苍穹外卖项目接入 AI 智能客服功能实战指南

Controller 层设计

在 AiController 中创建接口，接收前端请求并调用 Service 层处理。

@PostMapping("/getAnswer")
public Result<String> getAnswer(@RequestBody Map<String, String> payload) {
    String question = payload.get("question");
    return Result.success(aiService.getAnswer(question));
}

Service 层实现

这里需要重点处理 WebSocket 连接、鉴权以及对话历史的管理。为了避免代码冗余，我们将大模型连接逻辑封装在 Service 内部。

首先定义接口：

public interface AiService {
    String getAnswer(String question);
}

接着实现类。注意，这里移除了原测试代码中的 main 方法和静态变量，改为 Spring 标准的实例化方式，确保线程安全。同时修复了配置注入时的语法错误。

@Service
public class AiServiceImpl implements AiService {
    // 对话历史存储集合，限制上下文长度
    private final List<RoleContent> historyList = new ArrayList<>();
    // 大模型的答案汇总
    private StringBuilder totalAnswer = new StringBuilder();
    // Gson 实例，用于 JSON 解析
    private static final Gson gson = new Gson();
    
    // 配置参数注入
    @Value("${ai.hostUrl}")
    private String hostUrl;
    
    @Value("${ai.domain}")
    private String domain;
    
    @Value("${ai.appid}")
    private String appid;
    
    @Value("${ai.apiSecret}")
    private String apiSecret;
    
    @Value("${ai.apiKey}")
    private String apiKey;

    @Override
    public String getAnswer(String question) {
        if (question == null || question.trim().isEmpty()) {
            return "请输入有效问题";
        }
        try {
            // 构建鉴权 URL
            String authUrl = getAuthUrl(hostUrl, apiKey, apiSecret);
            OkHttpClient client = new OkHttpClient.Builder().build();
            String url = authUrl.toString().replace("http://", "ws://").replace("https://", "wss://");
            Request request = new Request.Builder().url(url).build();
            
            // 使用 CountDownLatch 等待响应完成
            CountDownLatch latch = new CountDownLatch(1);
            WebSocket webSocket = client.newWebSocket(request, new AiWebSocketListener(latch, question));
            
            // 等待响应完成或超时（30 秒）
            boolean completed = latch.await(30, TimeUnit.SECONDS);
            if (!completed) {
                webSocket.close(1000, "请求超时");
                return "请求超时，请重试";
            }
            
            log.info("AI 回答：{}", totalAnswer);
            return totalAnswer.toString();
        } catch (Exception e) {
            e.printStackTrace();
            return "处理请求时出错：" + e.getMessage();
        }
    }

    // WebSocket 监听器，处理连接和消息
    class AiWebSocketListener extends WebSocketListener {
        private final CountDownLatch latch;
        private final String question;
        private final AtomicBoolean isResponseComplete = new AtomicBoolean(false);

        public AiWebSocketListener(CountDownLatch latch, String question) {
            this.latch = latch;
            this.question = question;
        }

        @Override
        public void onOpen(WebSocket webSocket, Response response) {
            totalAnswer.setLength(0); // 清空之前的回答
            super.onOpen(webSocket, response);
            sendRequest(webSocket, question);
        }

        @Override
        public void onMessage(WebSocket webSocket, String text) {
            try {
                JsonParse myJsonParse = gson.fromJson(text, JsonParse.class);
                if (myJsonParse.header.code != 0) {
                    System.err.println("发生错误，错误码为：" + myJsonParse.header.code);
                    webSocket.close(1000, "Error code: " + myJsonParse.header.code);
                    latch.countDown();
                    return;
                }

                // 处理回答内容
                List<Text> textList = myJsonParse.payload.choices.text;
                for (Text temp : textList) {
                    totalAnswer.append(temp.content);
                }

                // 判断是否为完整响应
                if (myJsonParse.header.status == 2) {
                    addToHistory(question, totalAnswer.toString());
                    isResponseComplete.set(true);
                    latch.countDown();
                }
            } catch (Exception e) {
                e.printStackTrace();
                latch.countDown();
            }
        }

        @Override
        public void onFailure(WebSocket webSocket, Throwable t, Response response) {
            super.onFailure(webSocket, t, response);
            System.err.println("WebSocket 连接失败");
            if (response != null) {
                System.err.println("错误码：" + response.code());
            }
            t.printStackTrace();
            latch.countDown();
        }

        @Override
        public void onClosed(WebSocket webSocket, int code, String reason) {
            super.onClosed(webSocket, code, reason);
            if (!isResponseComplete.get()) {
                latch.countDown();
            }
        }

        // 发送请求参数
        private void sendRequest(WebSocket webSocket, String question) {
            try {
                JSONObject requestJson = new JSONObject();
                JSONObject header = new JSONObject();
                header.put("app_id", appid);
                header.put("uid", UUID.randomUUID().toString().substring(0, 10));

                JSONObject parameter = new JSONObject();
                JSONObject chat = new JSONObject();
                chat.put("domain", domain);
                chat.put("temperature", 0.5);
                chat.put("max_tokens", 4096);
                parameter.put("chat", chat);

                JSONObject payload = new JSONObject();
                JSONObject message = new JSONObject();
                JSONArray text = new JSONArray();

                // 添加系统提示词，设定客服人设
                RoleContent roleSystem = new RoleContent();
                roleSystem.setRole("system");
                roleSystem.setContent("你是一个智能客服，接下来你要用客服的语气和我对话");
                text.add(JSON.toJSON(roleSystem));

                // 添加历史记录
                for (RoleContent history : historyList) {
                    text.add(JSON.toJSON(history));
                }

                // 添加当前问题
                RoleContent userQuestion = new RoleContent();
                userQuestion.setRole("user");
                userQuestion.setContent(question);
                text.add(JSON.toJSON(userQuestion));

                message.put("text", text);
                payload.put("message", message);
                requestJson.put("header", header);
                requestJson.put("parameter", parameter);
                requestJson.put("payload", payload);

                webSocket.send(requestJson.toString());
            } catch (Exception e) {
                e.printStackTrace();
                latch.countDown();
            }
        }
    }

    // 添加到对话历史，控制上下文长度
    private void addToHistory(String question, String answer) {
        if (historyList.size() >= 10) {
            historyList.remove(0);
        }
        RoleContent userContent = new RoleContent();
        userContent.setRole("user");
        userContent.setContent(question);
        historyList.add(userContent);

        RoleContent aiContent = new RoleContent();
        aiContent.setRole("assistant");
        aiContent.setContent(answer);
        historyList.add(aiContent);
    }

    // 鉴权方法
    private String getAuthUrl(String hostUrl, String apiKey, String apiSecret) throws Exception {
        URL url = new URL(hostUrl);
        SimpleDateFormat format = new SimpleDateFormat("EEE, dd MMM yyyy HH:mm:ss z", Locale.US);
        format.setTimeZone(TimeZone.getTimeZone("GMT"));
        String date = format.format(new Date());

        String preStr = "host: " + url.getHost() + "\n" + "date: " + date + "\n" + "GET " + url.getPath() + " HTTP/1.1";
        Mac mac = Mac.getInstance("HmacSHA256");
        SecretKeySpec spec = new SecretKeySpec(apiSecret.getBytes(StandardCharsets.UTF_8), "HmacSHA256");
        mac.init(spec);
        byte[] hexDigits = mac.doFinal(preStr.getBytes(StandardCharsets.UTF_8));
        String sha = Base64.getEncoder().encodeToString(hexDigits);

        String authorization = String.format(
            "api_key=\"%s\", algorithm=\"%s\", headers=\"%s\", signature=\"%s\"",
            apiKey, "hmac-sha256", "host date request-line", sha
        );

        HttpUrl httpUrl = Objects.requireNonNull(HttpUrl.parse("https://" + url.getHost() + url.getPath())).newBuilder()
            .addQueryParameter("authorization", Base64.getEncoder().encodeToString(authorization.getBytes(StandardCharsets.UTF_8)))
            .addQueryParameter("date", date)
            .addQueryParameter("host", url.getHost())
            .build();
        return httpUrl.toString();
    }

    // 内部数据类
    class JsonParse {
        Header header;
        Payload payload;
    }

    class Header {
        int code;
        int status;
        String sid;
    }

    class Payload {
        Choices choices;
    }

    class Choices {
        List<Text> text;
    }

    class Text {
        String role;
        String content;
    }

    class RoleContent {
        String role;
        String content;
        public String getRole() { return role; }
        public void setRole(String role) { this.role = role; }
        public String getContent() { return content; }
        public void setContent(String content) { this.content = content; }
    }
}

效果验证

完成上述代码后，重启服务进行测试。前端点击菜单项即可唤起 AI 客服窗口，输入问题后，后端会返回大模型生成的回复，并在界面上展示对话历史。