⁷

⁴

Spring AI 工具调用回调与流式前端展示的完整落地方案

leikooo

2026-01-20 17:25

阅读1623

leikooo

2026-01-20 17:25

阅读 1623

那些坑

用 SpringAI 重写零代码生成时，前端展示工具调用这件事把我卡住了。

Langchain4j 写回调多舒服啊

▼
java
复制代码
public interface StreamingChatResponseHandler {  
    default void onPartialToolCall(PartialToolCall partialToolCall) {}  
    default void onPartialToolCall(PartialToolCall partialToolCall, PartialToolCallContext context) {}  
    default void onCompleteToolCall(CompleteToolCall completeToolCall) {}  
    void onCompleteResponse(ChatResponse completeResponse);  
    void onError(Throwable error);  
}

再看看 @ToolMemoryId，直接往方法参数里一扔，conversationId 就到手了，多省心：

▼
java
复制代码
class Tools {  
    @Tool  
    String addCalendarEvent(CalendarEvent event, @ToolMemoryId memoryId) {  
        // memoryId 直接能用  
    }  
}

SpringAI 呢？这些它都没有（也有可能是我没找到）。adviseStream 工具调用的时候也感知不到

为什么一定要 conversationId？

主要有下面几点

生成的代码需要区分目录，方便管理
隔离每个单独 APP 生成的路径
记录工具调用次数（后续分析用）

整体思路

用户发请求 → Ai2ChatClient 接收 → SpringAI 处理 → 切面拦截工具调用 → 事件发布 → 实时推给前端

就这么几条线。核心其实就三件事：切面拦截、事件发布、流合并。

AOP 依赖

▼
xml
复制代码
<dependency>    
	<groupId>org.springframework.boot</groupId>    
    <artifactId>spring-boot-starter-aop
    </artifactId>  
</dependency>

举个例子：TodoList 工具

这是我们项目里实际在用的工具类：

具体提示词参考的是 OpenCode 的 https://github.com/anomalyco/opencode/blob/dev/packages/opencode/src/tool/todoread.txt

▼
java
复制代码
@Component
public class TodolistTools extends BaseTools {

    /**
     * cache
     */
    private static final Cache<String, String> TODOLIST_CACHE = Caffeine.newBuilder()
            .maximumSize(10_00)
            .expireAfterWrite(Duration.ofMinutes(30))
            .build();

    @Tool(description = "Write or update the todo list for current task. " +
            "Use this to track progress and plan remaining work. " +
            "Each todo item should be a clear, actionable task. " +
            "Format: numbered list with status markers like [ ], [x], [>], [-]. " +
            "Status meanings: [ ] pending, [x] completed, [>] in progress, [-] blocked/cancelled."
    )
    public String todoWrite(
            @ToolParam(description = "The todo list content to save. Format as a structured list with status indicators.")
            String todoContent,
            ToolContext toolContext
    ) {
        String conversationId = ConversationIdUtils.getConversationId(toolContext);
        if (StringUtils.isBlank(todoContent)) {
            TODOLIST_CACHE.invalidate(conversationId);
            return "Todo list cleared.";
        }
        TODOLIST_CACHE.put(conversationId, todoContent);
        return "Todo list saved successfully.\n\nCurrent todo list:\n" + todoContent;
    }

    @Tool(description = "Read the current todo list for this conversation. " +
            "Use this to check progress and see what tasks remain. " +
            "Returns an empty message if no todo list exists yet."
    )
    public String todoRead(ToolContext toolContext) {
        String conversationId = ConversationIdUtils.getConversationId(toolContext);
        String todoContent = TODOLIST_CACHE.getIfPresent(conversationId);
        if (StringUtils.isBlank(todoContent)) {
            return "No todo list for this conversation.";
        }
        return "Current todo list:\n" + todoContent;
    }

    @Override
    String getToolName() {
        return "Todo List Tool";
    }

    @Override
    String getToolDes() {
        return "Read and write task todo lists to track progress";
    }
}

切面是怎么工作的

SpringAI 没给我们留回调接口，那就自己造一个。切面这东西好就好在不改动原代码，加个注解就能生效。

我们用 @Before 抓工具调用开始的那一刻，用 @AfterReturning 抓调用结束的那一刻。工具类丢给 Spring 容器，切面自己就找上门来了。

▼
java
复制代码
@Aspect
@Component
@Slf4j
public class ToolContextAspect {

    private final ToolEventPublisher toolEventPublisher;

    public ToolContextAspect(ToolEventPublisher toolEventPublisher) {
        this.toolEventPublisher = toolEventPublisher;
    }

    @Pointcut("execution(* com.leikooo.codemother.ai.tools..*.*(..)) && @annotation(org.springframework.ai.tool.annotation.Tool)")
    public void anyToolExecution() {

    }

    @Before("anyToolExecution()")
    public void beforeToolCall(JoinPoint joinPoint) {
        ToolContext toolContext = getToolContext(joinPoint);
        String className = joinPoint.getTarget().getClass().getSimpleName();
        String methodName = joinPoint.getSignature().getName();
        if (Objects.isNull(toolContext)) {
            log.warn("SKIPPED: Tool method [{}.{}] was called but does not accept ToolContext as a parameter.",
                    className, methodName);
            return;
        }
        handleToolContext(toolContext, className, methodName, null, true);
    }

    @AfterReturning(pointcut = "anyToolExecution()", returning = "result")
    public void afterToolCall(JoinPoint joinPoint, Object result) {
        ToolContext toolContext = getToolContext(joinPoint);
        String className = joinPoint.getTarget().getClass().getSimpleName();
        String methodName = joinPoint.getSignature().getName();
        if (toolContext != null) {
            handleToolContext(toolContext, className, methodName, result, false);
        }
    }

    private void handleToolContext(ToolContext context, String className, String methodName, Object result, boolean isBefore) {
        Message message = context.getToolCallHistory().getLast();
        AssistantMessage.ToolCall toolCallInfo = ((AssistantMessage) message).getToolCalls().getLast();
        String toolCallId = toolCallInfo.id();
        String sessionId = ConversationIdUtils.getConversationId(context);
        if (isBefore) {
            toolEventPublisher.publishToolCall(sessionId, className, methodName, toolCallId);
        } else {
            toolEventPublisher.publishToolResult(sessionId, className, methodName, toolCallId, result);
        }
    }


    /**
     * toolContext
     * @param joinPoint joinPoint
     * @return ToolContext
     */
    private ToolContext getToolContext(JoinPoint joinPoint) {
        Object[] args = joinPoint.getArgs();
        ToolContext toolContext = null;
        for (Object arg : args) {
            if (arg instanceof ToolContext) {
                toolContext = (ToolContext) arg;
                break;
            }
        }
        return toolContext;
    }
}

测试结果

事件发布：把消息送出去

这里用到了 Project Reactor 的 Sinks。每个会话一个 Sink，多线程环境下也能正常工作。

▼
java
复制代码
@Component  
public class ToolEventPublisher {    
    private final Map<String, Sinks.Many<ToolEvent>> sinks = new ConcurrentHashMap<>();    
    
    private Sinks.Many<ToolEvent> getSink(String sessionId) {    
        return sinks.computeIfAbsent(sessionId, k -> Sinks.many().multicast().onBackpressureBuffer());    
    }    
    
    public void publishToolCall(String sessionId, String toolName, String methodName, String toolCallId) {    
        getSink(sessionId).tryEmitNext(new ToolEvent(sessionId, "tool_call", toolName, methodName, toolCallId, null));    
    }    

    public void publishToolResult(String sessionId, String toolName, String methodName, String toolCallId, Object result) {    
        getSink(sessionId).tryEmitNext(new ToolEvent(sessionId, "tool_result", toolName, methodName, toolCallId, result));    
    }    

    public Flux<ToolEvent> events(String sessionId) {    
        return getSink(sessionId).asFlux();    
    }    

    public void complete(String sessionId) {    
        Sinks.Many<ToolEvent> sink = sinks.remove(sessionId);    
        if (sink != null) sink.tryEmitComplete();    
    }    

    public record ToolEvent(String sessionId, String type, String toolName, String methodName, String toolCallId, Object result) {} 
}

流怎么合并到主响应里

这里有两种玩法

玩法一：自己动手丰衣足食

直接在业务方法里把两个流 merge 起来。好处是代码都在明面上，坏处是每个方法都得写一遍。

一个小细节：mainFlux 结束时会触发 doFinally，但 toolEventFlux 不会。所以必须在 doFinally 里手动调用 complete 关掉事件流。否则这个流会一直挂在那儿，等不到终点。
前端就会一直这样：

▼
java
复制代码
@Component
public class Ai2ChatClient {
    private final ChatClient chatClient;
    private ToolEventPublisher toolEventPublisher;

    public Ai2ChatClient(ChatModel openAiChatModel,
                         TodolistTools todolistTools, ToolAdvisor toolAdvisor,
                         ToolEventPublisher toolEventPublisher) {
        this.toolEventPublisher = toolEventPublisher;
        this.chatClient = ChatClient
                .builder(openAiChatModel)
                .defaultTools(todolistTools)
                .build();
    }


    public Flux<String> chat2Ai(String msg, String appId) {
        Flux<String> mainFlux = chatClient.prompt()
                .system("""
                        You are a helpful, precise, and reliable AI assistant.
                        Respond clearly and concisely.
                        Prioritize correctness, safety, and practicality.
                        If information is uncertain, state the uncertainty explicitly.
                        
                        """)
                .user(msg)
                .advisors(advisorSpec -> advisorSpec.param(CONVERSATION_ID, appId))
                .toolContext(Map.of(CONVERSATION_ID, appId))
                .stream().content()
                .doFinally(s -> toolEventPublisher.complete(appId));
        Flux<String> toolEventFlux = toolEventPublisher.events(appId)
                .map(event -> {
                    Object result = Optional.ofNullable(event.result()).orElse("");
                    String message = switch (event.type()) {
                        case "tool_call" -> String.format("%s: %s", "正在进行工具调用", result);
                        case "tool_result" -> String.format("%s: %s", "工具调用完成", result);
                        default -> "";
                    };
                    return String.format("\n\n[选择工具] %s \n\n", message);
                });
        // 合并流
        return Flux.merge(mainFlux, toolEventFlux);
    }
}

玩法二：把脏活累活扔给 Advisor

写个 StreamAdvisor，让它自己处理流合并。业务代码瞬间清爽了。

▼
java
复制代码

/**
 * @author <a href="https://github.com/lieeew">leikooo</a>
 * @date 2025/12/31
 * @description
 */
@Slf4j
@Component
public class ToolAdvisor implements CallAdvisor, StreamAdvisor {

    private final ToolEventPublisher toolEventPublisher;

    public ToolAdvisor(ToolEventPublisher toolEventPublisher) {
        this.toolEventPublisher = toolEventPublisher;
    }

    @Override
    public ChatClientResponse adviseCall(ChatClientRequest chatClientRequest, CallAdvisorChain callAdvisorChain) {
        return callAdvisorChain.nextCall(chatClientRequest);
    }

    @Override
    public Flux<ChatClientResponse> adviseStream(ChatClientRequest chatClientRequest, StreamAdvisorChain streamAdvisorChain) {
        String appId = ConversationIdUtils.getConversationId(chatClientRequest.context());
        Flux<ChatClientResponse> toolEventFlux = getToolEventFlux(appId);
        Flux<ChatClientResponse> mainFlux = streamAdvisorChain.nextStream(chatClientRequest)
                .doFinally(signalType -> toolEventPublisher.complete(appId));
        return Flux.merge(mainFlux, toolEventFlux);
    }

    @Override
    public String getName() {
        return "ToolAdvisor";
    }

    @Override
    public int getOrder() {
        return Integer.MIN_VALUE + 100;
    }

    /**
     * 工具调用推送流
     * @param sessionId sessionId
     * @return flux
     */
    private Flux<ChatClientResponse> getToolEventFlux(String sessionId) {
        return toolEventPublisher.events(sessionId)
                .map(event -> {
                    Object result = Optional.ofNullable(event.result()).orElse("");
                    final String methodName = event.methodName();
                    String message = switch (event.type()) {
                        case "tool_call" -> String.format("正在进行工具调用 %s: %s", methodName, result);
                        case "tool_result" -> String.format("工具调用完成 %s: %s", methodName, result);
                        default -> "";
                    };
                    return String.format("\n\n[选择工具] %s \n\n", message);
                }).map(message -> {
                    AssistantMessage assistantMessage = new AssistantMessage(message);
                    Generation generation = new Generation(assistantMessage);
                    ChatResponse chatResponse = ChatResponse.builder()
                            .generations(List.of(generation))
                            .build();
                    return ChatClientResponse.builder().chatResponse(chatResponse).build();
                });
    }
}

注册一下，全局生效：

▼
java
复制代码
@Component  
public class Ai2ChatClient {    
	private final ChatClient chatClient;    
	public Ai2ChatClient(ChatModel openAiChatModel, FileTools fileTools, ToolAdvisor toolAdvisor) {    
		this.chatClient = ChatClient.builder(openAiChatModel)    
				.defaultTools(fileTools)    
				.defaultAdvisors(toolAdvisor)    
				.build();    
	}    
    public Flux<String> chat(String msg, String appId) {    
	return chatClient.prompt()  
                     .system("""  
                            You are a helpful, precise, and reliable AI assistant.                        Respond clearly and concisely.                        Prioritize correctness, safety, and practicality.                        If information is uncertain, state the uncertainty explicitly.                                                """)  
                    .user(msg)    
                    .advisors(spec -> spec.param(CONVERSATION_ID, appId))    
                    .toolContext(Map.of(CONVERSATION_ID, appId))    
                    .stream().content();    
	} 
}

两种方案怎么选

看什么	方案一自己写	方案二用 Advisor
代码位置	业务方法里	单独一个类
复用性	惨不忍睹	一次编写到处使用
代码量	挺长	业务方法就几行

我的建议是使用 advisor

怎么拿到 conversationId

Langchain4j 那个 @ToolMemoryId 是真方便。SpringAI 不给咱们就自己想办法。

在工具方法里加个 ToolContext 参数，从里面把 id 拽出来：

▼
java
复制代码
@Slf4j  
@Component  
public class FileWriteTool {    
	@Tool("写入文件到指定路径")    
	public String writeFile(    
	@P("文件的相对路径") String relativeFilePath,    
	@P("要写入文件的内容") String content,    
	ToolContext toolContext    
	) {    
        String conversationId = toolContext.getContext()    
                .get(ChatMemory.CONVERSATION_ID).toString();    
        // 接下来就能使用了
    }  
}

这个 id 是在调用链上通过 toolContext 传进来的：

▼
java
复制代码
public Flux<String> chat2AiAdvisor(String msg, String appId) {  
    return chatClient.prompt()  
            .system("你是有用的小助手")  
            .user(msg)  
            .advisors(advisorSpec -> advisorSpec.param(CONVERSATION_ID, appId))  
            // 这里设置到 ToolContext            .toolContext(Map.of(CONVERSATION_ID, appId))  
            .stream().content();  
}

跑一下看看

写个测试用例，验证整个链路通不通：

▼
java
复制代码
@SpringBootTest  
class Ai2ChatClientTest {    
	@Resource 
	private Ai2ChatClient ai2ChatClient;    
    
	@Test    
	void chat2Ai() throws InterruptedException {    
        Flux<String> flux = ai2ChatClient.chat("帮我生成一个企业级别的后端，帮我生成 todolist", "12345");    
        flux.doOnNext(System.out::println).subscribe(); 
        // 这里需要睡眠主线程，否则拿不到结果
        Thread.sleep(10000);    
	}  
}

跑起来，控制台陆陆续续打出日志，工具调用的事件也正常推送。整个链路是通的。