Multi-Model LLM Integration

NiteMoon AI Platform connects to 15+ mainstream large language models through a unified abstraction layer, supporting text chat, embedding generation, image generation, web search, and other model types, allowing users to leverage various LLM capabilities without worrying about underlying implementation details.

Key Features

• 15+ LLM providers: Covering mainstream large models both domestically and internationally
• 4 model types: CHAT, EMBEDDING, TEXT_IMAGE, WEB_SEARCH
• Strategy pattern abstraction: Unified interface with dynamic switching
• Dynamic model configuration: Database-driven configuration with no restart required

Supported LLM Providers

International Providers

Provider	Model Examples	Features
OpenAI	GPT-4, GPT-3.5	World-leading LLM
Azure OpenAI	GPT-4	Enterprise deployment
Claude	Claude 3	Long-text understanding
Gemini	Gemini Pro	Multimodal capabilities
Ollama	Llama, Mistral	Local deployment

Domestic Providers

Provider	Model Examples	Features
Qwen (Alibaba)	Tongyi Qianwen	Chinese-optimized
Qianfan (Baidu)	ERNIE Bot	Chinese-optimized
Zhipu AI	ChatGLM	Chinese-optimized
DeepSeek	DeepSeek-V2	High cost-effectiveness
Spark (iFlytek)	Spark	Chinese-optimized
Yi (01.AI)	Yi-Large	Long text
Douyin (ByteDance)	Doubao	Chinese-optimized
Gitee AI	Various models	Open-source models
Silicon	Various models	Open-source models
MiMo	MiMo	Chinese-optimized

Model Types

CHAT - Text Chat

Used for multi-turn conversations, Q&A, content generation, and similar scenarios.

ChatModel model = modelStoreFactory.createChatModel(provider, modelName);
String response = model.generate("Hello, please introduce yourself");

EMBEDDING - Embedding Generation

Used for text vectorization, supporting RAG knowledge base, semantic search, and similar scenarios.

EmbeddingModel model = modelStoreFactory.createEmbeddingModel(provider, modelName);
float[] embedding = model.embed("This is a test text");

TEXT_IMAGE - Image Generation

Used for text-to-image generation scenarios.

ImageModel model = modelStoreFactory.createImageModel(provider, modelName);
byte[] image = model.generate("A cute cat");

WEB_SEARCH - Web Search

Used for real-time web search scenarios.

WebSearchModel model = modelStoreFactory.createWebSearchModel(provider, modelName);
List<SearchResult> results = model.search("Today's weather");

Technical Implementation

Strategy Pattern

Each LLM provider has a corresponding ModelBuildHandler implementation:

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

Factory Pattern

Model instances are created uniformly through ModelStoreFactory:

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

Dynamic Configuration

Model configurations are stored in the database and support dynamic updates:

CREATE TABLE llm_model (
    id BIGINT PRIMARY KEY,
    provider VARCHAR(50),
    model_name VARCHAR(100),
    api_key VARCHAR(500),
    api_endpoint VARCHAR(500),
    config JSON
);

Use Cases

Intelligent Conversation

// Create chat model
ChatModel model = modelStoreFactory.createChatModel("openai", "gpt-4");

// Multi-turn conversation
ChatMemory memory = new MessageWindowChatMemory(10);
ChatAssistant assistant = AiServices.builder(ChatAssistant.class)
    .chatModel(model)
    .chatMemory(memory)
    .build();

String response = assistant.chat("Hello");

RAG Knowledge Base

// Create embedding model
EmbeddingModel embeddingModel = modelStoreFactory.createEmbeddingModel("openai", "text-embedding-3-small");

// Text vectorization
float[] embedding = embeddingModel.embed("This is a test text");

Content Generation

// Create image model
ImageModel imageModel = modelStoreFactory.createImageModel("openai", "dall-e-3");

// Generate image
byte[] image = imageModel.generate("A cute cat sitting on the moon");

Configuration Examples

OpenAI Configuration

llm:
  providers:
    openai:
      api-key: ${OPENAI_API_KEY}
      base-url: https://api.openai.com/v1
      models:
        - name: gpt-4
          type: CHAT
        - name: text-embedding-3-small
          type: EMBEDDING

Ollama Local Configuration

llm:
  providers:
    ollama:
      base-url: http://localhost:11434
      models:
        - name: llama3
          type: CHAT
        - name: nomic-embed-text
          type: EMBEDDING

Related Links

• Visual Workflow Engine
• RAG Knowledge Base
• Quick Start