(See the docs/assets/ directory in our repository for screenshots of the dashboard UI and passion gauge in action).

Code

How I Built It

The application is built using a modern full-stack developer architecture:

• 
  Backend: FastAPI (Python) and Uvicorn for fast asynchronous endpoints. We integrated the official Google Generative AI SDK (google-generativeai) to connect to the gemini-1.5-flash model.


• 
  Frontend: React, Vite, TypeScript, Tailwind CSS v4, and Framer Motion for a premium, responsive, dark-mode neon dashboard.

🧠 Gemini Prompts & Integration

1. Structured JSON Passion Score Estimation

We instruct Gemini to output structured JSON data directly by passing a precise scoring template:

prompt = (

    f"Analyze the following sports fan rant text: '{rant}'. "

    f"Calculate a 'passion_score' representing how obsessed, devoted, and emotional the fan is on a scale from 0 to 100. "

    f"Also write a brief 1-sentence supportive response acknowledging their obsession. "

    f"Return ONLY a clean JSON object with keys: 'passion_score' (integer) and 'response_summary' (string)."

)

In the backend services, we configured generation_config={"response_mime_type": "application/json"} to guarantee a clean, parseable JSON block returned to the React frontend.

2. PERSONA-Biased Commentary Generation

To capture the real feeling of sports rivalries, we feed Gemini a commentator persona biased towards a specific team:

prompt = (

    f"You are a fanatical, obsessed football commentator who is highly devoted to {bias_team}. "

    f"Generate a brief, emotional, 3-sentence live commentary stream of a hypothetical match "

    f"between {team_a} and {team_b}. Your tone must show absolute passion, bias, and excitement!"

)

Prize Categories

• 
  Best Use of Google AI: The entire system logic is powered by Google Gemini 1.5 Flash to perform structured sentiment scoring and persona-driven creative content generation.