sidebar chat window. The AI must be fed the 'Live Data' as context so it can answer fan questions accurately."The Navigation Logic Prompt (The Brain)
"You are the VibeFlow Fan Concierge. When a user asks for an exit route, your goal is to minimize their wait time.
Data Rules: If 'North Gate' is High density, never suggest it. If 'South Gate' is Low density, prioritize it.
Tone: Be energetic, clear, and safety-conscious. Give instructions like a professional stadium announcer."The "Pulse" Simulation Prompt (The Real-Time Factor)
"Write a Python function that uses st.empty() and a 10-second loop to update the stadium occupancy metrics. For each refresh, add a +/- 5% random jitter to the data to simulate fans moving through turnstiles in real-time. Ensure the Plotly bar chart re-renders instantly to reflect these changes."
🚀 Key Technical Wins
Dynamic Simulation: The app features a pulse-generator that randomizes crowd data every 10 seconds, proving the system's ability to handle streaming data.
Serverless Scalability: By deploying on Cloud Run, the app is production-ready. I configured autoscaling to handle sudden surges in traffic during peak exit times.
Contextual Awareness: VibeFlow uses Gemini to bridge the gap between "raw data" and "human advice." The AI doesn't just say a gate is 80% full; it explains why you should take the South exit to save 15 minutes.
🏁 Conclusion
PromptWars 2026 isn't just about code; it's about intent. VibeFlow demonstrates that with the right AI orchestration and cloud infrastructure, we can solve complex physical logistics—like crowd movement—in a matter of hours.
🔗 Links:
Live App: https://vibeflow-96920820644.europe-west1.run.app
Source Code: https://github.com/tejas-18/VibeFlow
Top comments (0)