LangChain4j—聊天记忆 Chat memory(四)

1、测试对话是否有记忆

java 复制代码

package com.atguigu.java.ai.langchain4j;

@SpringBootTest
public class ChatMemoryTest {

    @Autowired
    private Assistant assistant;

    @Test
    public void testChatMemory() {
        String answer1 = assistant.chat("我是环环");
        System.out.println(answer1);
        String answer2 = assistant.chat("我是谁");
        System.out.println(answer2);
    }
}

很显然目前的接入方式，大模型是没有记忆的

2、聊天记忆的简单实现

可以使用下面的的方式实现对话记忆

java 复制代码

@Autowired
private QwenChatModel qwenChatModel;

@Test
public void testChatMemory2() {

//第一轮对话
UserMessage userMessage1 = UserMessage.userMessage("我是环环");
ChatResponse chatResponse1 = qwenChatModel.chat(userMessage1);
AiMessage aiMessage1 = chatResponse1.aiMessage();

//输出大语言模型的回复
System.out.println(aiMessage1.text());

//第二轮对话
UserMessage userMessage2 = UserMessage.userMessage("你知道我是谁吗");
ChatResponse chatResponse2 = qwenChatModel.chat(Arrays.asList(userMessage1,
aiMessage1, userMessage2));
AiMessage aiMessage2 = chatResponse2.aiMessage();

//输出大语言模型的回复
System.out.println(aiMessage2.text());
}

3、使用ChatMemory实现聊天记忆

使用AIService可以封装多轮对话的复杂性，使聊天记忆功能的实现变得简单

java 复制代码

@Test
public void testChatMemory3() {

    //创建chatMemory
    MessageWindowChatMemory chatMemory = MessageWindowChatMemory.withMaxMessages(10);

    //创建AIService
    Assistant assistant = AiServices
        .builder(Assistant.class)
        .chatLanguageModel(qwenChatModel)
        .chatMemory(chatMemory)
        .build();

    //调用service的接口
    String answer1 = assistant.chat("我是环环");
    System.out.println(answer1);
    String answer2 = assistant.chat("我是谁");
    System.out.println(answer2);
}

4、使用AIService实现聊天记忆

创建记忆对话智能体

当AIService由多个组件（大模型，聊天记忆，等）组成的时候，我们就可以称他为 智能体 了

java 复制代码

package com.atguigu.java.ai.langchain4j.assistant;

@AiService(
    wiringMode = EXPLICIT,
    chatModel = "qwenChatModel",
    chatMemory = "chatMemory"
)

public interface MemoryChatAssistant {
    String chat(String message);
}

配置Chat Memory

java 复制代码

package com.atguigu.java.ai.langchain4j.config;

@Configuration
public class MemoryChatAssistantConfig {

@Bean
ChatMemory chatMemory() {
    //设置聊天记忆记录的message数量
    return MessageWindowChatMemory.withMaxMessages(10);
}
}

5、隔离聊天记忆

为每个用户的新聊天或者不同的用户区分聊天记忆

创建记忆隔离对话智能体

java 复制代码

package com.atguigu.java.ai.langchain4j.assistant;

@AiService(
    wiringMode = EXPLICIT,
    chatMemory = "chatMemory",
    chatMemoryProvider = "chatMemoryProvider"
)

public interface SeparateChatAssistant {

/**
* 分离聊天记录
* @param memoryId 聊天id
* @param userMessage 用户消息
* @return
*/
String chat(@MemoryId int memoryId, @UserMessage String userMessage);

}

配置 chatMemoryProvider

java 复制代码

package com.atguigu.java.ai.langchain4j.config;

@Configuration
public class SeparateChatAssistantConfig {

@Bean
ChatMemoryProvider chatMemoryProvider() {
    return memoryId -> MessageWindowChatMemory.builder()
        .id(memoryId)
        .maxMessages(10)
        .build();
}
}