Qloofone

Stop the endless scrolling. Dial 1-877-361-7566, say two things you love, and Qloofone gives you one perfect recommendation, no app, no account, no personal data.

Comment

Inspiration

Growing up, I could simply dial a hotline for movie times, weather, even recipes, no apps, no log-ins, no tracking.

Qloofone revives that privacy-first nostalgia for today’s universal question:

“What should we watch, listen to, or do?”

What It Does

Stuck scrolling or debating with friends? Call 1-877-361-7566, say any two things you enjoy, and hear a single recommendation that bridges both tastes, no personal data, no account, no download.

Examples

  • “Star Wars” + “Pride & Prejudice” → The Princess Bride
  • “Taylor Swift” + “Paris” → Emily in Paris
  • “McDonald’s” + “Whole Foods” → Chipotle

Give It a Try

  1. Dial 1-877-361-7566 from any phone.
  2. When prompted, say two things you love “Harry Potter” and “The Shining” “Beyoncé” and “Marilyn Manson” anything.
  3. Stay on the line; within a few seconds you’ll hear a recommendation that bridges both tastes.

Want to see the magic under the hood? Open Live monitor while you call.
You’ll watch in real time the speech transcript, GPT-4 function-call output, Qloo API response, and the final sentence that gets spoken back to you.

How We Built It

Voice → AI → Recommendation

Started with OpenAI Realtime and added Qloo's cultural intelligence. When you call, your voice streams to an AI that extracts two preferences and immediately says "Got it!" while Qloo finds connections.

Smart recommendation engine:

  • Caches common requests for speed
  • Filters obvious sequels (Star Wars → not Star Wars 2)
  • Transforms data into friendly conversation

Polish that matters:

  • Animated 3d Qloofone logo
  • Retro Futuristic styled branding
  • Live dashboard showing transcripts and function calls in real time

Framework & Language

  • TypeScript everywhere for type-safety
  • Node.js backend with WebSocket server for real-time audio
  • Next.js for the monitoring dashboard and landing page

Key Libraries

  • '@twilio/voice-sdk' – real-time call streaming
  • '@openai/openai-realtime-twilio-demo' – GPT-4 function calling + text-to-speech
  • 'axios' – Qloo API client (with retry & caching layer)
  • 'three.js' + react-three-fiber – landing-page 3-D logo
  • 'tailwindcss' – retro CRT styling

Core Services

  • Twilio Media Streams – bidirectional call audio
  • OpenAI Streaming API – function calls for entity extraction and copywriting
  • Qloo Taste AI – /search + /v2/insights for cross-domain matches

Challenges We Ran Into

  • Speech-to-text homophones (“metal” vs “meddle”) → added re-search and confirmation prompts.
  • Latency budget → cached entity IDs and injected a quick “Got it!” filler response.
  • filter.type pitfalls → dynamic chooser selects movie / artist / place automatically.
  • Sequel fatigue → substring + ID filter keeps results fresh.

Accomplishments We’re Proud Of

  1. Privacy-first: personalized advice without storing any PII.
  2. Thoughtful LLM fusion: GPT-4 handles dialog; Qloo supplies cultural depth.
  3. Nostalgia First Nostalgia powered by modern technology.

What We Learned

  • Designing for voice UX pauses, barge-in, and clarity all matter.
  • Privacy can be a feature, not a hurdle.
  • LLMs + specialized graphs beat LLMs alone for meaningful recommendations.

What’s Next for Qloofone

This solo sprint proved the concept; next I’ll explore elderly-friendly, phone-based AI companions and add SMS/chat interfaces for broader reach.

Built With

Share this project:

Updates