Spring AI Hands‑On for Java Developers: Connecting ChatClient to the MiniMax LLM

Project setup

pom.xml includes Spring Boot 4.0.5, Spring AI 2.0.0‑M4, WebFlux starter, and the MiniMax auto‑configuration starter.

<!-- Spring Boot 4.0.5 + Spring AI 2.0.0-M4 -->
<parent>
  <groupId>org.springframework.boot</groupId>
  <artifactId>spring-boot-starter-parent</artifactId>
  <version>4.0.5</version>
</parent>

<dependencies>
  <!-- WebFlux for streaming responses -->
  <dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-webflux</artifactId>
  </dependency>

  <!-- MiniMax auto‑configuration -->
  <dependency>
    <groupId>org.springframework.ai</groupId>
    <artifactId>spring-ai-starter-model-minimax</artifactId>
  </dependency>
</dependencies>

Configuration

Three properties in application.properties are sufficient:

spring.ai.minimax.base-url=api.minimax.chat
spring.ai.minimax.api-key=YOUR_API_KEY
spring.ai.minimax.chat.options.model=minimax-m2.7

No custom configuration class is required; Spring Boot auto‑configuration creates the necessary beans.

Simple chat endpoint

A @RestController exposing GET /minimax/call?question=... builds a ChatClient from the injected MiniMaxChatModel, sends the prompt, and returns the response as a Mono<String>:

@RestController
@RequestMapping("minimax")
public class MiniMaxController {

    @Resource
    private MiniMaxChatModel miniMaxChatModel; // auto‑injected

    @GetMapping("/call")
    public Mono<String> call(@RequestParam("question") String question) {
        ChatClient chatClient = ChatClient.builder(miniMaxChatModel).build();
        return Mono.fromCallable(() ->
            chatClient.prompt(new Prompt(question))
                     .call()
                     .content())
                     .subscribeOn(Schedulers.boundedElastic());
    }
}

Streaming (SSE) response

Endpoint GET /stream/str produces MediaType.TEXT_EVENT_STREAM_VALUE and returns a Flux<String> where each token is emitted separately, enabling a “typewriter” effect:

@GetMapping(value = "/stream/str", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
public Flux<String> streamStr(@RequestParam("question") String question) {
    ChatClient chatClient = ChatClient.create(miniMaxChatModel);
    return chatClient.prompt(question)
                     .stream()
                     .content(); // each character emitted
}

Prompt templates with dynamic parameters

Using the template syntax, variables are written as {variableName} and replaced at runtime via .param(). Example endpoint /call/prompt injects a player parameter into the prompt:

@GetMapping("/call/prompt")
public Mono<String> callPrompt(@RequestParam("player") String player) {
    ChatClient chatClient = ChatClient.create(miniMaxChatModel);
    return Mono.fromCallable(() ->
        chatClient.prompt()
                 .user(u -> u.text("请输出NBA球员{player}的最高得分")
                           .param("player", player))
                 .call()
                 .content())
                 .subscribeOn(Schedulers.boundedElastic());
}

Metadata and advisors

Metadata can be attached to a request, and advisors act as interceptors. The following endpoint adds a metadata pair and a logging advisor:

@GetMapping(value = "/stream/metadata", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
public Flux<String> streamMetadata(@RequestParam("question") String question) {
    ChatClient chatClient = ChatClient.create(miniMaxChatModel);
    return chatClient.prompt()
                     .user(u -> u.text(question).metadata("my name", "why"))
                     .advisors(new SimpleLoggerAdvisor())
                     .stream()
                     .content();
}

Built‑in advisors listed by Spring AI: SimpleLoggerAdvisor – logs request and response. MessageChatMemoryAdvisor – provides multi‑turn conversation memory. QuestionAnswerAdvisor – adds retrieval‑augmented generation.

Multi‑model support

Switching to a different provider only requires changing the injected model bean; the same ChatClient code works unchanged:

// Same ChatClient, different model bean
ChatClient chatClient = ChatClient.create(miniMaxChatModel);   // MiniMax
ChatClient chatClient = ChatClient.create(openAiChatModel);    // OpenAI
ChatClient chatClient = ChatClient.create(deepseekChatModel); // DeepSeek

Source code

Full source code is available at https://github.com/你的github