English | 中文

PlayForge is an AI-powered game design platform built with Spring Boot and React. It integrates multiple LLM providers (OpenAI, Anthropic Claude, Google Gemini) via LangChain4J to help users brainstorm, iterate, and produce professional game design documents through real-time streaming AI chat.

• AI Chat — Real-time streaming conversations with multiple LLM providers via WebSocket
• Multi-Model Support — Switch between OpenAI GPT, Anthropic Claude, and Google Gemini per conversation
• User Authentication — Registration, login, logout with JWT access/refresh tokens and Redis-backed session management
• Profile Management — Edit nickname, bio, and avatar (direct upload to Alibaba Cloud OSS)
• Token Auto-Refresh — Transparent access token refresh via Axios interceptors
• Admin Access Control — Only admin users can create AI agents and send chat messages
• Markdown Rendering — AI responses rendered in real-time with full Markdown support (tables, code blocks, lists)
• Web Search — Tavily-powered web search tool for agent research
• Skill System — Modular prompt skills (system design, combat, narrative, level design, etc.)
• Auto-Summarization — LLM-generated summaries compress long conversations to stay within context limits

The backend follows Domain-Driven Design with a Maven multi-module layout:

PlayForge/
├── common/           # Shared utilities, constants, exceptions
├── domain/           # Entities, value objects, repository interfaces
├── infrastructure/   # Repository implementations, MyBatis, Redis, JWT, OSS, LLM providers
├── application/      # Application services, use-case orchestration, AI agent factory
├── api/              # REST controllers, request/response DTOs, interceptors, WebSocket handler
├── playforge-start/  # Spring Boot entry point, configuration, Flyway migrations
├── frontend/         # React + TypeScript SPA
└── deploy/           # Docker build & Alibaba Cloud deployment scripts

• Java 25+
• Node.js 20+
• MySQL 8+
• Redis 7+
• Maven 3.9+ (or use the included mvnw wrapper)

git clone https://github.com/BrotherOrange/PlayForge.git
cd PlayForge

cp deploy/env.example .env
# Edit .env with your database, Redis, OSS, and LLM API keys

Key variables:

LLM API keys are optional — providers with missing keys are automatically disabled at startup.

./mvnw spring-boot:run -pl playforge-start

Flyway will automatically run database migrations on startup.

cd frontend
npm install
npm start

The frontend dev server runs on http://localhost:3000 and proxies API requests to the backend on port 8080.

# Build all backend modules
./mvnw clean package

# Run backend tests
./mvnw test

# Type-check frontend
cd frontend && npx tsc --noEmit

# Build frontend for production
cd frontend && npm run build

The project uses a two-step build: deploy/package.sh builds the frontend and backend locally into a fat JAR, then docker build creates a minimal runtime image (Eclipse Temurin 25 JRE).

# 1. Build the fat JAR locally (frontend + backend)
bash deploy/package.sh

# 2. Build the runtime Docker image
docker build -t playforge .

# 3. Run the container
docker run -p 8080:8080 --env-file .env playforge

Deploy scripts are provided in deploy/ for Alibaba Cloud SAE:

./deploy/deploy-all.sh [tag]   # Full pipeline: local build → Docker image → push ACR → deploy SAE
./deploy/package.sh            # Build frontend + backend fat JAR locally
./deploy/build.sh              # Build Docker image only
./deploy/push.sh               # Push image to ACR
./deploy/deploy.sh             # Deploy to SAE

See deploy/env.example for the full list of required environment variables.

This project is licensed under the Apache License 2.0.