AURA

Turn any mood into a curated collection of music, movies, books, and experiences—powered by Qloo’s cultural intelligence and LLMs for emotionally aware recommendations.

Comment

🌟 Project Story

🎭 About the Project

Aura bridges the gap between emotions and cultural experiences. Using OpenAI’s GPT-3.5 and Qloo’s recommendation engine, Aura intelligently extracts mood-based keywords and returns curated recommendations — from movie picks to local destinations — tailored to how you feel.

It factors in city-level context (e.g., Bangalore, San Jose, New York) and offers location-aware, multilingual suggestions. Whether you're nostalgic in New York or adventurous in Paris, Aura responds dynamically.

✨ Features include:

  • 🌐 Multi-language summaries
  • 💡 Mood interpretation
  • 🎬 Cultural content diversity (movies, books, artists, destinations)

Aura is a personal companion that understands your emotions and translates them into culturally rich suggestions.

💡 Inspiration

We wanted to create a meaningful interface between human feelings and cultural exploration. Instead of endlessly scrolling through generic content feeds, we asked:

"What if your mood could guide your journey?"

Aura was born from this idea — letting AI and culture work together to surface relevant, emotionally aligned experiences.

⚙️ How It Works

  1. User Input: Mood + (optional) city + language + content type
  2. GPT-3.5: Extracts 3 search keywords from the mood and context
  3. Qloo API: Searches for culturally relevant content
  4. GPT-3.5: Summarizes the mood in the selected language
  5. Frontend: Beautifully displays everything in a responsive interface

🛠️ Tech Stack

Layer Tech Stack
Frontend Next.js (TypeScript), Tailwind CSS
Backend FastAPI (Python), Pydantic
AI OpenAI GPT-3.5
API Qloo Hackathon API
Infrastructure Local Dev via Uvicorn + NPM

🚧 Challenges

  • Mapping abstract moods into precise, search-ready keywords
  • Normalizing location names for global compatibility
  • Rendering meaningful fallback results when Qloo returns empty results
  • Maintaining low latency between frontend-backend interactions

📚 What We Learned

  • Fine-tuning GPT prompts for consistent, concise keyword extraction
  • Building fallbacks to ensure reliable output even when APIs return sparse data
  • Using city-based filters to enrich personalization
  • Designing smooth UI/UX for emotional context-driven exploration

👥 Team

Made with ❤️ by Anirudh Esthuri
For the Qloo LLM Hackathon

Built With

  • fastapi
  • nextjs
  • openai
  • pydantic
  • qlooapi
  • tailwind
Share this project:

Updates