Skip to content

Latest commit

 

History

History
28 lines (20 loc) · 1.91 KB

File metadata and controls

28 lines (20 loc) · 1.91 KB

🤖 Gemini Built: Multiplay

This project was built and refined by Antigravity, a powerful agentic AI coding assistant designed by the Google DeepMind team.

🏗️ System Architecture & Logic Residence

Multiplay is designed with a strict separation of concerns to ensure UI performance and data integrity:

  • Game Engine (The Brain): Resides in src/worker/. All game state, SRS algorithms, and adaptive logic run in a dedicated Web Worker to prevent UI jank.
  • Mastery Engine: Located in src/worker/srs-algorithm.ts. This is the core pedagogical engine that handles specialized logic:
    • Tiered Fluency System: Priorities speed (automaticity) using weighted thresholds (<3s, 3-6s, >6s). Mastery requires speed, not just correctness.
    • Active Set Management: Constrains the learning focus to a rotating set of 15 facts.
    • Weighted Selection: Uses a probability distribution (60% Weak Pool / 20% Mastered / 20% Learning) for optimal drilling.
    • Pedagogical Progression: Automatically unlocks tables based on difficulty patterns (e.g., 2s and 5s before 7s) rather than numerical order.
  • UI State Proxy: Resides in src/lib/stores/game.svelte.ts. Uses Svelte 5 Runes to mirror worker state and expose it reactively to the frontend.
  • Persistence Layer: Found in src/worker/storage.ts. Implements scoped profile storage using a custom versioned IndexedDB (defined in src/lib/db/database.ts).

🛠️ Tech Stack (AI Curated)

  • Framework: Svelte 5 (Runes) for granular reactivity.
  • Threading: Web Workers + Comlink for type-safe RPC communication.
  • Styling: Tailwind CSS with custom Modern CSS Animations.
  • Persistence: Custom versioned IndexedDB for offline-first player profiles and progress.
  • Performance: Latency-masking patterns (700ms feedback windows) to hide background worker operations.

Built with ❤️ by Antigravity @ Google DeepMind.