BingeHouse

Spend your time watching movies and not searching movies.

Comment

📌 Inspiration

We noticed how frustrating it is to decide whether a movie is actually worth watching especially with fake reviews, inconsistent IMDb ratings, and biased opinions. Most apps just show raw ratings without context. We wanted to build an AI-powered, chat-based recommendation experience that feels personal, unbiased, and genuinely helpful right when you need it.

📌 What it does

BingeHouse is a chat-based movie recommendation mobile app built with Expo.

  • Users can ask if a movie is worth watching.
  • The app checks if the movie data exists in our Supabase database.
  • If missing, it fetches fresh details from the OMDb API (IMDb rating, vote count, reviews).
  • An AI assistant powered by OpenAI GPT-3.5 Turbo then generates a short, crisp recommendation verdict.
  • If ratings and reviews are strong, it recommends watching. Otherwise, it leaves the decision to the user.
  • The verdict is stored for future lookups to reduce redundant API calls.
  • Supports both guest mode and login accounts for personal watchlists and ratings after watching.

📌 How we built it

  • Expo SDK 53 for cross-platform mobile app development.
  • Supabase for database management and secure Row-Level Security (RLS).
  • Supabase Edge Functions for safely handling OpenAI API calls and database queries.
  • OpenAI GPT-3.5 Turbo for generating natural language movie verdicts.
  • AsyncStorage for handling guest user chat history locally.
  • OMDb API for fetching movie ratings and metadata.

📌 Challenges we ran into

  • Ensuring API keys were never exposed in the frontend securely moved everything to Edge Functions.
  • Implementing Row-Level Security policies on Supabase to isolate user-specific data (watchlists, chats, ratings).
  • Handling guest and logged-in user flows without breaking the chat experience.
  • Keeping token usage efficient while maintaining natural AI conversations.
  • Avoiding redundant OMDb API calls and caching results smartly.

📌 Accomplishments that we're proud of

  • Delivered a clean, error-free Expo app fully compatible with SDK 53.
  • Integrated OpenAI GPT-3.5 Turbo securely via Edge Functions without exposing any secrets.
  • Built a chat-based recommendation system that feels genuinely helpful, not just a bot spitting out raw scores.
  • Set up Supabase with robust RLS security, guest mode, and persistent user watchlists.
  • Cleanly separated logic, database, and AI operations for easy future scaling.

📌 What we learned

  • How to effectively structure an AI-assisted app with secure backend functions on Supabase Edge.
  • The importance of thoughtful user flows, especially around guest vs. authenticated modes.
  • Efficient handling of third-party APIs, caching, and minimizing redundant external calls.
  • Best practices for managing secrets using .env files and serverless environments.
  • Optimizing AI token usage without compromising the UX.

📌 What's next for BingeHouse

  • Integrating multiple review sources like Metacritic and RottenTomatoes APIs.
  • Letting users rate movies post-watch to build a trusted community-based custom rating system.
  • Adding AI-powered hidden gem recommendations based on user preferences and watch history.
  • Implementing push notifications for new releases or top-rated recommendations.
  • Making it an installable PWA and Android/iOS release via EAS build.

Built With

Share this project:

Updates