[Open-source] Java AI Intelligent Body Full-scene Application Development Framework Supports Various Capabilities of Known AI Development

Open SourceOpen Source Systems

2026-06-08 22:41 7 次浏览 • 0 次下载 100KB

[Open-source] Java AI Intelligent Body Full-scene Application Development Framework Supports Various Capabilities of Known AI Development

1. Project Overview

Solon-AI

Java AI (intelligent body) Full-scene application development framework (supports various capabilities of known AI development)
[Based on Solon application development framework]

Java AI (Intelligent Entity) Full-scene Application Development Framework (LLM, Function Call, RAG, Embedding, Reranking, Flow, MCP Server, Mcp Client, Mcp Proxy). It is also compatible with java8 ~ java24.

Full-scene Java AI Application Development Framework (Supporting Various Capabilities of Known AI Development). It is part of the Solon project. It can also be embedded in SpringBoot2, jFinal, Vert.x, etc. frameworks for use.

2. Open Source License

Licensed under Apache-2.0

3. Screenshots

4. Feature Overview

(v3.1.0 supports)

This series mainly introduces the use of Solon AI plugins (AI 'General' Application Development Framework). Solon-AI is designed with dialect adaptation, which can be compatible with various large language model (LLM) interface calls.

Here, the AI mainly refers to Generative Artificial Intelligence. It is also called 'Large Model', or 'Large Language Model'. By dividing the generation content, there are commonly:

Model	Function
Chat Generation Model (ChatModel)	Used to generate chat dialogues, or text creation, or explanation and answer questions, etc.
Image Generation Model (ImageModel)	Used to generate images
Video Generation Model (VideoModel)	Used to generate videos

Other models will also have (more, not listed):

Model	Function
Embedding Model (EmbeddingModel)	Used to generate vector data, and then implement similar queries
Ranking Model (RankingModel)	Used to sort

Different models will have mutual cooperation. For example, ChatModel generated content can be used for ImageModel and VideoModel. While EmbeddingModel and ChatModel cooperate, they can implement RAG (i.e. local data and large model cooperation, enhance generation effect). It also supports MCP protocol, implements Tool service publication, and Tool service use.

When used, it can be roughly thought of as an http-api interface (more human-friendly). See common terms, which are almost the same meaning:

ai
Large Model
Generative Large Language Model
Large Language Model
Language Model

Current AI common application construction (solon-flow provides 'Workflow Arrangement' support)

Source: Solon-AI Official Documentation

5. Tech Stack

Main Interface Experience Examples

ChatModel (Common Interface, based on dialect adaptation to implement different providers and models extension)

ChatModel chatModel = ChatModel.of("http://127.0.0.1:11434/api/chat")
                .provider("ollama") // Need to specify the supplier, used to identify the interface style (also called dialect)
                .model("qwen2.5:1.5b")
                .build();
// Synchronous call, and print response message
System.out.println(chatModel.prompt("hello").call().getMessage());
// Reactive call
chatModel.prompt("hello").stream(); //Publisher<ChatResponse>

Function Calling (or Tool Calling)

// Can add default tools (i.e. all requests have products), or request time tools
chatModel.prompt("What is the weather situation in Hangzhou today?")
    .options(op->op.toolsAdd(new FunctionTools()))
    .call();

Vision (Multimedia Perception)

chatModel.prompt(ChatMessage.ofUser("Is there a square in this picture?", Image.ofUrl(imageUrl)))
    .call();

RAG（EmbeddingModel，Repository，DocumentLoader，RerankingModel）

// Build knowledge base
EmbeddingModel embeddingModel = EmbeddingModel.of(apiUrl).apiKey(apiKey).provider(provider).model(model).batchSize(10).build();
RerankingModel rerankingModel = RerankingModel.of(apiUrl).apiKey(apiKey).provider(provider).model(model).build();
InMemoryRepository repository = new InMemoryRepository(TestUtils.getEmbeddingModel()); // 3. Initialize knowledge base
repository.insert(new PdfLoader(pdfUri).load());
// Search
List<Document> docs = repository.search(query);
// If needed, can re-arrange
docs = rerankingModel.rerank(query, docs);
// Prompt phrase enhancement is
ChatMessage message = ChatMessage.augment(query, docs);
// Call large model
chatModel.prompt(message)
    .call();

Ai Flow (Simulate Dify's Workflow Application)

id: demo1
layout:
  - title: "Start"
    type: start
  - title: "File Extraction"
    meta.input: "file" # Visual interface configuration (through meta information representation)
    meta.output: "fileTxt"
    task: @FileLoaderCom
  - title: "LLM"
    meta.model: "Qwen/Qwen2.5-72B-Instruct" # Visual interface configuration (through meta information representation)
    meta.input: "fileTxt"
    meta.messages:
      - role: system
        content: "#Role
You are a data expert, delete data format and conversion

#Context
${fileTxt}

#Task
Extract csv format string"
    task: @ChatModelCom
  - title: "Parameter Extractor"
    meta.model: "Qwen/Qwen2.5-72B-Instruct" # Visual interface configuration (through meta information representation)
    meta.output: "csvData"
    task: @ParamExtractionCom
  - title: "Execute Code"
    meta.input: "csvData"
    task: |
      import com.demo.DataUtils;

      String json = DataUtils.csvToJson(node.meta().get("meta.input"));  // Convert to json data
      String echatCode = DataUtils.jsonAsEchatCode(json); // Convert to echart chart code
      context.result = echatCode; // As a result return
  - title: "End"
    type: end
# FlowEngine flowEngine = FlowEngine.newInstance();
# ...
# flowEngine.eval("demo1");

MCP server (Supports Multi-Endpoint)

// Component way construction
@McpServerEndpoint(name="mcp-case1", sseEndpoint = "/case1/sse")
public class McpServerTool {
    @ToolMapping(description = "Query Weather Forecast")
    public String getWeather(@ToolParam(description = "City Location") String location) {
        return "Sunny, 14 degrees";
    }
}
// Native java way construction
McpServerEndpointProvider serverEndpoint = McpServerEndpointProvider.builder()
        .name("mcp-case2")
        .sseEndpoint("/case2/sse")
        .build();
serverEndpoint.addTool(new MethodToolProvider(new McpServerTool()));
serverEndpoint.postStart();

MCP client

McpClientToolProvider clientToolProvider = McpClientToolProvider.builder()
                .apiUrl("http://localhost:8080/case1/sse")
                .build();
String rst = clientToolProvider.callToolAsText("getWeather", Map.of("location", "Hangzhou"));

下载地址免费

资源直链https://gitee.com/opensolon/solon-ai?url=eflyos.zip

去下载