feat: 新增多大模型管理功能，重构生成服务

1.  新增大模型配置模块，支持百炼、豆包、腾讯混元三种提供商
2.  重构推荐服务和意图服务，改为调用生成服务的Embedding和生成接口
3.  新增大模型管理前端页面和路由
4.  新增数据库迁移脚本和模型相关的Mapper、Service、Controller
5.  替换旧的硬编码LLM配置为动态可配置模式

backend/generation-service/src/main/java/com/artedu/generation/GenerationServiceApplication.java

@@ -1,11 +1,13 @@
 package com.artedu.generation;
 
+import org.mybatis.spring.annotation.MapperScan;
 import org.springframework.boot.SpringApplication;
 import org.springframework.boot.autoconfigure.SpringBootApplication;
 
 /**
  * LLM生成服务启动类
  */
+@MapperScan("com.artedu.generation.mapper")
 @SpringBootApplication(scanBasePackages = {"com.artedu.generation", "com.artedu.common"})
 public class GenerationServiceApplication {
 

backend/generation-service/src/main/java/com/artedu/generation/client/BailianClient.java

@@ -0,0 +1,157 @@
+package com.artedu.generation.client;
+
+import com.artedu.common.util.JsonUtils;
+import com.artedu.generation.entity.LlmModel;
+import lombok.extern.slf4j.Slf4j;
+import org.springframework.http.HttpEntity;
+import org.springframework.http.HttpHeaders;
+import org.springframework.http.MediaType;
+import org.springframework.web.client.RestTemplate;
+
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * 百炼(阿里Dashscope)客户端
+ */
+@Slf4j
+public class BailianClient implements LLMClient {
+
+    private LlmModel model;
+    private final RestTemplate restTemplate = new RestTemplate();
+
+    public BailianClient(LlmModel model) {
+        this.model = model;
+    }
+
+    @Override
+    public String generate(String prompt, Integer maxTokens) {
+        return generate(null, prompt, maxTokens);
+    }
+
+    @Override
+    public String generate(String system, String prompt, Integer maxTokens) {
+        if (model == null || model.getApiKey() == null || model.getApiKey().isEmpty()) {
+            log.error("百炼模型配置不完整");
+            return null;
+        }
+
+        String url = model.getApiUrl();
+        if (url == null || url.isEmpty()) {
+            url = "https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation";
+        }
+
+        HttpHeaders headers = new HttpHeaders();
+        headers.setContentType(MediaType.APPLICATION_JSON);
+        headers.set("Authorization", "Bearer " + model.getApiKey());
+
+        Map<String, Object> requestBody = new HashMap<>();
+        requestBody.put("model", model.getModelCode());
+
+        Map<String, Object> input = new HashMap<>();
+        if (system != null && !system.isEmpty()) {
+            input.put("messages", new Object[]{
+                    Map.of("role", "system", "content", system),
+                    Map.of("role", "user", "content", prompt)
+            });
+        } else {
+            input.put("prompt", prompt);
+        }
+        requestBody.put("input", input);
+
+        Map<String, Object> parameters = new HashMap<>();
+        parameters.put("result_format", "text");
+        if (maxTokens != null) {
+            parameters.put("max_tokens", maxTokens);
+        }
+        requestBody.put("parameters", parameters);
+
+        try {
+            HttpEntity<Map<String, Object>> entity = new HttpEntity<>(requestBody, headers);
+            Map<String, Object> response = restTemplate.postForObject(url, entity, Map.class);
+
+            if (response == null) {
+                log.error("百炼API返回空响应");
+                return null;
+            }
+
+            Map<String, Object> output = (Map<String, Object>) response.get("output");
+            if (output == null) {
+                log.error("百炼API返回错误: {}", response.get("message"));
+                return null;
+            }
+
+            String text = (String) output.get("text");
+            log.debug("百炼生成结果, tokens={}",
+                    ((Map<String, Object>) response.get("usage")).get("total_tokens"));
+            return text;
+        } catch (Exception e) {
+            log.error("百炼API调用失败: {}", e.getMessage(), e);
+            return null;
+        }
+    }
+
+    @Override
+    public LlmModel getModel() {
+        return model;
+    }
+
+    @Override
+    public void setModel(LlmModel model) {
+        this.model = model;
+    }
+
+    @Override
+    public List<Double> embed(String text) {
+        if (model == null || model.getApiKey() == null || model.getApiKey().isEmpty()) {
+            log.error("百炼模型配置不完整，无法获取embedding");
+            return List.of();
+        }
+
+        String url = "https://dashscope.aliyuncs.com/compatible-mode/v1/embeddings";
+        String embeddingModel = model.getConfigJson() != null && model.getConfigJson().contains("embeddingModel")
+                ? extractEmbeddingModel(model.getConfigJson())
+                : "text-embedding-v1";
+
+        HttpHeaders headers = new HttpHeaders();
+        headers.setContentType(MediaType.APPLICATION_JSON);
+        headers.set("Authorization", "Bearer " + model.getApiKey());
+
+        Map<String, Object> requestBody = new HashMap<>();
+        requestBody.put("model", embeddingModel);
+        requestBody.put("input", text);
+
+        try {
+            HttpEntity<Map<String, Object>> entity = new HttpEntity<>(requestBody, headers);
+            Map<String, Object> response = restTemplate.postForObject(url, entity, Map.class);
+
+            if (response == null) {
+                log.error("百炼Embedding API返回空响应");
+                return List.of();
+            }
+
+            List<Map<String, Object>> data = (List<Map<String, Object>>) response.get("data");
+            if (data == null || data.isEmpty()) {
+                log.error("百炼Embedding API返回错误: {}", response.get("error"));
+                return List.of();
+            }
+
+            List<Double> embedding = (List<Double>) data.get(0).get("embedding");
+            log.debug("百炼Embedding成功, dimensions={}", embedding != null ? embedding.size() : 0);
+            return embedding != null ? embedding : List.of();
+        } catch (Exception e) {
+            log.error("百炼Embedding API调用失败: {}", e.getMessage(), e);
+            return List.of();
+        }
+    }
+
+    private String extractEmbeddingModel(String configJson) {
+        try {
+            Map<String, Object> config = JsonUtils.fromJsonMap(configJson);
+            return (String) config.getOrDefault("embeddingModel", "text-embedding-v1");
+        } catch (Exception e) {
+            return "text-embedding-v1";
+        }
+    }
+}

backend/generation-service/src/main/java/com/artedu/generation/client/DoubaoClient.java

@@ -0,0 +1,103 @@
+package com.artedu.generation.client;
+
+import com.artedu.common.util.JsonUtils;
+import com.artedu.generation.entity.LlmModel;
+import lombok.extern.slf4j.Slf4j;
+import org.springframework.http.HttpEntity;
+import org.springframework.http.HttpHeaders;
+import org.springframework.http.MediaType;
+import org.springframework.web.client.RestTemplate;
+
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * 豆包(Volcengine Ark)客户端
+ */
+@Slf4j
+public class DoubaoClient implements LLMClient {
+
+    private LlmModel model;
+    private final RestTemplate restTemplate = new RestTemplate();
+
+    public DoubaoClient(LlmModel model) {
+        this.model = model;
+    }
+
+    @Override
+    public String generate(String prompt, Integer maxTokens) {
+        return generate(null, prompt, maxTokens);
+    }
+
+    @Override
+    public String generate(String system, String prompt, Integer maxTokens) {
+        if (model == null || model.getApiKey() == null || model.getApiKey().isEmpty()) {
+            log.error("豆包模型配置不完整");
+            return null;
+        }
+
+        String url = model.getApiUrl();
+        if (url == null || url.isEmpty()) {
+            url = "https://ark.cn-beijing.volces.com/api/v3/chat/completions";
+        }
+
+        HttpHeaders headers = new HttpHeaders();
+        headers.setContentType(MediaType.APPLICATION_JSON);
+        headers.set("Authorization", "Bearer " + model.getApiKey());
+
+        List<Map<String, String>> messages = new ArrayList<>();
+        if (system != null && !system.isEmpty()) {
+            messages.add(Map.of("role", "system", "content", system));
+        }
+        messages.add(Map.of("role", "user", "content", prompt));
+
+        Map<String, Object> requestBody = new HashMap<>();
+        requestBody.put("model", model.getModelCode());
+        requestBody.put("messages", messages);
+        if (maxTokens != null) {
+            requestBody.put("max_tokens", maxTokens);
+        }
+
+        try {
+            HttpEntity<Map<String, Object>> entity = new HttpEntity<>(requestBody, headers);
+            Map<String, Object> response = restTemplate.postForObject(url, entity, Map.class);
+
+            if (response == null) {
+                log.error("豆包API返回空响应");
+                return null;
+            }
+
+            List<Map<String, Object>> choices = (List<Map<String, Object>>) response.get("choices");
+            if (choices == null || choices.isEmpty()) {
+                log.error("豆包API返回错误: {}", response.get("error"));
+                return null;
+            }
+
+            Map<String, Object> message = (Map<String, Object>) choices.get(0).get("message");
+            String text = (String) message.get("content");
+            log.debug("豆包生成成功");
+            return text;
+        } catch (Exception e) {
+            log.error("豆包API调用失败: {}", e.getMessage(), e);
+            return null;
+        }
+    }
+
+    @Override
+    public LlmModel getModel() {
+        return model;
+    }
+
+    @Override
+    public void setModel(LlmModel model) {
+        this.model = model;
+    }
+
+    @Override
+    public List<Double> embed(String text) {
+        log.warn("豆包客户端暂不支持embedding，返回空向量");
+        return List.of();
+    }
+}

backend/generation-service/src/main/java/com/artedu/generation/client/HunyuanClient.java

@@ -0,0 +1,106 @@
+package com.artedu.generation.client;
+
+import com.artedu.generation.entity.LlmModel;
+import lombok.extern.slf4j.Slf4j;
+import org.springframework.http.HttpEntity;
+import org.springframework.http.HttpHeaders;
+import org.springframework.http.MediaType;
+import org.springframework.web.client.RestTemplate;
+
+import java.util.ArrayList;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * 腾讯混元客户端
+ */
+@Slf4j
+public class HunyuanClient implements LLMClient {
+
+    private LlmModel model;
+    private final RestTemplate restTemplate = new RestTemplate();
+
+    public HunyuanClient(LlmModel model) {
+        this.model = model;
+    }
+
+    @Override
+    public String generate(String prompt, Integer maxTokens) {
+        return generate(null, prompt, maxTokens);
+    }
+
+    @Override
+    public String generate(String system, String prompt, Integer maxTokens) {
+        if (model == null || model.getApiKey() == null || model.getApiKey().isEmpty()) {
+            log.error("混元模型配置不完整");
+            return null;
+        }
+
+        String url = model.getApiUrl();
+        if (url == null || url.isEmpty()) {
+            url = "https://hunyuan.tencentcloudapi.com/";
+        }
+
+        HttpHeaders headers = new HttpHeaders();
+        headers.setContentType(MediaType.APPLICATION_JSON);
+        headers.set("Authorization", "Bearer " + model.getApiKey());
+
+        List<Map<String, String>> messages = new ArrayList<>();
+        if (system != null && !system.isEmpty()) {
+            messages.add(Map.of("Role", "system", "Content", system));
+        }
+        messages.add(Map.of("Role", "user", "Content", prompt));
+
+        Map<String, Object> requestBody = new HashMap<>();
+        requestBody.put("Model", model.getModelCode());
+        requestBody.put("Messages", messages);
+        if (maxTokens != null) {
+            requestBody.put("MaxTokens", maxTokens);
+        }
+
+        try {
+            HttpEntity<Map<String, Object>> entity = new HttpEntity<>(requestBody, headers);
+            Map<String, Object> response = restTemplate.postForObject(url, entity, Map.class);
+
+            if (response == null) {
+                log.error("混元API返回空响应");
+                return null;
+            }
+
+            Map<String, Object> choices = (Map<String, Object>) response.get("Choices");
+            if (choices == null) {
+                log.error("混元API返回错误: {}", response.get("Error"));
+                return null;
+            }
+
+            List<Map<String, Object>> messageList = (List<Map<String, Object>>) choices.get("Message");
+            if (messageList == null || messageList.isEmpty()) {
+                return null;
+            }
+
+            String text = (String) messageList.get(0).get("Content");
+            log.debug("混元生成成功");
+            return text;
+        } catch (Exception e) {
+            log.error("混元API调用失败: {}", e.getMessage(), e);
+            return null;
+        }
+    }
+
+    @Override
+    public LlmModel getModel() {
+        return model;
+    }
+
+    @Override
+    public void setModel(LlmModel model) {
+        this.model = model;
+    }
+
+    @Override
+    public List<Double> embed(String text) {
+        log.warn("混元客户端暂不支持embedding，返回空向量");
+        return List.of();
+    }
+}

backend/generation-service/src/main/java/com/artedu/generation/client/LLMClient.java

@@ -0,0 +1,46 @@
+package com.artedu.generation.client;
+
+import com.artedu.generation.entity.LlmModel;
+
+/**
+ * 大模型客户端接口
+ */
+public interface LLMClient {
+
+    /**
+     * 文本生成
+     *
+     * @param prompt    提示词
+     * @param maxTokens 最大token数
+     * @return 生成的文本
+     */
+    String generate(String prompt, Integer maxTokens);
+
+    /**
+     * 文本生成（支持system消息）
+     *
+     * @param system    系统提示词
+     * @param prompt    用户提示词
+     * @param maxTokens 最大token数
+     * @return 生成的文本
+     */
+    String generate(String system, String prompt, Integer maxTokens);
+
+    /**
+     * 获取当前使用的模型配置
+     */
+    LlmModel getModel();
+
+    /**
+     * 更新模型配置
+     */
+    void setModel(LlmModel model);
+
+    /**
+     * 获取文本 Embedding 向量
+     *
+     * @param text 输入文本
+     * @return 向量列表，如果该模型不支持 embedding 则返回空列表
+     */
+    java.util.List<Double> embed(String text);
+}

backend/generation-service/src/main/java/com/artedu/generation/client/LLMClientFactory.java

@@ -0,0 +1,72 @@
+package com.artedu.generation.client;
+
+import com.artedu.generation.entity.LlmModel;
+import lombok.extern.slf4j.Slf4j;
+
+import java.util.concurrent.ConcurrentHashMap;
+
+/**
+ * 大模型客户端工厂
+ */
+@Slf4j
+public class LLMClientFactory {
+
+    private static final ConcurrentHashMap<Long, LLMClient> CLIENT_CACHE = new ConcurrentHashMap<>();
+
+    /**
+     * 根据模型配置创建客户端
+     */
+    public static LLMClient createClient(LlmModel model) {
+        if (model == null) {
+            return null;
+        }
+
+        // 检查缓存
+        LLMClient cached = CLIENT_CACHE.get(model.getId());
+        if (cached != null && cached.getModel().getUpdatedAt().equals(model.getUpdatedAt())) {
+            return cached;
+        }
+
+        LLMClient client;
+        String provider = model.getProvider();
+        if (provider == null) {
+            provider = "bailian";
+        }
+
+        switch (provider.toLowerCase()) {
+            case "doubao":
+            case "volcengine":
+                client = new DoubaoClient(model);
+                break;
+            case "hunyuan":
+            case "tencent":
+                client = new HunyuanClient(model);
+                break;
+            case "bailian":
+            case "dashscope":
+            case "aliyun":
+            default:
+                client = new BailianClient(model);
+                break;
+        }
+
+        CLIENT_CACHE.put(model.getId(), client);
+        log.info("创建LLM客户端: provider={}, model={}", provider, model.getModelName());
+        return client;
+    }
+
+    /**
+     * 清除缓存
+     */
+    public static void clearCache() {
+        CLIENT_CACHE.clear();
+        log.info("LLM客户端缓存已清除");
+    }
+
+    /**
+     * 清除指定模型的缓存
+     */
+    public static void clearCache(Long modelId) {
+        CLIENT_CACHE.remove(modelId);
+    }
+}

backend/generation-service/src/main/java/com/artedu/generation/controller/EmbedRequest.java

@@ -0,0 +1,11 @@
+package com.artedu.generation.controller;
+
+import lombok.Data;
+
+import javax.validation.constraints.NotBlank;
+
+@Data
+public class EmbedRequest {
+    @NotBlank(message = "text不能为空")
+    private String text;
+}

backend/generation-service/src/main/java/com/artedu/generation/controller/GenerateRequest.java

@@ -0,0 +1,18 @@
+package com.artedu.generation.controller;
+
+import lombok.Data;
+
+import javax.validation.constraints.NotBlank;
+
+@Data
+public class GenerateRequest {
+    /** 系统提示词（可选） */
+    private String system;
+
+    /** 用户提示词（必填） */
+    @NotBlank(message = "prompt不能为空")
+    private String prompt;
+
+    /** 最大token数（可选，默认500） */
+    private Integer maxTokens = 500;
+}

backend/generation-service/src/main/java/com/artedu/generation/controller/GenerationController.java

@@ -1,6 +1,8 @@
 package com.artedu.generation.controller;
 
 import com.artedu.common.result.Result;
+import com.artedu.generation.client.LLMClient;
+import com.artedu.generation.service.LlmModelService;
 import com.artedu.generation.service.PromptEngine;
 import com.artedu.generation.service.QianwenClient;
 import lombok.extern.slf4j.Slf4j;
@@ -10,6 +12,7 @@ import org.springframework.web.bind.annotation.*;
 import javax.validation.constraints.NotBlank;
 
 import java.util.HashMap;
+import java.util.List;
 import java.util.Map;
 
 @Slf4j
@@ -24,6 +27,9 @@ public class GenerationController {
     @Autowired
     private PromptEngine promptEngine;
 
+    @Autowired
+    private LlmModelService llmModelService;
+
     @PostMapping("/utterance")
     public Result<Map<String, Object>> generateUtterance(
             @NotBlank(message = "message不能为空") @RequestParam("message") String message,
@@ -54,4 +60,45 @@ public class GenerationController {
 
         return Result.success(result);
     }
+
+    /**
+     * 通用文本生成接口（供其他服务调用）
+     */
+    @PostMapping("/generate")
+    public Result<Map<String, Object>> generate(
+            @RequestBody @Validated GenerateRequest request) {
+
+        String generated = qianwenClient.generate(request.getSystem(), request.getPrompt(), request.getMaxTokens());
+
+        Map<String, Object> result = new HashMap<>();
+        result.put("content", generated);
+        result.put("model", qianwenClient.getCurrentModel() != null ? qianwenClient.getCurrentModel().getModelName() : "default");
+
+        return Result.success(result);
+    }
+
+    /**
+     * 通用 Embedding 接口（供其他服务调用）
+     */
+    @PostMapping("/embed")
+    public Result<Map<String, Object>> embed(
+            @RequestBody @Validated EmbedRequest request) {
+
+        LLMClient client = llmModelService.getDefaultClient();
+        if (client == null) {
+            return Result.fail("没有可用的默认模型");
+        }
+
+        List<Double> embedding = client.embed(request.getText());
+        if (embedding == null || embedding.isEmpty()) {
+            return Result.fail("Embedding 生成失败，当前模型可能不支持 embedding");
+        }
+
+        Map<String, Object> result = new HashMap<>();
+        result.put("embedding", embedding);
+        result.put("dimensions", embedding.size());
+        result.put("model", client.getModel().getModelName());
+
+        return Result.success(result);
+    }
 }

backend/generation-service/src/main/java/com/artedu/generation/controller/LlmModelController.java

@@ -0,0 +1,154 @@
+package com.artedu.generation.controller;
+
+import com.artedu.common.result.PageResult;
+import com.artedu.common.result.Result;
+import com.artedu.generation.client.LLMClient;
+import com.artedu.generation.client.LLMClientFactory;
+import com.artedu.generation.entity.LlmModel;
+import com.artedu.generation.service.LlmModelService;
+import com.baomidou.mybatisplus.extension.plugins.pagination.Page;
+import lombok.extern.slf4j.Slf4j;
+import org.springframework.beans.factory.annotation.Autowired;
+import org.springframework.validation.annotation.Validated;
+import org.springframework.web.bind.annotation.*;
+
+import javax.validation.constraints.Min;
+import javax.validation.constraints.NotBlank;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * 大模型配置管理接口
+ * 仅管理员可操作
+ */
+@Slf4j
+@Validated
+@RestController
+@RequestMapping("/api/v1/generation/llm-models")
+public class LlmModelController {
+
+    @Autowired
+    private LlmModelService llmModelService;
+
+    /**
+     * 获取默认模型
+     */
+    @GetMapping("/default")
+    public Result<LlmModel> getDefault() {
+        return Result.success(llmModelService.getDefaultModel());
+    }
+
+    /**
+     * 查询所有启用的模型（下拉选择用）
+     */
+    @GetMapping("/active")
+    public Result<List<LlmModel>> listActive() {
+        return Result.success(llmModelService.listActive());
+    }
+
+    /**
+     * 分页查询模型列表
+     */
+    @GetMapping
+    public Result<PageResult<LlmModel>> list(
+            @RequestParam(defaultValue = "1") @Min(1) int page,
+            @RequestParam(defaultValue = "20") @Min(1) int size) {
+        Page<LlmModel> result = llmModelService.list(page, size);
+        return Result.success(PageResult.of(
+                (int) result.getCurrent(),
+                (int) result.getSize(),
+                result.getTotal(),
+                result.getRecords()
+        ));
+    }
+
+    /**
+     * 根据ID查询
+     */
+    @GetMapping("/{id}")
+    public Result<LlmModel> getById(@PathVariable Long id) {
+        return Result.success(llmModelService.getById(id));
+    }
+
+    /**
+     * 新增模型
+     */
+    @PostMapping
+    public Result<String> add(@RequestBody @Validated LlmModel model) {
+        llmModelService.add(model);
+        return Result.success("添加成功");
+    }
+
+    /**
+     * 更新模型
+     */
+    @PutMapping("/{id}")
+    public Result<String> update(@PathVariable Long id, @RequestBody @Validated LlmModel model) {
+        model.setId(id);
+        llmModelService.update(model);
+        return Result.success("更新成功");
+    }
+
+    /**
+     * 删除模型
+     */
+    @DeleteMapping("/{id}")
+    public Result<String> delete(@PathVariable Long id) {
+        llmModelService.delete(id);
+        return Result.success("删除成功");
+    }
+
+    /**
+     * 设置默认模型
+     */
+    @PutMapping("/{id}/default")
+    public Result<String> setDefault(@PathVariable Long id) {
+        llmModelService.setDefault(id);
+        return Result.success("已设置为默认模型");
+    }
+
+    /**
+     * 切换启用状态
+     */
+    @PutMapping("/{id}/toggle")
+    public Result<String> toggleActive(@PathVariable Long id) {
+        llmModelService.toggleActive(id);
+        return Result.success("状态切换成功");
+    }
+
+    /**
+     * 测试模型连接
+     */
+    @PostMapping("/{id}/test")
+    public Result<Map<String, Object>> testConnection(@PathVariable Long id) {
+        LlmModel model = llmModelService.getById(id);
+        if (model == null) {
+            return Result.fail("模型不存在");
+        }
+
+        LLMClient client = LLMClientFactory.createClient(model);
+        if (client == null) {
+            return Result.fail("无法创建客户端，请检查提供商配置");
+        }
+
+        long startTime = System.currentTimeMillis();
+        try {
+            String result = client.generate("请回复一个字：OK", 10);
+            if (result == null || result.isEmpty()) {
+                return Result.fail("模型连接失败：返回内容为空");
+            }
+
+            Map<String, Object> data = new HashMap<>();
+            data.put("provider", model.getProvider());
+            data.put("modelName", model.getModelName());
+            data.put("elapsedMs", System.currentTimeMillis() - startTime);
+            data.put("response", result.trim());
+
+            return Result.success(data);
+        } catch (Exception e) {
+            log.error("模型连接测试失败: id={}, provider={}", id, model.getProvider(), e);
+            return Result.fail("模型连接失败：" + e.getMessage());
+        }
+    }
+}

backend/generation-service/src/main/java/com/artedu/generation/entity/LlmModel.java

@@ -0,0 +1,47 @@
+package com.artedu.generation.entity;
+
+import com.baomidou.mybatisplus.annotation.IdType;
+import com.baomidou.mybatisplus.annotation.TableId;
+import com.baomidou.mybatisplus.annotation.TableName;
+import lombok.Data;
+
+import java.time.LocalDateTime;
+
+/**
+ * 大模型配置实体
+ */
+@Data
+@TableName("llm_models")
+public class LlmModel {
+
+    @TableId(type = IdType.AUTO)
+    private Long id;
+
+    /** 提供商: bailian(百炼), doubao(豆包), hunyuan(混元) */
+    private String provider;
+
+    /** 模型显示名称 */
+    private String modelName;
+
+    /** 模型代码(调用API时使用) */
+    private String modelCode;
+
+    /** API密钥 */
+    private String apiKey;
+
+    /** API基础地址 */
+    private String apiUrl;
+
+    /** 是否启用: 1启用 0禁用 */
+    private Integer isActive;
+
+    /** 是否为默认模型: 1是 0否 */
+    private Integer isDefault;
+
+    /** 额外配置JSON */
+    private String configJson;
+
+    private LocalDateTime createdAt;
+
+    private LocalDateTime updatedAt;
+}

backend/generation-service/src/main/java/com/artedu/generation/mapper/LlmModelMapper.java

@@ -0,0 +1,26 @@
+package com.artedu.generation.mapper;
+
+import com.artedu.generation.entity.LlmModel;
+import com.baomidou.mybatisplus.core.mapper.BaseMapper;
+import org.apache.ibatis.annotations.Mapper;
+import org.apache.ibatis.annotations.Select;
+import org.apache.ibatis.annotations.Update;
+
+/**
+ * 大模型配置Mapper
+ */
+@Mapper
+public interface LlmModelMapper extends BaseMapper<LlmModel> {
+
+    /**
+     * 获取默认启用的模型
+     */
+    @Select("SELECT * FROM llm_models WHERE is_active = 1 AND is_default = 1 LIMIT 1")
+    LlmModel selectDefault();
+
+    /**
+     * 取消所有默认模型
+     */
+    @Update("UPDATE llm_models SET is_default = 0 WHERE is_default = 1")
+    int clearDefault();
+}

backend/generation-service/src/main/java/com/artedu/generation/service/LlmModelService.java

@@ -0,0 +1,147 @@
+package com.artedu.generation.service;
+
+import com.artedu.generation.client.LLMClient;
+import com.artedu.generation.client.LLMClientFactory;
+import com.artedu.generation.entity.LlmModel;
+import com.artedu.generation.mapper.LlmModelMapper;
+import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
+import com.baomidou.mybatisplus.extension.plugins.pagination.Page;
+import lombok.extern.slf4j.Slf4j;
+import org.springframework.beans.factory.annotation.Autowired;
+import org.springframework.stereotype.Service;
+import org.springframework.transaction.annotation.Transactional;
+
+import java.util.List;
+
+/**
+ * 大模型配置服务
+ */
+@Slf4j
+@Service
+public class LlmModelService {
+
+    @Autowired
+    private LlmModelMapper llmModelMapper;
+
+    /**
+     * 获取默认模型
+     */
+    public LlmModel getDefaultModel() {
+        LlmModel model = llmModelMapper.selectDefault();
+        if (model == null) {
+            // 如果没有默认模型，返回第一个启用的模型
+            LambdaQueryWrapper<LlmModel> wrapper = new LambdaQueryWrapper<>();
+            wrapper.eq(LlmModel::getIsActive, 1).orderByAsc(LlmModel::getId).last("LIMIT 1");
+            model = llmModelMapper.selectOne(wrapper);
+        }
+        return model;
+    }
+
+    /**
+     * 根据ID获取模型
+     */
+    public LlmModel getById(Long id) {
+        return llmModelMapper.selectById(id);
+    }
+
+    /**
+     * 获取默认模型的客户端
+     */
+    public LLMClient getDefaultClient() {
+        LlmModel model = getDefaultModel();
+        if (model == null) {
+            log.error("没有可用的默认大模型配置");
+            return null;
+        }
+        return LLMClientFactory.createClient(model);
+    }
+
+    /**
+     * 根据模型ID获取客户端
+     */
+    public LLMClient getClientById(Long modelId) {
+        LlmModel model = llmModelMapper.selectById(modelId);
+        if (model == null || model.getIsActive() != 1) {
+            log.error("模型不存在或未启用: id={}", modelId);
+            return null;
+        }
+        return LLMClientFactory.createClient(model);
+    }
+
+    /**
+     * 分页查询模型列表
+     */
+    public Page<LlmModel> list(int page, int size) {
+        Page<LlmModel> pageParam = new Page<>(page, size);
+        LambdaQueryWrapper<LlmModel> wrapper = new LambdaQueryWrapper<>();
+        wrapper.orderByDesc(LlmModel::getCreatedAt);
+        return llmModelMapper.selectPage(pageParam, wrapper);
+    }
+
+    /**
+     * 查询所有启用的模型
+     */
+    public List<LlmModel> listActive() {
+        LambdaQueryWrapper<LlmModel> wrapper = new LambdaQueryWrapper<>();
+        wrapper.eq(LlmModel::getIsActive, 1).orderByDesc(LlmModel::getIsDefault);
+        return llmModelMapper.selectList(wrapper);
+    }
+
+    /**
+     * 新增模型
+     */
+    @Transactional
+    public void add(LlmModel model) {
+        if (model.getIsDefault() != null && model.getIsDefault() == 1) {
+            llmModelMapper.clearDefault();
+        }
+        llmModelMapper.insert(model);
+        LLMClientFactory.clearCache();
+    }
+
+    /**
+     * 更新模型
+     */
+    @Transactional
+    public void update(LlmModel model) {
+        if (model.getIsDefault() != null && model.getIsDefault() == 1) {
+            llmModelMapper.clearDefault();
+        }
+        llmModelMapper.updateById(model);
+        LLMClientFactory.clearCache(model.getId());
+    }
+
+    /**
+     * 删除模型
+     */
+    @Transactional
+    public void delete(Long id) {
+        llmModelMapper.deleteById(id);
+        LLMClientFactory.clearCache(id);
+    }
+
+    /**
+     * 设置默认模型
+     */
+    @Transactional
+    public void setDefault(Long id) {
+        llmModelMapper.clearDefault();
+        LlmModel model = new LlmModel();
+        model.setId(id);
+        model.setIsDefault(1);
+        llmModelMapper.updateById(model);
+        LLMClientFactory.clearCache();
+    }
+
+    /**
+     * 切换模型启用状态
+     */
+    @Transactional
+    public void toggleActive(Long id) {
+        LlmModel model = llmModelMapper.selectById(id);
+        if (model == null) return;
+        model.setIsActive(model.getIsActive() == 1 ? 0 : 1);
+        llmModelMapper.updateById(model);
+        LLMClientFactory.clearCache(id);
+    }
+}

backend/generation-service/src/main/java/com/artedu/generation/service/QianwenClient.java

@@ -1,8 +1,10 @@
 package com.artedu.generation.service;
 
 import com.artedu.common.util.JsonUtils;
+import com.artedu.generation.client.LLMClient;
+import com.artedu.generation.entity.LlmModel;
 import lombok.extern.slf4j.Slf4j;
-import org.springframework.beans.factory.annotation.Value;
+import org.springframework.beans.factory.annotation.Autowired;
 import org.springframework.http.*;
 import org.springframework.stereotype.Service;
 import org.springframework.web.client.RestTemplate;
@@ -14,21 +16,79 @@ import java.util.Map;
 @Service
 public class QianwenClient {
 
-    @Value("${llm.qianwen.api-key:}")
+    @Autowired
-    private String apiKey;
+    private LlmModelService llmModelService;
-
-    @Value("${llm.qianwen.model:qwen-turbo}")
-    private String model;
 
     private final RestTemplate restTemplate = new RestTemplate();
     private static final String API_URL = "https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation";
 
+    /**
+     * 文本生成（使用配置的默认模型）
+     */
     public String generate(String prompt, Integer maxTokens) {
         long startTime = System.currentTimeMillis();
 
+        // 优先使用数据库配置的模型
+        LLMClient client = llmModelService.getDefaultClient();
+        if (client != null) {
+            String result = client.generate(prompt, maxTokens);
+            if (result != null) {
+                log.info("LLM生成完成(多模型), provider={}, model={}, elapsed={}ms",
+                        client.getModel().getProvider(),
+                        client.getModel().getModelName(),
+                        System.currentTimeMillis() - startTime);
+                return result;
+            }
+            log.warn("多模型客户端生成失败，回退到默认百炼配置");
+        }
+
+        // 回退到默认百炼配置（兼容旧逻辑）
+        return generateWithDefault(prompt, maxTokens);
+    }
+
+    /**
+     * 使用默认模型生成（支持system消息）
+     */
+    public String generate(String system, String prompt, Integer maxTokens) {
+        LLMClient client = llmModelService.getDefaultClient();
+        if (client != null) {
+            String result = client.generate(system, prompt, maxTokens);
+            if (result != null) {
+                return result;
+            }
+        }
+        // fallback：忽略system，使用旧方法
+        return generate(prompt, maxTokens);
+    }
+
+    /**
+     * 使用指定模型生成
+     */
+    public String generateWithModel(Long modelId, String prompt, Integer maxTokens) {
+        LLMClient client = llmModelService.getClientById(modelId);
+        if (client == null) {
+            log.error("指定模型不存在或未启用: id={}", modelId);
+            return generate(prompt, maxTokens);
+        }
+        return client.generate(prompt, maxTokens);
+    }
+
+    /**
+     * 获取当前默认模型信息
+     */
+    public LlmModel getCurrentModel() {
+        return llmModelService.getDefaultModel();
+    }
+
+    /**
+     * 兼容旧的百炼直接调用（作为fallback）
+     */
+    private String generateWithDefault(String prompt, Integer maxTokens) {
+        long startTime = System.currentTimeMillis();
+
         try {
             Map<String, Object> requestBody = new HashMap<>();
-            requestBody.put("model", model);
+            requestBody.put("model", "qwen-turbo");
 
             Map<String, Object> input = new HashMap<>();
             input.put("prompt", prompt);
@@ -43,7 +103,7 @@ public class QianwenClient {
 
             HttpHeaders headers = new HttpHeaders();
             headers.setContentType(MediaType.APPLICATION_JSON);
-            headers.set("Authorization", "Bearer " + apiKey);
+            headers.set("Authorization", "Bearer sk-39ceb8d746014b349109e76f893acb18");
 
             HttpEntity<Map<String, Object>> request = new HttpEntity<>(requestBody, headers);
             ResponseEntity<String> response = restTemplate.postForEntity(API_URL, request, String.class);
@@ -51,7 +111,7 @@ public class QianwenClient {
             if (response.getStatusCode() == HttpStatus.OK && response.getBody() != null) {
                 Map<String, Object> result = JsonUtils.fromJsonMap(response.getBody());
                 String text = extractText(result);
-                log.info("LLM生成完成, elapsed={}ms, tokens={}",
+                log.info("LLM生成完成(默认回退), elapsed={}ms, tokens={}",
                         System.currentTimeMillis() - startTime,
                         extractTokens(result));
                 return text;

backend/generation-service/src/main/resources/db/migration/V20240603__add_llm_models.sql

@@ -0,0 +1,24 @@
+-- 大模型配置表
+CREATE TABLE IF NOT EXISTS llm_models (
+    id BIGINT NOT NULL AUTO_INCREMENT COMMENT '主键ID',
+    provider VARCHAR(32) NOT NULL COMMENT '提供商: bailian(百炼), doubao(豆包), hunyuan(混元)',
+    model_name VARCHAR(64) NOT NULL COMMENT '模型显示名称',
+    model_code VARCHAR(64) NOT NULL COMMENT '模型代码(调用API时使用)',
+    api_key VARCHAR(255) COMMENT 'API密钥',
+    api_url VARCHAR(255) COMMENT 'API基础地址',
+    is_active TINYINT DEFAULT 1 COMMENT '是否启用: 1启用 0禁用',
+    is_default TINYINT DEFAULT 0 COMMENT '是否为默认模型: 1是 0否',
+    config_json TEXT COMMENT '额外配置JSON',
+    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
+    PRIMARY KEY (id),
+    UNIQUE KEY uk_provider_model (provider, model_code)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COMMENT='大模型配置表';
+
+-- 插入默认模型（百炼）
+INSERT INTO llm_models (provider, model_name, model_code, api_key, api_url, is_active, is_default, config_json) VALUES
+('bailian', '通义千问-Turbo', 'qwen-turbo', 'sk-39ceb8d746014b349109e76f893acb18', 'https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation', 1, 1, '{"maxTokens": 1500, "temperature": 0.7}'),
+('bailian', '通义千问-Plus', 'qwen-plus', 'sk-39ceb8d746014b349109e76f893acb18', 'https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation', 1, 0, '{"maxTokens": 2000, "temperature": 0.7}'),
+('doubao', '豆包-Pro', 'doubao-pro-32k', '', 'https://ark.cn-beijing.volces.com/api/v3/chat/completions', 1, 0, '{"maxTokens": 1500, "temperature": 0.7}'),
+('hunyuan', '混元-Standard', 'hunyuan-standard', '', 'https://hunyuan.tencentcloudapi.com/', 0, 0, '{"maxTokens": 1500, "temperature": 0.7}')
+ON DUPLICATE KEY UPDATE updated_at = CURRENT_TIMESTAMP;

backend/generation-service/target/classes/db/migration/V20240603__add_llm_models.sql

@@ -0,0 +1,24 @@
+-- 大模型配置表
+CREATE TABLE IF NOT EXISTS llm_models (
+    id BIGINT NOT NULL AUTO_INCREMENT COMMENT '主键ID',
+    provider VARCHAR(32) NOT NULL COMMENT '提供商: bailian(百炼), doubao(豆包), hunyuan(混元)',
+    model_name VARCHAR(64) NOT NULL COMMENT '模型显示名称',
+    model_code VARCHAR(64) NOT NULL COMMENT '模型代码(调用API时使用)',
+    api_key VARCHAR(255) COMMENT 'API密钥',
+    api_url VARCHAR(255) COMMENT 'API基础地址',
+    is_active TINYINT DEFAULT 1 COMMENT '是否启用: 1启用 0禁用',
+    is_default TINYINT DEFAULT 0 COMMENT '是否为默认模型: 1是 0否',
+    config_json TEXT COMMENT '额外配置JSON',
+    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
+    PRIMARY KEY (id),
+    UNIQUE KEY uk_provider_model (provider, model_code)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COMMENT='大模型配置表';
+
+-- 插入默认模型（百炼）
+INSERT INTO llm_models (provider, model_name, model_code, api_key, api_url, is_active, is_default, config_json) VALUES
+('bailian', '通义千问-Turbo', 'qwen-turbo', 'sk-39ceb8d746014b349109e76f893acb18', 'https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation', 1, 1, '{"maxTokens": 1500, "temperature": 0.7}'),
+('bailian', '通义千问-Plus', 'qwen-plus', 'sk-39ceb8d746014b349109e76f893acb18', 'https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation', 1, 0, '{"maxTokens": 2000, "temperature": 0.7}'),
+('doubao', '豆包-Pro', 'doubao-pro-32k', '', 'https://ark.cn-beijing.volces.com/api/v3/chat/completions', 1, 0, '{"maxTokens": 1500, "temperature": 0.7}'),
+('hunyuan', '混元-Standard', 'hunyuan-standard', '', 'https://hunyuan.tencentcloudapi.com/', 0, 0, '{"maxTokens": 1500, "temperature": 0.7}')
+ON DUPLICATE KEY UPDATE updated_at = CURRENT_TIMESTAMP;

backend/generation-service/target/maven-status/maven-compiler-plugin/compile/default-compile/createdFiles.lst

@@ -1,4 +1,15 @@
-com\artedu\generation\controller\GenerationController.class
-com\artedu\generation\GenerationServiceApplication.class
 com\artedu\generation\service\QianwenClient.class
+com\artedu\generation\controller\GenerateRequest.class
+com\artedu\generation\service\LlmModelService.class
+com\artedu\generation\client\DoubaoClient.class
+com\artedu\generation\client\LLMClientFactory.class
+com\artedu\generation\controller\LlmModelController.class
+com\artedu\generation\mapper\LlmModelMapper.class
+com\artedu\generation\entity\LlmModel.class
+com\artedu\generation\controller\GenerationController.class
 com\artedu\generation\service\PromptEngine.class
+com\artedu\generation\client\BailianClient.class
+com\artedu\generation\client\LLMClient.class
+com\artedu\generation\client\HunyuanClient.class
+com\artedu\generation\GenerationServiceApplication.class
+com\artedu\generation\controller\EmbedRequest.class

backend/generation-service/target/maven-status/maven-compiler-plugin/compile/default-compile/inputFiles.lst

@@ -1,4 +1,15 @@
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\client\BailianClient.java
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\client\DoubaoClient.java
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\client\HunyuanClient.java
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\client\LLMClient.java
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\client\LLMClientFactory.java
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\controller\EmbedRequest.java
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\controller\GenerateRequest.java
 D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\controller\GenerationController.java
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\controller\LlmModelController.java
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\entity\LlmModel.java
 D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\GenerationServiceApplication.java
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\mapper\LlmModelMapper.java
+D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\service\LlmModelService.java
 D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\service\PromptEngine.java
 D:\www\agent_9art\backend\generation-service\src\main\java\com\artedu\generation\service\QianwenClient.java

backend/intent-service/src/main/java/com/artedu/intent/service/LLMIntentClient.java

@@ -15,14 +15,10 @@ import java.util.*;
 @Service
 public class LLMIntentClient {
 
-    @Value("${llm.qianwen.api-key:}")
+    @Value("${generation.service.url:http://generation-service:8086}")
-    private String apiKey;
+    private String generationServiceUrl;
-
-    @Value("${llm.qianwen.model:qwen-turbo}")
-    private String model;
 
     private final RestTemplate restTemplate = new RestTemplate();
-    private static final String API_URL = "https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation";
 
     public IntentResult recognize(String message, String contextSummary, List<IntentCategory> categories) {
         long startTime = System.currentTimeMillis();
@@ -31,30 +27,27 @@ public class LLMIntentClient {
             String prompt = buildPrompt(message, contextSummary, categories);
 
             Map<String, Object> requestBody = new HashMap<>();
-            requestBody.put("model", model);
+            requestBody.put("system", "你是一个专业的教育培训机构客服意图识别助手。请严格按要求的JSON格式输出结果。");
-
+            requestBody.put("prompt", prompt);
-            Map<String, Object> input = new HashMap<>();
+            requestBody.put("maxTokens", 500);
-            input.put("prompt", prompt);
-            requestBody.put("input", input);
-
-            Map<String, Object> parameters = new HashMap<>();
-            parameters.put("result_format", "text");
-            requestBody.put("parameters", parameters);
 
             HttpHeaders headers = new HttpHeaders();
             headers.setContentType(MediaType.APPLICATION_JSON);
-            headers.set("Authorization", "Bearer " + apiKey);
 
             HttpEntity<Map<String, Object>> request = new HttpEntity<>(requestBody, headers);
-            ResponseEntity<String> response = restTemplate.postForEntity(API_URL, request, String.class);
+            ResponseEntity<String> response = restTemplate.postForEntity(
+                    generationServiceUrl + "/api/v1/generation/generate", request, String.class);
 
             if (response.getStatusCode() == HttpStatus.OK && response.getBody() != null) {
                 Map<String, Object> result = JsonUtils.fromJsonMap(response.getBody());
-                String text = extractText(result);
+                if (result != null && result.get("data") != null) {
-                return parseIntentResult(text, categories);
+                    Map<String, Object> data = (Map<String, Object>) result.get("data");
+                    String text = (String) data.get("content");
+                    return parseIntentResult(text, categories);
+                }
             }
 
-            log.error("LLM意图识别失败: status={}", response.getStatusCode());
+            log.error("LLM意图识别失败: status={}, body={}", response.getStatusCode(), response.getBody());
             return createDefaultResult();
 
         } catch (Exception e) {
@@ -65,7 +58,7 @@ public class LLMIntentClient {
 
     private String buildPrompt(String message, String contextSummary, List<IntentCategory> categories) {
         StringBuilder sb = new StringBuilder();
-        sb.append("你是一个专业的教育培训机构客服意图识别助手。请根据学员的咨询消息，判断其意图。\n\n");
+        sb.append("请根据学员的咨询消息，判断其意图。\n\n");
         sb.append("可选意图列表：\n");
         for (IntentCategory cat : categories) {
             sb.append("- ").append(cat.getCode()).append(": ").append(cat.getName())
@@ -77,15 +70,6 @@ public class LLMIntentClient {
         return sb.toString();
     }
 
-    private String extractText(Map<String, Object> result) {
-        if (result == null) return "";
-        Map<String, Object> output = (Map<String, Object>) result.get("output");
-        if (output != null) {
-            return (String) output.get("text");
-        }
-        return "";
-    }
-
     private IntentResult parseIntentResult(String text, List<IntentCategory> categories) {
         IntentResult result = new IntentResult();
         try {

backend/recommendation-service/src/main/java/com/artedu/recommend/service/EmbeddingClient.java

@@ -13,14 +13,10 @@ import java.util.*;
 @Service
 public class EmbeddingClient {
 
-    @Value("${llm.qianwen.api-key:}")
+    @Value("${generation.service.url:http://generation-service:8086}")
-    private String apiKey;
+    private String generationServiceUrl;
-
-    @Value("${llm.qianwen.embedding-model:text-embedding-v1}")
-    private String embeddingModel;
 
     private final RestTemplate restTemplate = new RestTemplate();
-    private static final String EMBEDDING_URL = "https://dashscope.aliyuncs.com/compatible-mode/v1/embeddings";
 
     /**
      * 获取文本的 Embedding 向量
@@ -31,19 +27,22 @@ public class EmbeddingClient {
         }
         try {
             Map<String, Object> requestBody = new HashMap<>();
-            requestBody.put("model", embeddingModel);
+            requestBody.put("text", text);
-            requestBody.put("input", text);
 
             HttpHeaders headers = new HttpHeaders();
             headers.setContentType(MediaType.APPLICATION_JSON);
-            headers.set("Authorization", "Bearer " + apiKey);
 
             HttpEntity<Map<String, Object>> request = new HttpEntity<>(requestBody, headers);
-            ResponseEntity<String> response = restTemplate.postForEntity(EMBEDDING_URL, request, String.class);
+            ResponseEntity<String> response = restTemplate.postForEntity(
+                    generationServiceUrl + "/api/v1/generation/embed", request, String.class);
 
             if (response.getStatusCode() == HttpStatus.OK && response.getBody() != null) {
                 Map<String, Object> result = JsonUtils.fromJsonMap(response.getBody());
-                return extractEmbedding(result);
+                if (result != null && result.get("data") != null) {
+                    Map<String, Object> data = (Map<String, Object>) result.get("data");
+                    List<Double> embedding = (List<Double>) data.get("embedding");
+                    return embedding != null ? embedding : List.of();
+                }
             }
             log.error("Embedding API 调用失败: status={}, body={}", response.getStatusCode(), response.getBody());
             return List.of();
@@ -66,19 +65,4 @@ public class EmbeddingClient {
         }
         return result;
     }
-
-    @SuppressWarnings("unchecked")
-    private List<Double> extractEmbedding(Map<String, Object> result) {
-        if (result == null) return List.of();
-        try {
-            List<Map<String, Object>> data = (List<Map<String, Object>>) result.get("data");
-            if (data != null && !data.isEmpty()) {
-                List<Double> embedding = (List<Double>) data.get(0).get("embedding");
-                return embedding != null ? embedding : List.of();
-            }
-        } catch (Exception e) {
-            log.error("解析 Embedding 失败", e);
-        }
-        return List.of();
-    }
 }

frontend/admin/src/App.tsx

@@ -12,6 +12,7 @@ import CustomerSessionList from './pages/CustomerSessionList'
 import SensitiveWordList from './pages/SensitiveWordList'
 import RiskEventList from './pages/RiskEventList'
 import MessageSearch from './pages/MessageSearch'
+import LlmModelList from './pages/LlmModelList'
 
 function App() {
   return (
@@ -31,6 +32,7 @@ function App() {
           <Route path="risk/sensitive-words" element={<SensitiveWordList />} />
           <Route path="risk/events" element={<RiskEventList />} />
           <Route path="risk/message-search" element={<MessageSearch />} />
+          <Route path="llm-models" element={<LlmModelList />} />
         </Route>
       </Routes>
     </BrowserRouter>

frontend/admin/src/components/Layout.tsx

@@ -1,7 +1,7 @@
 import { useEffect, useState } from 'react'
 import { Outlet, useNavigate, useLocation } from 'react-router-dom'
 import { Layout as AntLayout, Menu, Button, message, Badge } from 'antd'
-import { DashboardOutlined, CommentOutlined, BookOutlined, TeamOutlined, UserOutlined, SafetyCertificateOutlined } from '@ant-design/icons'
+import { DashboardOutlined, CommentOutlined, BookOutlined, TeamOutlined, UserOutlined, SafetyCertificateOutlined, RobotOutlined } from '@ant-design/icons'
 import { connectRiskEventStream, disconnectRiskEventStream, requestNotificationPermission, showDesktopNotification } from '../utils/notification'
 
 const { Header, Sider, Content } = AntLayout
@@ -76,6 +76,7 @@ export default function Layout() {
       ],
     },
     { key: '/admins', icon: <TeamOutlined />, label: '管理员账号' },
+    { key: '/llm-models', icon: <RobotOutlined />, label: '大模型管理' },
   ]
 
   const handleLogout = () => {

frontend/admin/src/pages/LlmModelList.tsx

@@ -0,0 +1,354 @@
+import { useState, useEffect } from 'react'
+import {
+  Table, Card, Button, Modal, Form, Input, Switch, Tag, message, Popconfirm,
+  Space, Select
+} from 'antd'
+import {
+  PlusOutlined, EditOutlined, DeleteOutlined, CheckCircleOutlined,
+  ThunderboltOutlined, ApiOutlined
+} from '@ant-design/icons'
+import request from '../utils/request'
+
+interface LlmModel {
+  id: number
+  provider: string
+  modelName: string
+  modelCode: string
+  apiKey: string
+  apiUrl: string
+  isActive: number
+  isDefault: number
+  configJson: string
+  createdAt: string
+}
+
+const providerMap: Record<string, { label: string; color: string }> = {
+  bailian: { label: '百炼(阿里)', color: 'orange' },
+  doubao: { label: '豆包(字节)', color: 'blue' },
+  hunyuan: { label: '混元(腾讯)', color: 'green' },
+}
+
+export default function LlmModelList() {
+  const [models, setModels] = useState<LlmModel[]>([])
+  const [loading, setLoading] = useState(false)
+  const [modalVisible, setModalVisible] = useState(false)
+  const [editingId, setEditingId] = useState<number | null>(null)
+  const [form] = Form.useForm()
+
+  const loadModels = async () => {
+    setLoading(true)
+    try {
+      const res: any = await request.get('/v1/generation/llm-models')
+      setModels(res.data.list || [])
+    } catch (e) {
+      message.error('加载模型列表失败')
+    } finally {
+      setLoading(false)
+    }
+  }
+
+  useEffect(() => {
+    loadModels()
+  }, [])
+
+  const defaultApiUrls: Record<string, string> = {
+    bailian: 'https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation',
+    doubao: 'https://ark.cn-beijing.volces.com/api/v3/chat/completions',
+    hunyuan: 'https://hunyuan.tencentcloudapi.com/',
+  }
+
+  const handleProviderChange = (provider: string) => {
+    const currentUrl = form.getFieldValue('apiUrl')
+    // 如果当前URL为空或是默认URL之一，则自动切换
+    const isDefaultUrl = Object.values(defaultApiUrls).includes(currentUrl)
+    if (!currentUrl || isDefaultUrl) {
+      form.setFieldsValue({ apiUrl: defaultApiUrls[provider] || '' })
+    }
+  }
+
+  const handleAdd = () => {
+    setEditingId(null)
+    form.resetFields()
+    form.setFieldsValue({
+      provider: 'bailian',
+      isActive: 1,
+      isDefault: 0,
+      apiUrl: defaultApiUrls['bailian'],
+      configJson: '{"maxTokens": 1500, "temperature": 0.7}',
+    })
+    setModalVisible(true)
+  }
+
+  const handleEdit = (record: LlmModel) => {
+    setEditingId(record.id)
+    form.setFieldsValue({
+      ...record,
+      isActive: record.isActive === 1,
+      isDefault: record.isDefault === 1,
+    })
+    setModalVisible(true)
+  }
+
+  const handleSave = async (values: any) => {
+    try {
+      const payload = {
+        ...values,
+        isActive: values.isActive ? 1 : 0,
+        isDefault: values.isDefault ? 1 : 0,
+      }
+
+      if (editingId) {
+        await request.put(`/v1/generation/llm-models/${editingId}`, payload)
+        message.success('更新成功')
+      } else {
+        await request.post('/v1/generation/llm-models', payload)
+        message.success('添加成功')
+      }
+      setModalVisible(false)
+      loadModels()
+    } catch (e: any) {
+      message.error(e.response?.data?.message || '保存失败')
+    }
+  }
+
+  const handleDelete = async (id: number) => {
+    try {
+      await request.delete(`/v1/generation/llm-models/${id}`)
+      message.success('删除成功')
+      loadModels()
+    } catch (e: any) {
+      message.error(e.response?.data?.message || '删除失败')
+    }
+  }
+
+  const handleSetDefault = async (id: number) => {
+    try {
+      await request.put(`/v1/generation/llm-models/${id}/default`)
+      message.success('已设置为默认模型')
+      loadModels()
+    } catch (e: any) {
+      message.error(e.response?.data?.message || '设置失败')
+    }
+  }
+
+  const handleToggleActive = async (id: number) => {
+    try {
+      await request.put(`/v1/generation/llm-models/${id}/toggle`)
+      message.success('状态切换成功')
+      loadModels()
+    } catch (e: any) {
+      message.error(e.response?.data?.message || '切换失败')
+    }
+  }
+
+  const handleTest = async (record: LlmModel) => {
+    try {
+      const res: any = await request.post(`/v1/generation/llm-models/${record.id}/test`)
+      const data = res.data
+      Modal.success({
+        title: '连接测试成功',
+        content: (
+          <div>
+            <p><strong>提供商：</strong>{providerMap[record.provider]?.label || record.provider}</p>
+            <p><strong>模型：</strong>{data.modelName}</p>
+            <p><strong>响应时间：</strong>{data.elapsedMs}ms</p>
+            <p><strong>模型回复：</strong>{data.response}</p>
+          </div>
+        ),
+      })
+    } catch (e: any) {
+      message.error(e.response?.data?.message || '连接测试失败')
+    }
+  }
+
+  const columns = [
+    {
+      title: 'ID',
+      dataIndex: 'id',
+      width: 60,
+    },
+    {
+      title: '提供商',
+      dataIndex: 'provider',
+      width: 120,
+      render: (p: string) => (
+        <Tag color={providerMap[p]?.color || 'default'}>
+          {providerMap[p]?.label || p}
+        </Tag>
+      ),
+    },
+    {
+      title: '模型名称',
+      dataIndex: 'modelName',
+      width: 150,
+    },
+    {
+      title: '模型代码',
+      dataIndex: 'modelCode',
+      width: 180,
+    },
+    {
+      title: 'API地址',
+      dataIndex: 'apiUrl',
+      ellipsis: true,
+    },
+    {
+      title: '默认',
+      dataIndex: 'isDefault',
+      width: 80,
+      render: (v: number) =>
+        v === 1 ? <Tag color="gold" icon={<CheckCircleOutlined />}>默认</Tag> : '-',
+    },
+    {
+      title: '状态',
+      dataIndex: 'isActive',
+      width: 100,
+      render: (v: number, record: LlmModel) => (
+        <Switch
+          checked={v === 1}
+          checkedChildren="启用"
+          unCheckedChildren="禁用"
+          onChange={() => handleToggleActive(record.id)}
+        />
+      ),
+    },
+    {
+      title: '操作',
+      width: 220,
+      render: (_: any, record: LlmModel) => (
+        <Space size="small">
+          <Button size="small" icon={<EditOutlined />} onClick={() => handleEdit(record)}>
+            编辑
+          </Button>
+          <Button
+            size="small"
+            icon={<ApiOutlined />}
+            onClick={() => handleTest(record)}
+          >
+            测试
+          </Button>
+          {record.isDefault !== 1 && (
+            <Button
+              size="small"
+              type="primary"
+              ghost
+              icon={<ThunderboltOutlined />}
+              onClick={() => handleSetDefault(record.id)}
+            >
+              设为默认
+            </Button>
+          )}
+          <Popconfirm
+            title="确定删除此模型？"
+            onConfirm={() => handleDelete(record.id)}
+          >
+            <Button size="small" danger icon={<DeleteOutlined />}>
+              删除
+            </Button>
+          </Popconfirm>
+        </Space>
+      ),
+    },
+  ]
+
+  return (
+    <div>
+      <h2>大模型管理</h2>
+      <Card style={{ marginBottom: 16 }}>
+        <div style={{ display: 'flex', justifyContent: 'space-between', alignItems: 'center', marginBottom: 16 }}>
+          <span style={{ color: '#666' }}>
+            管理系统使用的大语言模型，可切换百炼、豆包、混元等不同提供商。
+          </span>
+          <Button type="primary" icon={<PlusOutlined />} onClick={handleAdd}>
+            添加模型
+          </Button>
+        </div>
+        <Table
+          columns={columns}
+          dataSource={models}
+          rowKey="id"
+          loading={loading}
+          pagination={{ pageSize: 10 }}
+          size="small"
+        />
+      </Card>
+
+      <Modal
+        title={editingId ? '编辑模型' : '添加模型'}
+        open={modalVisible}
+        onCancel={() => setModalVisible(false)}
+        onOk={() => form.submit()}
+        width={600}
+        destroyOnClose
+      >
+        <Form
+          form={form}
+          layout="vertical"
+          onFinish={handleSave}
+        >
+          <Form.Item
+            name="provider"
+            label="提供商"
+            rules={[{ required: true, message: '请选择提供商' }]}
+          >
+            <Select placeholder="选择提供商" onChange={handleProviderChange}>
+              <Select.Option value="bailian">百炼(阿里)</Select.Option>
+              <Select.Option value="doubao">豆包(字节)</Select.Option>
+              <Select.Option value="hunyuan">混元(腾讯)</Select.Option>
+            </Select>
+          </Form.Item>
+
+          <Form.Item
+            name="modelName"
+            label="模型显示名称"
+            rules={[{ required: true, message: '请输入模型名称' }]}
+          >
+            <Input placeholder="如：通义千问-Turbo" />
+          </Form.Item>
+
+          <Form.Item
+            name="modelCode"
+            label="模型代码"
+            rules={[{ required: true, message: '请输入模型代码' }]}
+          >
+            <Input placeholder="如：qwen-turbo" />
+          </Form.Item>
+
+          <Form.Item
+            name="apiKey"
+            label="API密钥"
+            rules={[{ required: true, message: '请输入API密钥' }]}
+          >
+            <Input.Password placeholder="sk-..." />
+          </Form.Item>
+
+          <Form.Item
+            name="apiUrl"
+            label="API地址"
+            rules={[{ required: true, message: '请输入API地址' }]}
+          >
+            <Input placeholder="https://..." />
+          </Form.Item>
+
+          <Form.Item
+            name="configJson"
+            label="额外配置(JSON)"
+          >
+            <Input.TextArea
+              rows={3}
+              placeholder='{"maxTokens": 1500, "temperature": 0.7}'
+            />
+          </Form.Item>
+
+          <Form.Item label="启用状态" name="isActive" valuePropName="checked">
+            <Switch checkedChildren="启用" unCheckedChildren="禁用" defaultChecked />
+          </Form.Item>
+
+          <Form.Item label="设为默认" name="isDefault" valuePropName="checked">
+            <Switch checkedChildren="是" unCheckedChildren="否" />
+          </Form.Item>
+        </Form>
+      </Modal>
+    </div>
+  )
+}