Implement Persistent Chat Memory in Spring Boot 3 with Spring AI Advisors

Environment

Spring Boot 3.4.2

1. Introduction

When interacting with Chat models (e.g., Wenxin Yiyan, Tongyi Qianwen, Chat‑DeepSeek), memory and history are essential for recalling past user inputs and topics, enabling more accurate responses. Without memory, each conversation is independent.

With conversation memory enabled, the dialogue becomes continuous.

2. Practical Example

2.1 Environment Preparation

We use Alibaba's Chat model.

<code><dependency>
  <groupId>com.alibaba.cloud.ai</groupId>
  <artifactId>spring-ai-alibaba-starter</artifactId>
  <version>1.0.0-M6.1</version>
</dependency></code>

2.2 Conversation Memory

Spring AI provides a default memory Advisor. Add it when building the ChatClient :

<code>var chatClient = ChatClient.builder(chatModel)
    .defaultAdvisors(
        new MyAdvisor(),
        ...
    )
    .build();</code>

Configure two advisors: one for request logging and one for memory.

<code>@Bean
ChatClient chatClient(ChatClient.Builder ccb, ChatMemory cm) {
    List<Advisor> advisors = List.of(
        new SimpleLoggerAdvisor(), // request log
        new MessageChatMemoryAdvisor(cm) // conversation memory
    );
    return ccb.defaultAdvisors(advisors).build();
}</code>

Define a ChatMemory bean (in‑memory implementation):

<code>@Bean
ChatMemory chatMemory() {
    return new InMemoryChatMemory();
}</code>

2.3 Unique Conversation IDs

To avoid sharing the same context across users, pass a unique chatId for each request:

<code>@GetMapping("/{chatId}")
public String chat(String prompt, @PathVariable String chatId) {
    String result = this.chatClient.prompt(prompt)
        .advisors(a -> a
            .param(CHAT_MEMORY_CONVERSATION_ID_KEY, chatId)
            .param(CHAT_MEMORY_RETRIEVE_SIZE_KEY, 100))
        .call()
        .content();
    return result;
}</code>

Now each conversation is isolated.

2.4 Conversation History

Retrieve past messages via ChatMemory :

<code>@RestController
@RequestMapping("/history")
public class ChatHistoryController {
    private final ChatMemory chatMemory;
    public ChatHistoryController(ChatMemory chatMemory) { this.chatMemory = chatMemory; }

    @GetMapping("/{chatId}")
    public ResponseEntity<?> queryChatId(@PathVariable String chatId) {
        List<MessageVO> result = this.chatMemory.get(chatId, 10).stream()
            .map(message -> new MessageVO(parseMessageType(message.getMessageType()), message.getText()))
            .collect(Collectors.toList());
        return ResponseEntity.ok(result);
    }

    public static String parseMessageType(MessageType type) {
        return switch (type) {
            case USER -> "user";
            case ASSISTANT -> "assistant";
            default -> throw new IllegalArgumentException("Unexpected value: " + type);
        };
    }
}
</code>

2.5 Persistence to Database

If you need to store chat history, add the JDBC memory starter:

<code><dependency>
  <groupId>org.springframework.ai</groupId>
  <artifactId>spring-ai-starter-model-chat-memory-jdbc</artifactId>
</dependency></code>

Note: This feature is available from version 1.0.0‑M7.

For the full article, see the accompanying PDF with 116 Spring Boot practical cases.