How DeepSeek V3-0324 Boosts Java AI Apps with Function Calling

DeepSeek V3-0324 Release Highlights

DeepSeek has quietly released the DeepSeek‑V3‑0324 model, delivering a large performance jump and new practical features such as function calling and standardized JSON output, addressing the shortcomings of the previous R1 version.

Chinese Writing Ability Test

Prompt: Assume you are an environmental expert who discovered a new sustainable energy technology. Describe its principle, advantages, and environmental impact.

The model shows a significant improvement in Chinese writing, as illustrated by the example “左脚踩右脚”.

Frontend Development Test

Request: “Implement a website login page .html without technology constraints.”

The response demonstrates the model’s ability to generate frontend code.

DeepSeek V3 Function Calling

Function calling enables the large language model to generate structured requests that specify a function name and required parameters, allowing seamless integration with external tools and APIs.

For example, a mathematical tool library can be invoked so the model indicates which function to call and with what arguments; the system executes the function and returns the result to the model.

Note: the model does not execute the function itself; it only suggests the call.

Practical Example

Add Dependency

<code>&lt;dependency&gt;
  &lt;groupId&gt;dev.langchain4j&lt;/groupId&gt;
  &lt;artifactId&gt;langchain4j-open-ai&lt;/artifactId&gt;
  &lt;version&gt;1.0.0-beta2&lt;/version&gt;
&lt;/dependency&gt;
</code>

Create AI Service Interface

<code>public interface FunctionAssistant {
    String chat(String message);
}
</code>

Implement Business Logic with Annotations

<code>public class InvoiceHandler {
    @Tool("Generate invoice based on user‑submitted information")
    public String handle(String companyName, String dutyNumber,
                         @P("Amount with two decimal places") String amount) {
        log.info("companyName >>>> {} dutyNumber >>>> {} amount >>>> {}", companyName, dutyNumber, amount);
        return "Invoice generated";
    }
}
</code>

Test Invocation

<code>@Test
public void test2() {
    ChatLanguageModel chatLanguageModel = OpenAiChatModel.builder()
        .apiKey("sk-")
        .modelName("deepseek-chat")
        .baseUrl("https://api.deepseek.com/v1")
        .logRequests(true)
        .logResponses(true)
        .build();

    FunctionAssistant assistant = AiServices.builder(FunctionAssistant.class)
        .chatLanguageModel(chatLanguageModel)
        .tools(new InvoiceHandler())
        .build();

    String chat = assistant.chat("Help me issue an invoice for Shenzhen Tencent Co., tax ID 91440300734991234A, amount 100.123456");
    System.out.println(chat);
}
</code>

Summary

From recent AI Chat2BI research, V3‑0324 shows higher playability than R1, and in our tests it uniquely supports function‑based Chat2BI among domestic models, outperforming most tuning methods. The update bridges many gaps, allowing AI to integrate deeper into business workflows.

Only AI that truly empowers business creates real value.