This is a submission for the Redis AI Challenge: Real-Time AI Innovators.
What I Built
I created Rock Paper Scissors Mind Reader - an AI-powered game where a machine learning model attempts to predict your next move in real-time by analyzing your playing patterns. The twist? The AI explains its reasoning, showing you exactly how it's learning to read your mind!
Key Features:
- Real-time AI Predictions: The AI analyzes your patterns and predicts your next move before you make it
- Explainable AI: See exactly why the AI made its prediction with detailed reasoning
- Pattern Visualization: Watch your playing patterns emerge in real-time
- Global Leaderboard: Compete with players worldwide
- Client-Side API Keys: Secure, privacy-focused design where players use their own Gemini API keys
The game creates a fascinating psychological battle where you try to be unpredictable while the AI gets smarter with every move!
Demo
- Live Demo: https://rock-paper-scissors-web-ebon.vercel.app
- Video:
-
Screenshot:
-
GitHub repo:
depapp
/
rock-paper-scissors
An innovative real-time AI-powered prediction game where players battle against an AI that learns their patterns and tries to predict their next move in Rock Paper Scissors. Built for the Redis "Real-Time AI Innovators" challenge.
🧠 Rock Paper Scissors Mind Reader - AI Prediction Game
An innovative real-time AI-powered prediction game where players battle against an AI that learns their patterns and tries to predict their next move in Rock Paper Scissors. Built for the Redis "Real-Time AI Innovators" challenge.
🎮 Game Concept
Players choose between Rock ✊, Paper 📄, or Scissors ✂️ while an AI analyzes their patterns in real-time and predicts their next choice. The AI's prediction is hidden until after the player makes their choice, creating true suspense. The AI gets smarter with each move, learning from:
- Frequency patterns
- Sequential patterns
- Complex multi-step patterns
- Meta-patterns (trying to be unpredictable)
- Psychological patterns (panic choices, pressure responses)
🚀 Key Features
Real-Time AI Learning
- Pattern Recognition: AI analyzes player behavior in real-time
- Adaptive Difficulty: AI confidence grows as it learns your patterns
- Explainable AI: AI explains its reasoning for each prediction
- Multiple…
How I Used Redis 8
Redis 8 is the backbone of this application, powering every real-time feature. Here's how I leveraged its capabilities:
1. Vector Search for Pattern Matching
// Find players with similar playing patterns
const similarPlayers = await redis.ft.search('idx:profiles',
`*=>[KNN 10 @embedding $vec]`
);
I use Redis Vector Search to find players with similar patterns, helping the AI learn from collective behavior.
2. Semantic Caching for AI Predictions
// Cache AI analysis for similar pattern combinations
await redis.setex(`analysis:${patternHash}`, 300, aiPrediction);
Frequently occurring patterns are cached, reducing API calls and providing instant predictions.
3. Redis Streams for Move History
// Record every move for pattern analysis
await redis.xadd(`moves:${gameId}`, '*', {
move: playerMove,
prediction: aiPrediction,
timestamp: Date.now()
});
Every game move is streamed in real-time, creating a comprehensive dataset for pattern analysis.
4. Sorted Sets for Global Leaderboard
// Update player rankings in real-time
await redis.zadd('leaderboard', {
score: winRate * 1000 + totalWins,
value: username
});
The leaderboard updates instantly as games finish, showing top players globally.
5. Pub/Sub for Real-Time Updates
// Broadcast game events to all connected clients
redis.publish('game:updates', JSON.stringify({
type: 'leaderboard_update',
data: newRankings
}));
Live updates keep all players synchronized with the latest statistics.
6. Time-Series for Performance Metrics
// Track AI accuracy over time
await redis.ts.add('ai:accuracy', '*', accuracyScore);
Monitor how the AI's prediction accuracy improves as it learns more patterns.
7. Hash Storage for Game State
// Store complex game state efficiently
await redis.hset(`game:${gameId}`, {
playerPattern: JSON.stringify(patterns),
aiConfidence: confidence,
moveHistory: JSON.stringify(history)
});
Technical Architecture
Stack:
- Frontend: Next.js 14, TypeScript, Tailwind CSS, Framer Motion
- Backend: Node.js, Express, Socket.io, TypeScript
- AI: Google Gemini for pattern analysis and explanations
- Database: Redis 8 (Cloud)
- Architecture: Monorepo with shared types
Security & Privacy:
- Client-side API key management (keys never touch the server)
- Local storage for user preferences
- No personal data collection
Challenges and Learnings
Building this project taught me:
- How to effectively combine multiple Redis features for real-time AI
- The importance of client-side security for API keys
- Creating engaging visualizations for complex data
- Balancing AI intelligence with game enjoyment
Future Enhancements
- Tournament mode with brackets
- More sophisticated pattern detection algorithms
- Mobile app version
- AI difficulty levels
- Pattern sharing between friends
This project showcases how Redis 8's real-time capabilities can create engaging AI-powered experiences. By combining vector search, caching, streams, and pub/sub, I've built a game that's not just fun to play, but also demonstrates the future of real-time AI applications.
The mind-reading aspect creates a unique psychological challenge that keeps players coming back, while the technical implementation shows the power of Redis as a complete real-time data platform for AI applications.
Top comments (0)