Gambit

Real-time multiplayer chess game. Create room, share code, play chess.

Tech Stack

Frontend: Next.js 14, TypeScript, Tailwind CSS, React Chessboard
Backend: Express.js, TypeScript, Socket.IO, Custom Chess Engine
Database: Supabase PostgreSQL
Real-time: WebSocket connections
AI Bots: LLM Integration (Gemini/OpenRouter) + Stockfish Fallback

Architecture

Frontend (Vercel)     Backend (Render)      Database (Supabase)
┌─────────────────┐   ┌─────────────────┐   ┌─────────────────┐
│ Next.js 14      │◄──┤ Express + Socket│◄──┤ PostgreSQL      │
│ TypeScript      │   │ Real-time WS    │   │ Real-time API   │
│ Tailwind CSS    │   │ Chess Engine    │   │ Row Level Sec   │
│ Socket.IO Client│   │ Move Validation │   │ Auto-scaling    │
│ Optimistic UI   │   │ Room Management │   │ Persistence     │
└─────────────────┘   └─────────────────┘   └─────────────────┘

How it works

Player creates room, gets 6-digit code
Player shares code with friend
Friend joins using code
Chess game starts automatically
Custom engine validates all moves
WebSocket syncs moves instantly between players

Run locally

git clone https://github.com/dushyant4665/gambit
cd gambit

# Terminal 1 - Server
cd server
npm install
npm run dev

# Terminal 2 - Client  
cd client
npm install
npm run dev

Open http://localhost:3000

Connect GitHub repo to both platforms. Auto-deploys on push.

Features

All chess rules implemented
Move validation and checkmate detection
Player names and room codes
Mobile responsive design
Database persistence
Real-time WebSocket sync
No authentication required
AI Chess Bots: Michael Tal (aggressive) & Bobby Fischer (precise) with LLM + Stockfish integration

AI Chess Bots

The application features two AI chess bots with distinct playing styles:

Michael Tal Bot

• Style: Aggressive, tactical, sacrifice-friendly
• Personality: Known for brilliant tactical sacrifices and attacking play
• Technology: Gemini LLM + Stockfish validation
• Sacrifice Tolerance: Up to 5 pawns for tactical advantage

Bobby Fischer Bot

• Style: Classical, precise, accuracy-first
• Personality: Perfect endgame technique and positional understanding
• Technology: OpenRouter/DeepSeek LLM + Stockfish validation
• Sacrifice Tolerance: Conservative, prefers high-evaluation moves

Bot Architecture

1. LLM Analysis: Bot analyzes position with personality-specific prompts
2. Move Suggestions: LLM provides 3 candidate moves with rationale
3. Stockfish Validation: Each move validated for legality and evaluation
4. Fallback System: Stockfish provides strong moves when LLM suggestions fail
5. Authoritative Server: All moves validated server-side before persistence

Configuration

• Free Tier Friendly: Works without API keys using Stockfish-only mode
• Rate Limited: 20 LLM calls per minute with automatic fallback
• Robust Fallback: Multiple layers ensure bots always provide legal moves
• Debug Endpoints: Analyze bot decisions and move sources

See /server/bots/README.md for detailed documentation.