feat：【ai 大模型】支持 Chat Role 的 mcp 配置

yudao-module-ai/pom.xml

@@ -197,6 +197,7 @@
             <groupId>org.springframework.ai</groupId>
             <artifactId>spring-ai-starter-mcp-client-webflux</artifactId>
             <version>${spring-ai.version}</version>
+            <optional>true</optional>
         </dependency>
 
         <!-- TinyFlow：AI 工作流 -->

yudao-module-ai/src/main/java/cn/iocoder/yudao/module/ai/controller/admin/model/vo/chatRole/AiChatRoleRespVO.java

@@ -52,6 +52,9 @@ public class AiChatRoleRespVO implements VO {
     @Schema(description = "引用的工具编号列表", example = "1,2,3")
     private List<Long> toolIds;
 
+    @Schema(description = "引用的 MCP Client 名字列表", example = "filesystem")
+    private List<String> mcpClientNames;
+
     @Schema(description = "是否公开", requiredMode = Schema.RequiredMode.REQUIRED, example = "1")
     private Boolean publicStatus;
 

yudao-module-ai/src/main/java/cn/iocoder/yudao/module/ai/controller/admin/model/vo/chatRole/AiChatRoleSaveMyReqVO.java

@@ -37,4 +37,7 @@ public class AiChatRoleSaveMyReqVO {
     @Schema(description = "引用的工具编号列表", example = "1,2,3")
     private List<Long> toolIds;
 
+    @Schema(description = "引用的 MCP Client 名字列表", example = "filesystem")
+    private List<String> mcpClientNames;
+
 }

yudao-module-ai/src/main/java/cn/iocoder/yudao/module/ai/controller/admin/model/vo/chatRole/AiChatRoleSaveReqVO.java

@@ -50,6 +50,9 @@ public class AiChatRoleSaveReqVO {
     @Schema(description = "引用的工具编号列表", example = "1,2,3")
     private List<Long> toolIds;
 
+    @Schema(description = "引用的 MCP Client 名字列表", example = "filesystem")
+    private List<String> mcpClientNames;
+
     @Schema(description = "是否公开", requiredMode = Schema.RequiredMode.REQUIRED, example = "1")
     @NotNull(message = "是否公开不能为空")
     private Boolean publicStatus;

yudao-module-ai/src/main/java/cn/iocoder/yudao/module/ai/dal/dataobject/model/AiChatRoleDO.java

@@ -3,6 +3,7 @@ package cn.iocoder.yudao.module.ai.dal.dataobject.model;
 import cn.iocoder.yudao.framework.common.enums.CommonStatusEnum;
 import cn.iocoder.yudao.framework.mybatis.core.dataobject.BaseDO;
 import cn.iocoder.yudao.framework.mybatis.core.type.LongListTypeHandler;
+import cn.iocoder.yudao.framework.mybatis.core.type.StringListTypeHandler;
 import cn.iocoder.yudao.module.ai.dal.dataobject.knowledge.AiKnowledgeDO;
 import com.baomidou.mybatisplus.annotation.KeySequence;
 import com.baomidou.mybatisplus.annotation.TableField;
@@ -80,6 +81,13 @@ public class AiChatRoleDO extends BaseDO {
      */
     @TableField(typeHandler = LongListTypeHandler.class)
     private List<Long> toolIds;
+    /**
+     * 引用的 MCP Client 名字列表
+     *
+     * 关联 spring.ai.mcp.client 下的名字
+     */
+    @TableField(typeHandler = StringListTypeHandler.class)
+    private List<String> mcpClientNames;
 
     /**
      * 是否公开

yudao-module-ai/src/main/java/cn/iocoder/yudao/module/ai/service/chat/AiChatMessageServiceImpl.java

@@ -36,6 +36,7 @@ import cn.iocoder.yudao.module.ai.service.model.AiToolService;
 import cn.iocoder.yudao.module.ai.util.AiUtils;
 import cn.iocoder.yudao.module.infra.framework.file.core.utils.FileTypeUtils;
 import com.google.common.collect.Maps;
+import io.modelcontextprotocol.client.McpSyncClient;
 import jakarta.annotation.Resource;
 import lombok.extern.slf4j.Slf4j;
 import org.springframework.ai.chat.messages.Message;
@@ -47,6 +48,10 @@ import org.springframework.ai.chat.model.ChatResponse;
 import org.springframework.ai.chat.model.StreamingChatModel;
 import org.springframework.ai.chat.prompt.ChatOptions;
 import org.springframework.ai.chat.prompt.Prompt;
+import org.springframework.ai.mcp.SyncMcpToolCallbackProvider;
+import org.springframework.ai.mcp.client.autoconfigure.properties.McpClientCommonProperties;
+import org.springframework.ai.tool.ToolCallback;
+import org.springframework.ai.tool.resolution.ToolCallbackResolver;
 import org.springframework.beans.factory.annotation.Autowired;
 import org.springframework.stereotype.Service;
 import org.springframework.transaction.annotation.Transactional;
@@ -119,6 +124,17 @@ public class AiChatMessageServiceImpl implements AiChatMessageService {
     @Autowired(required = false) // 由于 yudao.ai.web-search.enable 配置项，可以关闭 AiWebSearchClient 的功能，所以这里只能不强制注入
     private AiWebSearchClient webSearchClient;
 
+    @SuppressWarnings("SpringJavaAutowiredFieldsWarningInspection")
+    @Autowired(required = false) // 由于 yudao.ai.mcp.client.enable 配置项，可以关闭 McpSyncClient 的功能，所以这里只能不强制注入
+    private List<McpSyncClient> mcpClients;
+
+    @SuppressWarnings("SpringJavaAutowiredFieldsWarningInspection")
+    @Autowired(required = false) // 由于 yudao.ai.mcp.client.enable 配置项，可以关闭 McpSyncClient 的功能，所以这里只能不强制注入
+    private McpClientCommonProperties mcpClientCommonProperties;
+
+    @Resource
+    private ToolCallbackResolver toolCallbackResolver;
+
     @Transactional(rollbackFor = Exception.class)
     public AiChatMessageSendRespVO sendMessage(AiChatMessageSendReqVO sendReqVO, Long userId) {
         // 1.1 校验对话存在
@@ -334,23 +350,54 @@ public class AiChatMessageServiceImpl implements AiChatMessageService {
         }
 
         // 2.1 查询 tool 工具
-        Set<String> toolNames = null;
+        List<ToolCallback> toolCallbacks = getToolCallbackListByRoleId(conversation.getRoleId());
-        Map<String,Object> toolContext = Map.of();
+        Map<String,Object> toolContext = CollUtil.isNotEmpty(toolCallbacks) ? AiUtils.buildCommonToolContext()
-        if (conversation.getRoleId() != null) {
+                : Map.of();
-            AiChatRoleDO chatRole = chatRoleService.getChatRole(conversation.getRoleId());
-            if (chatRole != null && CollUtil.isNotEmpty(chatRole.getToolIds())) {
-                toolNames = convertSet(toolService.getToolList(chatRole.getToolIds()), AiToolDO::getName);
-                toolContext = AiUtils.buildCommonToolContext();
-            }
-        }
         // 2.2 构建 ChatOptions 对象
         AiPlatformEnum platform = AiPlatformEnum.validatePlatform(model.getPlatform());
         ChatOptions chatOptions = AiUtils.buildChatOptions(platform, model.getModel(),
                 conversation.getTemperature(), conversation.getMaxTokens(),
-                toolNames, toolContext);
+                toolCallbacks, toolContext);
         return new Prompt(chatMessages, chatOptions);
     }
 
+    private List<ToolCallback> getToolCallbackListByRoleId(Long roleId) {
+        if (roleId == null) {
+            return null;
+        }
+        AiChatRoleDO chatRole = chatRoleService.getChatRole(roleId);
+        if (chatRole == null) {
+            return null;
+        }
+        List<ToolCallback> toolCallbacks = new ArrayList<>();
+        // 1. 通过 toolIds
+        if (CollUtil.isNotEmpty(chatRole.getToolIds())) {
+            Set<String> toolNames = convertSet(toolService.getToolList(chatRole.getToolIds()), AiToolDO::getName);
+            toolNames.forEach(toolName -> {
+                ToolCallback toolCallback = toolCallbackResolver.resolve(toolName);
+                if (toolCallback != null) {
+                    toolCallbacks.add(toolCallback);
+                }
+            });
+        }
+        // 2. 通过 mcpClients
+        if (CollUtil.isNotEmpty(mcpClients) && CollUtil.isNotEmpty(chatRole.getMcpClientNames())) {
+            chatRole.getMcpClientNames().forEach(mcpClientName -> {
+                // 2.1 标准化名字，参考 McpClientAutoConfiguration 的 connectedClientName 方法
+                String finalMcpClientName = mcpClientCommonProperties.getName() + " - " + mcpClientName;
+                // 2.2 匹配对应的 McpSyncClient
+                mcpClients.forEach(mcpClient -> {
+                    if (ObjUtil.notEqual(mcpClient.getClientInfo().name(), finalMcpClientName)) {
+                        return;
+                    }
+                    ToolCallback[] mcpToolCallBacks = new SyncMcpToolCallbackProvider(mcpClient).getToolCallbacks();
+                    CollUtil.addAll(toolCallbacks, mcpToolCallBacks);
+                });
+            });
+        }
+        return toolCallbacks;
+    }
+
     /**
      * 从历史消息中，获得倒序的 n 组消息作为消息上下文
      * <p>

yudao-module-ai/src/main/java/cn/iocoder/yudao/module/ai/util/AiUtils.java

@@ -18,12 +18,10 @@ import org.springframework.ai.deepseek.DeepSeekChatOptions;
 import org.springframework.ai.minimax.MiniMaxChatOptions;
 import org.springframework.ai.ollama.api.OllamaOptions;
 import org.springframework.ai.openai.OpenAiChatOptions;
+import org.springframework.ai.tool.ToolCallback;
 import org.springframework.ai.zhipuai.ZhiPuAiChatOptions;
 
-import java.util.Collections;
+import java.util.*;
-import java.util.HashMap;
-import java.util.Map;
-import java.util.Set;
 
 /**
  * Spring AI 工具类
@@ -40,15 +38,15 @@ public class AiUtils {
     }
 
     public static ChatOptions buildChatOptions(AiPlatformEnum platform, String model, Double temperature, Integer maxTokens,
-                                               Set<String> toolNames, Map<String, Object> toolContext) {
+                                               List<ToolCallback> toolCallbacks, Map<String, Object> toolContext) {
-        toolNames = ObjUtil.defaultIfNull(toolNames, Collections.emptySet());
+        toolCallbacks = ObjUtil.defaultIfNull(toolCallbacks, Collections.emptyList());
         toolContext = ObjUtil.defaultIfNull(toolContext, Collections.emptyMap());
         // noinspection EnhancedSwitchMigration
         switch (platform) {
             case TONG_YI:
                 return DashScopeChatOptions.builder().withModel(model).withTemperature(temperature).withMaxToken(maxTokens)
                         .withEnableThinking(true) // TODO 芋艿：默认都开启 thinking 模式，后续可以让用户配置
-                        .withToolNames(toolNames).withToolContext(toolContext).build();
+                        .withToolCallbacks(toolCallbacks).withToolContext(toolContext).build();
             case YI_YAN:
                 return QianFanChatOptions.builder().model(model).temperature(temperature).maxTokens(maxTokens).build();
             case DEEP_SEEK:
@@ -57,30 +55,30 @@ public class AiUtils {
             case SILICON_FLOW: // 复用 DeepSeek 客户端
             case XING_HUO: // 复用 DeepSeek 客户端
                 return DeepSeekChatOptions.builder().model(model).temperature(temperature).maxTokens(maxTokens)
-                        .toolNames(toolNames).toolContext(toolContext).build();
+                        .toolCallbacks(toolCallbacks).toolContext(toolContext).build();
             case ZHI_PU:
                 return ZhiPuAiChatOptions.builder().model(model).temperature(temperature).maxTokens(maxTokens)
-                        .toolNames(toolNames).toolContext(toolContext).build();
+                        .toolCallbacks(toolCallbacks).toolContext(toolContext).build();
             case MINI_MAX:
                 return MiniMaxChatOptions.builder().model(model).temperature(temperature).maxTokens(maxTokens)
-                        .toolNames(toolNames).toolContext(toolContext).build();
+                        .toolCallbacks(toolCallbacks).toolContext(toolContext).build();
             case MOONSHOT:
                 return MoonshotChatOptions.builder().model(model).temperature(temperature).maxTokens(maxTokens)
-                        .toolNames(toolNames).toolContext(toolContext).build();
+                        .toolCallbacks(toolCallbacks).toolContext(toolContext).build();
             case OPENAI:
             case GEMINI: // 复用 OpenAI 客户端
             case BAI_CHUAN: // 复用 OpenAI 客户端
                 return OpenAiChatOptions.builder().model(model).temperature(temperature).maxTokens(maxTokens)
-                        .toolNames(toolNames).toolContext(toolContext).build();
+                        .toolCallbacks(toolCallbacks).toolContext(toolContext).build();
             case AZURE_OPENAI:
                 return AzureOpenAiChatOptions.builder().deploymentName(model).temperature(temperature).maxTokens(maxTokens)
-                        .toolNames(toolNames).toolContext(toolContext).build();
+                        .toolCallbacks(toolCallbacks).toolContext(toolContext).build();
             case ANTHROPIC:
                 return AnthropicChatOptions.builder().model(model).temperature(temperature).maxTokens(maxTokens)
-                        .toolNames(toolNames).toolContext(toolContext).build();
+                        .toolCallbacks(toolCallbacks).toolContext(toolContext).build();
             case OLLAMA:
                 return OllamaOptions.builder().model(model).temperature(temperature).numPredict(maxTokens)
-                        .toolNames(toolNames).toolContext(toolContext).build();
+                        .toolCallbacks(toolCallbacks).toolContext(toolContext).build();
             default:
                 throw new IllegalArgumentException(StrUtil.format("未知平台({})", platform));
         }

yudao-module-ai/src/test/java/cn/iocoder/yudao/module/ai/framework/ai/core/model/mcp/DouBaoMcpTests.java

@@ -36,44 +36,47 @@ public class DouBaoMcpTests {
 
     @Test
     public void testMcpGetUserInfo() {
-
         // 打印结果
         System.out.println(chatClient.prompt()
                 .user("目前有哪些工具可以使用")
                 .call()
                 .content());
         System.out.println("====================================");
+
         // 打印结果
         System.out.println(chatClient.prompt()
                 .user("小新的年龄是多少")
                 .call()
                 .content());
         System.out.println("====================================");
+
         // 打印结果
         System.out.println(chatClient.prompt()
                 .user("获取小新的基本信息")
                 .call()
                 .content());
         System.out.println("====================================");
+
         // 打印结果
         System.out.println(chatClient.prompt()
                 .user("小新是什么职业的")
                 .call()
                 .content());
         System.out.println("====================================");
+
         // 打印结果
         System.out.println(chatClient.prompt()
                 .user("小新的教育背景")
                 .call()
                 .content());
         System.out.println("====================================");
+
         // 打印结果
         System.out.println(chatClient.prompt()
                 .user("小新的兴趣爱好是什么")
                 .call()
                 .content());
         System.out.println("====================================");
-
     }