Technical Architecture

NiteMoon AI Platform adopts a modern microservices architecture design, supporting both monolith and microservices dual-mode deployment, providing a stable and scalable technical foundation for enterprise AI applications.

System Architecture Diagram

┌─────────────────────────────────────────────────────────────┐
│                      Client Layer                            │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐   │
│  │ Web UI   │  │ Mini App │  │ Mobile   │  │  API     │   │
│  └──────────┘  └──────────┘  └──────────┘  └──────────┘   │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                    API Gateway                               │
│              Spring Cloud Gateway : 7770                    │
│                    Nacos Service Discovery                   │
└─────────────────────────────────────────────────────────────┘
                              │
      ┌───────────────────────┼───────────────────────┐
      │                       │                       │
      ▼                       ▼                       ▼
┌──────────┐           ┌──────────┐           ┌──────────┐
│   Auth   │           │   Biz    │           │ Monitor  │
│  Service │           │  Service │           │ Service  │
└──────────┘           └──────────┘           └──────────┘
                              │
      ┌───────────────────────┼───────────────────────┐
      │                       │                       │
      ▼                       ▼                       ▼
┌──────────┐           ┌──────────┐           ┌──────────┐
│   LLM    │           │Knowledge │           │ Digital  │
│  Service │           │  Graph   │           │  Human   │
└──────────┘           └──────────┘           └──────────┘
      │                       │                       │
      ▼                       ▼                       ▼
┌──────────┐           ┌──────────┐           ┌──────────┐
│ Training │           │   RAG    │           │   AI     │
│  Service │           │  Service │           │  Excel   │
└──────────┘           └──────────┘           └──────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                      Data Layer                              │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐   │
│  │  MySQL   │  │  Redis   │  │  Neo4j   │  │  Milvus  │   │
│  └──────────┘  └──────────┘  └──────────┘  └──────────┘   │
└─────────────────────────────────────────────────────────────┘

Tech Stack Details

Core Framework

Layer	Technology	Version	Description
Language	Java	17	LTS version
Framework	Spring Boot	3.4.7	Microservices framework
Cloud	Spring Cloud	2024.0.0	Cloud-native support
Cloud Alibaba	Spring Cloud Alibaba	2023.0.3.2	Alibaba ecosystem
RPC	Apache Dubbo	3.2.14	High-performance RPC

AI / LLM

Technology	Version	Description
LangChain4j	1.0.0-beta1	LLM application framework
HanLP	latest	Chinese NLP
Apache Tika	2.9.1	Document parsing
ONNX Runtime	latest	Model inference
OpenCV	latest	Image processing

Data Storage

Technology	Version	Description
MySQL	9.2.0	Relational database
Neo4j	5.27.0	Graph database
Redis	6.0+	Cache and vector store
Milvus	latest	Vector database
MinIO	latest	Object storage

Middleware

Technology	Version	Description
RocketMQ	5.2.0	Message queue
Sa-Token	1.44.0	Authentication & authorization
MyBatis-Plus	3.5.12	ORM framework
Redisson	3.38.1	Redis client
Sentinel	latest	Flow control

Utility Libraries

Technology	Version	Description
Hutool	5.8.33	Java utility library
Fastjson2	2.0.57	JSON processing
OkHttp	4.12.0	HTTP client
Knife4j	4.6.0	API documentation

Deployment Modes

Monolith Mode (boot profile)

Suitable for development, testing, and small-scale deployment scenarios.

Architecture Characteristics:

• Single JVM deployment
• Local method invocation
• No infrastructure dependencies like Nacos
• Simple and fast

Deployment Requirements:

• Java 17+
• MySQL 8.0+
• Redis 6.0+
• Memory: 4GB+

Startup Command:

java -jar nitemoon-boot.jar --spring.profiles.active=boot

Microservices Mode (cloud profile)

Suitable for production environments and large-scale deployment scenarios.

Architecture Characteristics:

• Multi-service independent deployment
• Dubbo RPC communication
• Nacos service discovery and configuration
• Sentinel flow control
• Seata distributed transactions

Deployment Requirements:

• Java 17+
• MySQL 8.0+
• Redis 6.0+
• Neo4j 5.0+
• Nacos 2.0+
• RocketMQ 5.0+
• Memory: 8GB+ (per service)

Service List:

Service	Port	Description
gateway	7770	API Gateway
auth	7771	Authentication Service
llm	-	LLM Service
onnx	-	ONNX Inference Service
upms	-	User Permission Service

Module Dependencies

nitemoon-cloud (root)
├── nitemoon-gateway          # API Gateway
├── nitemoon-auth             # Authentication Service
├── nitemoon-common           # Common Modules
│   ├── common-core           # Core Utilities
│   ├── common-security       # Security Module
│   ├── common-mybatis        # ORM Module
│   ├── common-redis          # Cache Module
│   ├── common-sse            # SSE Module
│   ├── common-storage        # Storage Module
│   └── ...                   # Other Common Modules
├── nitemoon-api              # Service Interfaces
│   ├── nitemoon-llm-api      # LLM Interface
│   ├── nitemoon-kg-api       # Knowledge Graph Interface
│   ├── nitemoon-human-api    # Digital Human Interface
│   └── ...                   # Other Interfaces
├── nitemoon-biz              # Business Modules
│   ├── nitemoon-llm          # LLM Service
│   ├── nitemoon-kg           # Knowledge Graph Service
│   ├── nitemoon-human        # Digital Human Service
│   ├── nitemoon-training     # Training Service
│   ├── nitemoon-onnx         # Inference Service
│   ├── nitemoon-aiexcel      # AI Excel Service
│   ├── nitemoon-generator    # Code Generator
│   ├── nitemoon-monitor      # Monitor Service
│   └── nitemoon-upms         # User Permission Service
└── nitemoon-boot             # Monolith Startup Module

Design Patterns

Strategy Pattern

Used for abstraction of LLM providers, TTS providers, storage backends, and similar scenarios.

public interface ModelBuildHandler {
    ChatModel buildChatModel(LLMModelConfig config);
    EmbeddingModel buildEmbeddingModel(LLMModelConfig config);
}

Factory Pattern

Used for creating model instances, storage instances, and more.

@Component
public class ModelStoreFactory {
    public ChatModel createChatModel(String provider, String modelName) {
        ModelBuildHandler handler = getHandler(provider);
        return handler.buildChatModel(loadConfig(modelName));
    }
}

Observer Pattern

Used for event-driven cache refresh, configuration updates, and more.

@Component
public class ProviderRefreshListener implements ApplicationListener<ProviderRefreshEvent> {
    @Override
    public void onApplicationEvent(ProviderRefreshEvent event) {
        // Refresh provider cache
    }
}

Performance Optimization

Caching Strategy

• Redis cache for hot data
• Local cache for frequently accessed data
• Multi-level cache architecture

Async Processing

• RocketMQ asynchronous messaging
• CompletableFuture async invocation
• Thread pool isolation

Connection Pooling

• HikariCP database connection pool
• Redisson connection pool
• OkHttp connection pool

Related Links

• Product Overview
• Quick Start
• API Documentation