PuzzleRush
Inspiration
PuzzleRush was born from the idea of merging classic jigsaw puzzles with modern AI to create a unique, adaptive challenge every time you play.
We wanted to eliminate the repetitive nature of traditional puzzle games by generating fresh, themed images and missions on demand.
What it does
PuzzleRush is an AI-powered jigsaw puzzle game.
Game Flow
Pick a Theme
Choose from Animals, Space, Cities, Cartoons, Nature, or Random.AI Generates the Challenge
A Large Language Model creates a unique title, mission story, difficulty recommendation, and strategic hint.AI Generates the Image
An AI image generation model creates a custom puzzle image based on the selected theme.Play the Game
Drag and drop the shuffled pieces into the correct grid positions before the timer runs out.Save & Replay
Every generated puzzle is saved to a personal Library, so users can replay their favorites anytime.
How we built it
Frontend
- React
- Vite
- Tailwind CSS
- Custom 8-bit pixel-art aesthetic
Game Logic
- Custom drag-and-drop puzzle mechanics
- Framer Motion for smooth interactions
State Management
- Zustand for managing game state, scores, and challenge data
AI Integration
- Gemini 2.5 Flash for generating:
- Challenge configurations
- Puzzle titles
- Mission stories
- Hints
- Difficulty recommendations
- AI Image Generation for creating unique puzzle visuals
Backend
- Supabase for storing generated puzzles in a PostgreSQL database
- Supabase Storage for hosting puzzle images
Edge Functions
- Serverless functions securely handle API calls to AI services
Challenges we ran into
API Polling
Image generation is asynchronous and can take up to 60 seconds. We built a polling mechanism to track task status such as:
PROCESSINGPENDINGSUCCESS
This prevents the UI from crashing while waiting for generated images.
Response Parsing
The image generation API returns base64 data nested inside markdown syntax:
Built With
- vibecoding
Log in or sign up for Devpost to join the conversation.