Agent开发

1
@RestController
2
@RequestMapping("/model")
3
public class ChatModelController {
4

5
    @Autowired
6
    private DashScopeChatModel dashScopeChatModel;
7

8
    @RequestMapping("/call/string")
9
    public String callString(String message) {
10
        return dashScopeChatModel.call(message);
11
    }

1
    @RequestMapping("/stream/string")
2
    public Flux<String> callStreamString(String message, HttpServletResponse response) {
3
        response.setCharacterEncoding("UTF-8");
4
        //防止变成乱码
5
        return dashScopeChatModel.stream(message);
6
    }

1
DashScopeChatOptions.builder().withModel("qwen-plus").build()

1
resp.getResult().getOutput().getText()

1
@GetMapping("/promptsEngineer3")
2
public Flux<String> chat3(@RequestParam(value = "message") String message, HttpServletResponse response) {
3
    response.setCharacterEncoding("UTF-8");
4

5
    return chatClient.prompt("请生成包括书名、作者和类别的三本虚构的、非真实存在的中文书籍清单，并以 JSON 格式提供，其中包含以下键:book_id、title、author、genre。").system("你是一个富有创意的作家").user(message).stream().content();
6
}

1
@GetMapping("/chat4")
2
public Flux<String> chat4(@RequestParam(value = "message") String message, HttpServletResponse response) {
3
    response.setCharacterEncoding("UTF-8");
4

5
    return chatClient.prompt("""
6
                执行以下操作：
7
                  step1-用一句话概括下面文本。
8
                  step 2-将摘要翻译成英语。
9
                  step 3-在英语摘要中列出每个人名。
10
                  step 4-输出一个 JSON 对象，其中包含以下键：english_summary，num_names。
11

12
                    请用换行符分隔您的答案。
13
            """).system("你是个ai").user(message).stream().content();
14
}

1
@GetMapping("/chat5")
2
public Flux<String> chat5(@RequestParam(value = "message") String message, HttpServletResponse response) {
3
    response.setCharacterEncoding("UTF-8");
4

5
    return chatClient.prompt("""
6
                一个水果摊有5箱苹果，每箱重15公斤。今天卖掉了35公斤，还剩下多少公斤苹果？
7

8
                                请一步一步思考，并给出最终答案。
9
            """).system("你是个ai").user(message).stream().content();
10
}

1
<dependency>
2
     <groupId>org.springframework.boot</groupId>
3
     <artifactId>spring-boot-starter-webflux</artifactId>
4
</dependency>

1
@GetMapping(value = "/sse/flux")
2
public Flux<String> fluxStream() {
3
    return Flux.interval(Duration.ofSeconds(1))//相隔时间
4
        .take(100)
5
            .map(seq -> "Stream element - " + seq);
6

7
}

1
@Override
2
public String getFormat() {
3
    String template = """
4
            Your response should be in JSON format.
5
            Do not include any explanations, only provide a RFC8259 compliant JSON response following this format without deviation.
6
            Do not include markdown code blocks in your response.
7
            Remove the ```json markdown from the output.
8
            Here is the JSON Schema instance your output must adhere to:
9
            ```%s```
10
            """;
11
                return String.format(template, this.jsonSchema);
12
}

1
public record Book(@JsonPropertyDescription("书籍名称") String title,
2
                   @JsonPropertyDescription("作者") String author,
3
                   @JsonPropertyDescription("书籍介绍") String description,
4
                   @JsonPropertyDescription("价格") BigDecimal price) {
5

6
}

1
  @GetMapping("/bean")
2
    public Book bean(String format) {
3
        BeanOutputConverter<Book> beanOutputConverter = new BeanOutputConverter<>(Book.class);
4
        return chatClient.prompt(new PromptTemplate("你需要推荐一本书，请以{format}格式输出")
5
                .create(Map.of("format", beanOutputConverter.getFormat())))
6
                .system("你是一个专业的图书推荐人员")
7
                .call()
8
                .entity(beanOutputConverter);
9
    }

1
@GetMapping("/chat2")
2
public String chat2(HttpServletResponse response) {
3
    Book book = chatClient.prompt("请帮我推荐几本java相关的书").system("你是一个专业的图书推荐人员").call().entity(Book.class);
4
    return book.toString();
5
}

1
List<Book> result = chatClient.prompt("请帮我推荐几本java相关的书")
2
.system("你是一个专业的图书推荐人员")
3
.call().entity(
4

5
new ParameterizedTypeReference<
6
//抽象类需要重写方法
7
List<Book>>() {
8
  });

1
Map<String, Object> book = chatClient.prompt("请给我推荐几本心理学有关的书，书的内容包括书名、作者、价格、上市时间等信息，以书名作为key，书的信息作为value")
2
        .call().entity(new MapOutputConverter());

1
<!--    chatMemoy基于内存对话记忆-->
2
    <dependency>
3
        <groupId>org.springframework.ai</groupId>
4
        <artifactId>spring-ai-autoconfigure-model-chat-memory</artifactId>
5
        <version>1.1.0</version>
6
    </dependency>

1
ChatClient.builder(chatModel)        .defaultAdvisors(MessageChatMemoryAdvisor.builder(chatMemory).build())

1
.advisors(spec -> spec.param(ChatMemory.CONVERSATION_ID, chatId))

1
// 伪代码
2
history = chatMemory.get(chatId);
3
prompt = history + 当前用户消息;
4
answer = model.generate(prompt);
5
chatMemory.save(chatId, 用户消息, answer);
6

7
请求 -> ChatClient -> MessageChatMemoryAdvisor -> ChatMemory -> ChatMemoryRepository -> 存储后端

1
dependency>
2
    <groupId>org.springframework.ai</groupId>
3
    <artifactId>spring-ai-starter-model-chat-memory-repository-jdbc</artifactId>
4
    <version>1.1.0</version>
5
</dependency>

1
spring:
2
  application:
3
    name: demo
4
  ai:
5
    dashscope:
6
      api-key: sk-ws-H.EHMYHYX.VeUc.MEUCIQDyLgMI5VSHAOyUHnbblT4muC2q2DhOXkCPY9OY_a9GCAIga6OiX6wD-bOZtgzy4iGErxOZiP4E_QPbJtUcEPGYYr4
7
    chat:
8
        memory:
9
          repository:
10
            jdbc:
11
              platform: mysql
12
              initialize-schema: always

1
CREATE TABLE `spring_ai_chat_memory` (
2
  `conversation_id` varchar(36) CHARACTER SET utf8mb4 NOT NULL,
3
  `content` text CHARACTER SET utf8mb4 NOT NULL,
4
  `type` enum('USER','ASSISTANT','SYSTEM','TOOL') CHARACTER SET utf8mb4 NOT NULL,
5
  `timestamp` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
6
  KEY `SPRING_AI_CHAT_MEMORY_CONVERSATION_ID_TIMESTAMP_IDX` (`conversation_id`,`timestamp`)
7
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_general_ci
8
;

1
<dependency>
2
    <groupId>mysql</groupId>
3
    <artifactId>mysql-connector-java</artifactId>
4
    <version>8.0.33</version>
5
</dependency>

1
  datasource:
2
    url: jdbc:mysql://localhost:3306/springai?useUnicode=true&characterEncoding=UTF-8  # 记得改成你自己的
3
    username: root  # 记得改成你自己的
4
    password: 123456  # 记得改成你自己的
5
    driver-class-name: com.mysql.cj.jdbc.Driver

1
@Configuration
2
public class JdbcChatMemoryConfiguration {
3

4
    @Bean
5
    public ChatMemory jdbcChatMemory(JdbcChatMemoryRepository jdbcChatMemoryRepository) {
6
        return MessageWindowChatMemory.builder().chatMemoryRepository(jdbcChatMemoryRepository).maxMessages(20).build();
7
    }
8
}

1
<dependency>
2
    <groupId>org.springframework.ai</groupId>
3
    <artifactId>spring-ai-starter-model-chat-memory-repository-cassandra</artifactId>
4
</dependency>

1
<dependency>
2
    <groupId>org.springframework.ai</groupId>
3
    <artifactId>spring-ai-starter-model-chat-memory-repository-neo4j</artifactId>
4
</dependency>

1
.defaultAdvisors(MessageChatMemoryAdvisor.builder(chatMemory).build())

1
.defaultAdvisors(
2
                        new SimpleLoggerAdvisor()

1
public interface Advisor extends Ordered {
2
    int DEFAULT_CHAT_MEMORY_PRECEDENCE_ORDER = -2147482648;
3
    String getName();
4
}

1
@Override
2
public int getOrder() {
3
    return Ordered.LOWEST_PRECEDENCE;

1
ChatClientRequest processedRequest = before(chatClientRequest, callAdvisorChain);
2

3
ChatClientResponse response = callAdvisorChain.nextCall(processedRequest);
4

5
return after(response, callAdvisorChain);

1
用户这次的问题 + 历史对话记录
2
        ↓
3
组合成新的 Prompt
4
        ↓
5
再交给后面的 advisor / ChatModel

1
before 阶段：保存 user message
2
after 阶段：保存 assistant message
3
before：读取历史记忆 + 拼进本次请求 + 保存用户消息
4
after：提取模型回复 + 保存助手消息

1
      <dependency>
2
            <groupId>org.springframework.ai</groupId>
3
            <artifactId>spring-ai-ollama</artifactId>
4
            <version>1.1.0</version>
5
        </dependency>

1
<dependency>
2
     <groupId>org.springframework.ai</groupId>
3
     <artifactId>spring-ai-autoconfigure-model-ollama</artifactId>
4
     <version>1.1.0</version>
5
 </dependency>

1
@Autowired
2
@Qualifier("ollamaChatModel")
3
private ChatModel ollamaChatModel;