diff --git a/main/manager-api/src/main/java/xiaozhi/common/config/SwaggerConfig.java b/main/manager-api/src/main/java/xiaozhi/common/config/SwaggerConfig.java
index cbdb5a19..3efe261c 100644
--- a/main/manager-api/src/main/java/xiaozhi/common/config/SwaggerConfig.java
+++ b/main/manager-api/src/main/java/xiaozhi/common/config/SwaggerConfig.java
@@ -95,6 +95,14 @@ public class SwaggerConfig {
                 .build();
     }
 
+    @Bean
+    public GroupedOpenApi correctWordApi() {
+        return GroupedOpenApi.builder()
+                .group("correct-word")
+                .pathsToMatch("/correct-word/**")
+                .build();
+    }
+
     @Bean
     public OpenAPI customOpenAPI() {
         return new OpenAPI().info(new Info()
diff --git a/main/manager-api/src/main/java/xiaozhi/common/exception/ErrorCode.java b/main/manager-api/src/main/java/xiaozhi/common/exception/ErrorCode.java
index 820f4089..134b66c5 100644
--- a/main/manager-api/src/main/java/xiaozhi/common/exception/ErrorCode.java
+++ b/main/manager-api/src/main/java/xiaozhi/common/exception/ErrorCode.java
@@ -256,4 +256,8 @@ public interface ErrorCode {
     int MCP_ACCESS_POINT_ADDRESS_NO_PERMISSION = 10200; // 没有权限查看该智能体的MCP接入点地址
     int MCP_ACCESS_POINT_ADDRESS_NOT_CONFIGURED = 10201; // 请联系管理员进入参数管理配置mcp接入点地址
     int MCP_ACCESS_POINT_TOOLS_LIST_NO_PERMISSION = 10202; // 没有权限查看该智能体的MCP工具列表
+
+    // 替换词相关错误码
+    int CORRECT_WORD_FILE_NAME_EXISTS = 10203; // 文件名已存在
+    int FILE_SIZE_OVER_LIMIT = 10204; // 文件大小超过限制
 }
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/agent/controller/AgentController.java b/main/manager-api/src/main/java/xiaozhi/modules/agent/controller/AgentController.java
index 77907cc5..5e2435e9 100644
--- a/main/manager-api/src/main/java/xiaozhi/modules/agent/controller/AgentController.java
+++ b/main/manager-api/src/main/java/xiaozhi/modules/agent/controller/AgentController.java
@@ -52,6 +52,7 @@ import xiaozhi.modules.agent.service.AgentContextProviderService;
 import xiaozhi.modules.agent.service.AgentPluginMappingService;
 import xiaozhi.modules.agent.service.AgentService;
 import xiaozhi.modules.agent.service.AgentTemplateService;
+import xiaozhi.modules.correctword.service.CorrectWordFileService;
 import xiaozhi.modules.agent.vo.AgentChatHistoryUserVO;
 import xiaozhi.modules.agent.vo.AgentInfoVO;
 import xiaozhi.modules.device.entity.DeviceEntity;
@@ -73,6 +74,7 @@ public class AgentController {
     private final AgentChatSummaryService agentChatSummaryService;
     private final RedisUtils redisUtils;
     private final AgentTagService agentTagService;
+    private final CorrectWordFileService correctWordFileService;
 
     @GetMapping("/list")
     @Operation(summary = "获取用户智能体列表")
@@ -177,6 +179,8 @@ public class AgentController {
         agentPluginMappingService.deleteByAgentId(id);
         // 删除关联的上下文源配置
         agentContextProviderService.deleteByAgentId(id);
+        // 删除关联的替换词文件关联记录
+        correctWordFileService.deleteMappingsByAgentId(id);
         // 再删除智能体
         agentService.deleteById(id);
         return new Result<>();
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/agent/dao/AgentCorrectWordMappingDao.java b/main/manager-api/src/main/java/xiaozhi/modules/agent/dao/AgentCorrectWordMappingDao.java
new file mode 100644
index 00000000..6e327dcc
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/agent/dao/AgentCorrectWordMappingDao.java
@@ -0,0 +1,21 @@
+package xiaozhi.modules.agent.dao;
+
+import java.util.List;
+
+import org.apache.ibatis.annotations.Mapper;
+import org.apache.ibatis.annotations.Param;
+
+import xiaozhi.common.dao.BaseDao;
+import xiaozhi.modules.agent.entity.AgentCorrectWordMappingEntity;
+
+@Mapper
+public interface AgentCorrectWordMappingDao extends BaseDao<AgentCorrectWordMappingEntity> {
+
+    int deleteByAgentId(@Param("agentId") String agentId);
+
+    int deleteByFileId(@Param("fileId") String fileId);
+
+    int batchInsertMapping(@Param("list") List<AgentCorrectWordMappingEntity> mappings);
+
+    List<AgentCorrectWordMappingEntity> selectByAgentId(@Param("agentId") String agentId);
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/agent/dto/AgentUpdateDTO.java b/main/manager-api/src/main/java/xiaozhi/modules/agent/dto/AgentUpdateDTO.java
index 13d95ddc..f2053ac5 100644
--- a/main/manager-api/src/main/java/xiaozhi/modules/agent/dto/AgentUpdateDTO.java
+++ b/main/manager-api/src/main/java/xiaozhi/modules/agent/dto/AgentUpdateDTO.java
@@ -88,6 +88,9 @@ public class AgentUpdateDTO implements Serializable {
     @Schema(description = "上下文源配置", nullable = true)
     private List<ContextProviderDTO> contextProviders;
 
+    @Schema(description = "替换词文件ID列表", nullable = true)
+    private List<String> correctWordFileIds;
+
     @Data
     @Schema(description = "插件函数信息")
     public static class FunctionInfo implements Serializable {
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/agent/entity/AgentCorrectWordMappingEntity.java b/main/manager-api/src/main/java/xiaozhi/modules/agent/entity/AgentCorrectWordMappingEntity.java
new file mode 100644
index 00000000..6b7d0c5a
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/agent/entity/AgentCorrectWordMappingEntity.java
@@ -0,0 +1,38 @@
+package xiaozhi.modules.agent.entity;
+
+import java.util.Date;
+
+import com.baomidou.mybatisplus.annotation.IdType;
+import com.baomidou.mybatisplus.annotation.TableId;
+import com.baomidou.mybatisplus.annotation.TableName;
+
+import io.swagger.v3.oas.annotations.media.Schema;
+import lombok.Data;
+
+@Data
+@TableName("ai_agent_correct_word_mapping")
+@Schema(description = "智能体替换词文件关联")
+public class AgentCorrectWordMappingEntity {
+
+    @TableId(type = IdType.ASSIGN_UUID)
+    @Schema(description = "主键")
+    private String id;
+
+    @Schema(description = "智能体ID")
+    private String agentId;
+
+    @Schema(description = "替换词文件ID")
+    private String fileId;
+
+    @Schema(description = "创建者")
+    private Long creator;
+
+    @Schema(description = "创建时间")
+    private Date createdAt;
+
+    @Schema(description = "更新者")
+    private Long updater;
+
+    @Schema(description = "更新时间")
+    private Date updatedAt;
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/agent/service/impl/AgentServiceImpl.java b/main/manager-api/src/main/java/xiaozhi/modules/agent/service/impl/AgentServiceImpl.java
index ff8add76..56cd70a6 100644
--- a/main/manager-api/src/main/java/xiaozhi/modules/agent/service/impl/AgentServiceImpl.java
+++ b/main/manager-api/src/main/java/xiaozhi/modules/agent/service/impl/AgentServiceImpl.java
@@ -48,6 +48,7 @@ import xiaozhi.modules.agent.service.AgentService;
 import xiaozhi.modules.agent.service.AgentTagService;
 import xiaozhi.modules.agent.service.AgentTemplateService;
 import xiaozhi.modules.agent.vo.AgentInfoVO;
+import xiaozhi.modules.correctword.service.CorrectWordFileService;
 import xiaozhi.modules.device.entity.DeviceEntity;
 import xiaozhi.modules.device.service.DeviceService;
 import xiaozhi.modules.model.dto.ModelProviderDTO;
@@ -74,6 +75,7 @@ public class AgentServiceImpl extends BaseServiceImpl<AgentDao, AgentEntity> imp
     private final ModelProviderService modelProviderService;
     private final AgentContextProviderService agentContextProviderService;
     private final AgentTagService agentTagService;
+    private final CorrectWordFileService correctWordFileService;
 
     @Override
     public PageData<AgentEntity> adminAgentList(Map<String, Object> params) {
@@ -104,6 +106,10 @@ public class AgentServiceImpl extends BaseServiceImpl<AgentDao, AgentEntity> imp
             agent.setContextProviders(contextProviderEntity.getContextProviders());
         }
 
+        // 查询替换词文件ID列表
+        List<String> correctWordFileIds = correctWordFileService.getAgentCorrectWordFileIds(id);
+        agent.setCorrectWordFileIds(correctWordFileIds);
+
         // 无需额外查询插件列表，已通过SQL查询出来
         return agent;
     }
@@ -427,6 +433,11 @@ public class AgentServiceImpl extends BaseServiceImpl<AgentDao, AgentEntity> imp
             agentContextProviderService.saveOrUpdateByAgentId(contextEntity);
         }
 
+        // 更新替换词文件关联
+        if (dto.getCorrectWordFileIds() != null) {
+            correctWordFileService.saveAgentCorrectWords(agentId, dto.getCorrectWordFileIds());
+        }
+
         boolean b = validateLLMIntentParams(dto.getLlmModelId(), dto.getIntentModelId());
         if (!b) {
             throw new RenException(ErrorCode.LLM_INTENT_PARAMS_MISMATCH);
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/agent/vo/AgentInfoVO.java b/main/manager-api/src/main/java/xiaozhi/modules/agent/vo/AgentInfoVO.java
index 3da8b71c..658ed6f2 100644
--- a/main/manager-api/src/main/java/xiaozhi/modules/agent/vo/AgentInfoVO.java
+++ b/main/manager-api/src/main/java/xiaozhi/modules/agent/vo/AgentInfoVO.java
@@ -25,4 +25,7 @@ public class AgentInfoVO extends AgentEntity
 
     @Schema(description = "上下文源配置")
     private List<ContextProviderDTO> contextProviders;
+
+    @Schema(description = "替换词文件ID列表")
+    private List<String> correctWordFileIds;
 }
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/config/controller/ConfigController.java b/main/manager-api/src/main/java/xiaozhi/modules/config/controller/ConfigController.java
index f21795a3..1bbbcaa4 100644
--- a/main/manager-api/src/main/java/xiaozhi/modules/config/controller/ConfigController.java
+++ b/main/manager-api/src/main/java/xiaozhi/modules/config/controller/ConfigController.java
@@ -1,5 +1,7 @@
 package xiaozhi.modules.config.controller;
 
+import java.util.List;
+
 import org.springframework.web.bind.annotation.PostMapping;
 import org.springframework.web.bind.annotation.RequestBody;
 import org.springframework.web.bind.annotation.RequestMapping;
@@ -12,6 +14,7 @@ import lombok.AllArgsConstructor;
 import xiaozhi.common.utils.Result;
 import xiaozhi.common.validator.ValidatorUtils;
 import xiaozhi.modules.config.dto.AgentModelsDTO;
+import xiaozhi.modules.config.dto.CorrectWordsDTO;
 import xiaozhi.modules.config.service.ConfigService;
 
 /**
@@ -41,4 +44,12 @@ public class ConfigController {
         Object models = configService.getAgentModels(dto.getMacAddress(), dto.getSelectedModule());
         return new Result<Object>().ok(models);
     }
+
+    @PostMapping("correct-words")
+    @Operation(summary = "获取智能体替换词")
+    public Result<Object> getCorrectWords(@Valid @RequestBody CorrectWordsDTO dto) {
+        ValidatorUtils.validateEntity(dto);
+        List<String> list = configService.getCorrectWords(dto.getMacAddress());
+        return new Result<Object>().ok(list);
+    }
 }
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/config/dto/CorrectWordsDTO.java b/main/manager-api/src/main/java/xiaozhi/modules/config/dto/CorrectWordsDTO.java
new file mode 100644
index 00000000..0c689d5d
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/config/dto/CorrectWordsDTO.java
@@ -0,0 +1,14 @@
+package xiaozhi.modules.config.dto;
+
+import io.swagger.v3.oas.annotations.media.Schema;
+import jakarta.validation.constraints.NotBlank;
+import lombok.Data;
+
+@Data
+@Schema(description = "获取智能体替换词DTO")
+public class CorrectWordsDTO {
+
+    @NotBlank(message = "设备MAC地址不能为空")
+    @Schema(description = "设备MAC地址")
+    private String macAddress;
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/config/service/ConfigService.java b/main/manager-api/src/main/java/xiaozhi/modules/config/service/ConfigService.java
index 93845e8e..e234e22d 100644
--- a/main/manager-api/src/main/java/xiaozhi/modules/config/service/ConfigService.java
+++ b/main/manager-api/src/main/java/xiaozhi/modules/config/service/ConfigService.java
@@ -1,11 +1,12 @@
 package xiaozhi.modules.config.service;
 
+import java.util.List;
 import java.util.Map;
 
 public interface ConfigService {
     /**
      * 获取服务器配置
-     * 
+     *
      * @param isCache 是否缓存
      * @return 配置信息
      */
@@ -13,10 +14,18 @@ public interface ConfigService {
 
     /**
      * 获取智能体模型配置
-     * 
+     *
      * @param macAddress     MAC地址
      * @param selectedModule 客户端已实例化的模型
      * @return 模型配置信息
      */
     Map<String, Object> getAgentModels(String macAddress, Map<String, String> selectedModule);
+
+    /**
+     * 获取智能体替换词
+     *
+     * @param macAddress 设备MAC地址
+     * @return 替换词列表，格式如 ["模板1|模板01", "模板2|模板02"]
+     */
+    List<String> getCorrectWords(String macAddress);
 }
\ No newline at end of file
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/config/service/impl/ConfigServiceImpl.java b/main/manager-api/src/main/java/xiaozhi/modules/config/service/impl/ConfigServiceImpl.java
index 278d4299..8e35e0b7 100644
--- a/main/manager-api/src/main/java/xiaozhi/modules/config/service/impl/ConfigServiceImpl.java
+++ b/main/manager-api/src/main/java/xiaozhi/modules/config/service/impl/ConfigServiceImpl.java
@@ -1,10 +1,12 @@
 package xiaozhi.modules.config.service.impl;
 
 import java.util.ArrayList;
+import java.util.Collections;
 import java.util.HashMap;
 import java.util.List;
 import java.util.Map;
 import java.util.Objects;
+import java.util.stream.Collectors;
 
 import org.apache.commons.lang3.StringUtils;
 import org.springframework.stereotype.Service;
@@ -30,7 +32,9 @@ import xiaozhi.modules.agent.service.AgentMcpAccessPointService;
 import xiaozhi.modules.agent.service.AgentPluginMappingService;
 import xiaozhi.modules.agent.service.AgentService;
 import xiaozhi.modules.agent.service.AgentTemplateService;
+import xiaozhi.modules.correctword.service.CorrectWordFileService;
 import xiaozhi.modules.agent.vo.AgentVoicePrintVO;
+import xiaozhi.modules.correctword.vo.CorrectWordSimpleVO;
 import xiaozhi.modules.config.service.ConfigService;
 import xiaozhi.modules.device.entity.DeviceEntity;
 import xiaozhi.modules.device.service.DeviceService;
@@ -58,6 +62,7 @@ public class ConfigServiceImpl implements ConfigService {
     private final AgentContextProviderService agentContextProviderService;
     private final VoiceCloneService cloneVoiceService;
     private final AgentVoicePrintDao agentVoicePrintDao;
+    private final CorrectWordFileService correctWordFileService;
 
     @Override
     public Object getConfig(Boolean isCache) {
@@ -242,6 +247,18 @@ public class ConfigServiceImpl implements ConfigService {
         return result;
     }
 
+    @Override
+    public List<String> getCorrectWords(String macAddress) {
+        DeviceEntity device = deviceService.getDeviceByMacAddress(macAddress);
+        if (device == null) {
+            return Collections.emptyList();
+        }
+        List<CorrectWordSimpleVO> items = correctWordFileService.getAllItemsByAgentId(device.getAgentId());
+        return items.stream()
+                .map(item -> item.getSourceWord() + "|" + item.getTargetWord())
+                .collect(Collectors.toList());
+    }
+
     /**
      * 构建配置信息
      * 
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/correctword/controller/CorrectWordController.java b/main/manager-api/src/main/java/xiaozhi/modules/correctword/controller/CorrectWordController.java
new file mode 100644
index 00000000..5569587f
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/correctword/controller/CorrectWordController.java
@@ -0,0 +1,116 @@
+package xiaozhi.modules.correctword.controller;
+
+import java.net.URLEncoder;
+import java.nio.charset.StandardCharsets;
+import java.util.List;
+import java.util.Map;
+
+import org.apache.shiro.authz.annotation.RequiresPermissions;
+import org.springframework.http.HttpHeaders;
+import org.springframework.http.MediaType;
+import org.springframework.http.ResponseEntity;
+import org.springframework.web.bind.annotation.DeleteMapping;
+import org.springframework.web.bind.annotation.GetMapping;
+import org.springframework.web.bind.annotation.PathVariable;
+import org.springframework.web.bind.annotation.PostMapping;
+import org.springframework.web.bind.annotation.PutMapping;
+import org.springframework.web.bind.annotation.RequestBody;
+import org.springframework.web.bind.annotation.RequestMapping;
+import org.springframework.web.bind.annotation.RequestParam;
+import org.springframework.web.bind.annotation.RestController;
+
+import io.swagger.v3.oas.annotations.Operation;
+import io.swagger.v3.oas.annotations.Parameter;
+import io.swagger.v3.oas.annotations.Parameters;
+import io.swagger.v3.oas.annotations.tags.Tag;
+import jakarta.validation.Valid;
+import lombok.AllArgsConstructor;
+import xiaozhi.common.constant.Constant;
+import xiaozhi.common.page.PageData;
+import xiaozhi.common.utils.Result;
+import xiaozhi.modules.correctword.dto.CorrectWordFileCreateDTO;
+import xiaozhi.modules.correctword.service.CorrectWordFileService;
+import xiaozhi.modules.correctword.vo.CorrectWordFileVO;
+
+@RestController
+@RequestMapping("/correct-word")
+@Tag(name = "替换词管理")
+@AllArgsConstructor
+public class CorrectWordController {
+
+    private final CorrectWordFileService correctWordFileService;
+
+    @PostMapping("/file")
+    @Operation(summary = "创建替换词文件")
+    @RequiresPermissions("sys:role:normal")
+    public Result<CorrectWordFileVO> createFile(@Valid @RequestBody CorrectWordFileCreateDTO dto) {
+        CorrectWordFileVO vo = correctWordFileService.createFile(dto);
+        return new Result<CorrectWordFileVO>().ok(vo);
+    }
+
+    @PutMapping("/file/{fileId}")
+    @Operation(summary = "修改替换词文件")
+    @RequiresPermissions("sys:role:normal")
+    public Result<Void> updateFile(@PathVariable String fileId, @Valid @RequestBody CorrectWordFileCreateDTO dto) {
+        correctWordFileService.updateFile(fileId, dto);
+        return new Result<>();
+    }
+
+    @GetMapping("/file/list")
+    @Operation(summary = "分页获取当前用户替换词文件列表")
+    @RequiresPermissions("sys:role:normal")
+    @Parameters({
+            @Parameter(name = Constant.PAGE, description = "当前页码，从1开始", required = true),
+            @Parameter(name = Constant.LIMIT, description = "每页显示记录数", required = true),
+    })
+    public Result<PageData<CorrectWordFileVO>> listFiles(
+            @Parameter(hidden = true) @RequestParam Map<String, Object> params) {
+        PageData<CorrectWordFileVO> page = correctWordFileService.listFiles(params);
+        return new Result<PageData<CorrectWordFileVO>>().ok(page);
+    }
+
+    @GetMapping("/file/select")
+    @Operation(summary = "智能体获取当前用户替换词文件列表")
+    @RequiresPermissions("sys:role:normal")
+    public Result<List<CorrectWordFileVO>> listAllFiles() {
+        List<CorrectWordFileVO> list = correctWordFileService.listAllFiles();
+        return new Result<List<CorrectWordFileVO>>().ok(list);
+    }
+
+    @GetMapping("/file/download/{fileId}")
+    @Operation(summary = "下载替换词文件")
+    @RequiresPermissions("sys:role:normal")
+    public ResponseEntity<byte[]> downloadFile(@PathVariable String fileId) {
+        CorrectWordFileVO vo = correctWordFileService.getFileContent(fileId);
+        if (vo == null || vo.getContent() == null || vo.getContent().isEmpty()) {
+            return ResponseEntity.notFound().build();
+        }
+        byte[] bytes = String.join("\n", vo.getContent()).getBytes(StandardCharsets.UTF_8);
+        String encodedFileName = URLEncoder.encode(vo.getFileName(), StandardCharsets.UTF_8).replace("+", "%20");
+        String asciiFileName = vo.getFileName().replaceAll("[^\\x00-\\x7F]", "_");
+        return ResponseEntity.ok()
+                .header(HttpHeaders.CONTENT_DISPOSITION,
+                        "attachment; filename=\"" + asciiFileName + "\"; filename*=UTF-8''" + encodedFileName)
+                .contentType(MediaType.APPLICATION_OCTET_STREAM)
+                .body(bytes);
+    }
+
+    @DeleteMapping("/file/{fileId}")
+    @Operation(summary = "删除替换词文件")
+    @RequiresPermissions("sys:role:normal")
+    public Result<Void> deleteFile(@PathVariable String fileId) {
+        correctWordFileService.deleteFile(fileId);
+        return new Result<>();
+    }
+
+    @PostMapping("/file/batch-delete")
+    @Operation(summary = "批量删除替换词文件")
+    @RequiresPermissions("sys:role:normal")
+    public Result<Void> batchDeleteFiles(@RequestBody List<String> fileIds) {
+        if (fileIds == null || fileIds.isEmpty()) {
+            return new Result<>();
+        }
+        correctWordFileService.batchDeleteFiles(fileIds);
+        return new Result<>();
+    }
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/correctword/dao/CorrectWordFileDao.java b/main/manager-api/src/main/java/xiaozhi/modules/correctword/dao/CorrectWordFileDao.java
new file mode 100644
index 00000000..769c23c7
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/correctword/dao/CorrectWordFileDao.java
@@ -0,0 +1,10 @@
+package xiaozhi.modules.correctword.dao;
+
+import org.apache.ibatis.annotations.Mapper;
+
+import xiaozhi.common.dao.BaseDao;
+import xiaozhi.modules.correctword.entity.CorrectWordFileEntity;
+
+@Mapper
+public interface CorrectWordFileDao extends BaseDao<CorrectWordFileEntity> {
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/correctword/dao/CorrectWordItemDao.java b/main/manager-api/src/main/java/xiaozhi/modules/correctword/dao/CorrectWordItemDao.java
new file mode 100644
index 00000000..cd1696b5
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/correctword/dao/CorrectWordItemDao.java
@@ -0,0 +1,15 @@
+package xiaozhi.modules.correctword.dao;
+
+import java.util.List;
+
+import org.apache.ibatis.annotations.Mapper;
+import org.apache.ibatis.annotations.Param;
+
+import xiaozhi.common.dao.BaseDao;
+import xiaozhi.modules.correctword.entity.CorrectWordItemEntity;
+
+@Mapper
+public interface CorrectWordItemDao extends BaseDao<CorrectWordItemEntity> {
+
+    int batchInsert(@Param("list") List<CorrectWordItemEntity> items);
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/correctword/dto/CorrectWordFileCreateDTO.java b/main/manager-api/src/main/java/xiaozhi/modules/correctword/dto/CorrectWordFileCreateDTO.java
new file mode 100644
index 00000000..9941abda
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/correctword/dto/CorrectWordFileCreateDTO.java
@@ -0,0 +1,24 @@
+package xiaozhi.modules.correctword.dto;
+
+import java.util.List;
+
+import io.swagger.v3.oas.annotations.media.Schema;
+import jakarta.validation.constraints.NotBlank;
+import jakarta.validation.constraints.NotEmpty;
+import lombok.Data;
+
+@Data
+@Schema(description = "创建替换词文件DTO")
+public class CorrectWordFileCreateDTO {
+
+    @NotBlank(message = "文件名不能为空")
+    @Schema(description = "文件名")
+    private String fileName;
+
+    @NotEmpty(message = "替换词内容不能为空")
+    @Schema(description = "替换词内容，每条格式：原词|替换词")
+    private List<String> content;
+
+    @Schema(description = "文件大小（字节），不能超过1MB")
+    private Long fileSize;
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/correctword/entity/CorrectWordFileEntity.java b/main/manager-api/src/main/java/xiaozhi/modules/correctword/entity/CorrectWordFileEntity.java
new file mode 100644
index 00000000..1d32623e
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/correctword/entity/CorrectWordFileEntity.java
@@ -0,0 +1,41 @@
+package xiaozhi.modules.correctword.entity;
+
+import java.util.Date;
+
+import com.baomidou.mybatisplus.annotation.IdType;
+import com.baomidou.mybatisplus.annotation.TableId;
+import com.baomidou.mybatisplus.annotation.TableName;
+
+import io.swagger.v3.oas.annotations.media.Schema;
+import lombok.Data;
+
+@Data
+@TableName("ai_agent_correct_word_file")
+@Schema(description = "智能体替换词文件")
+public class CorrectWordFileEntity {
+
+    @TableId(type = IdType.ASSIGN_UUID)
+    @Schema(description = "替换词文件ID")
+    private String id;
+
+    @Schema(description = "原始文件名")
+    private String fileName;
+
+    @Schema(description = "替换词数量")
+    private Integer wordCount;
+
+    @Schema(description = "文件原始内容（用于下载）")
+    private String content;
+
+    @Schema(description = "创建者")
+    private Long creator;
+
+    @Schema(description = "创建时间")
+    private Date createdAt;
+
+    @Schema(description = "更新者")
+    private Long updater;
+
+    @Schema(description = "更新时间")
+    private Date updatedAt;
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/correctword/entity/CorrectWordItemEntity.java b/main/manager-api/src/main/java/xiaozhi/modules/correctword/entity/CorrectWordItemEntity.java
new file mode 100644
index 00000000..42b9f188
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/correctword/entity/CorrectWordItemEntity.java
@@ -0,0 +1,27 @@
+package xiaozhi.modules.correctword.entity;
+
+import com.baomidou.mybatisplus.annotation.IdType;
+import com.baomidou.mybatisplus.annotation.TableId;
+import com.baomidou.mybatisplus.annotation.TableName;
+
+import io.swagger.v3.oas.annotations.media.Schema;
+import lombok.Data;
+
+@Data
+@TableName("ai_agent_correct_word_item")
+@Schema(description = "替换词词条")
+public class CorrectWordItemEntity {
+
+    @TableId(type = IdType.ASSIGN_UUID)
+    @Schema(description = "词条ID")
+    private String id;
+
+    @Schema(description = "所属文件ID")
+    private String fileId;
+
+    @Schema(description = "原词")
+    private String sourceWord;
+
+    @Schema(description = "替换词")
+    private String targetWord;
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/correctword/service/CorrectWordFileService.java b/main/manager-api/src/main/java/xiaozhi/modules/correctword/service/CorrectWordFileService.java
new file mode 100644
index 00000000..31e29c8e
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/correctword/service/CorrectWordFileService.java
@@ -0,0 +1,96 @@
+package xiaozhi.modules.correctword.service;
+
+import java.util.List;
+import java.util.Map;
+
+import xiaozhi.common.page.PageData;
+import xiaozhi.modules.correctword.dto.CorrectWordFileCreateDTO;
+import xiaozhi.modules.correctword.vo.CorrectWordFileVO;
+import xiaozhi.modules.correctword.vo.CorrectWordSimpleVO;
+
+public interface CorrectWordFileService {
+
+    /**
+     * 创建替换词文件
+     *
+     * @param dto 创建参数
+     * @return 文件VO
+     */
+    CorrectWordFileVO createFile(CorrectWordFileCreateDTO dto);
+
+    /**
+     * 修改替换词文件（全量替换词条）
+     *
+     * @param fileId 文件ID
+     * @param dto    修改参数
+     */
+    void updateFile(String fileId, CorrectWordFileCreateDTO dto);
+
+    /**
+     * 获取当前用户的替换词文件列表
+     *
+     * @param params 分页参数
+     * @return 分页数据
+     */
+    PageData<CorrectWordFileVO> listFiles(Map<String, Object> params);
+
+    /**
+     * 获取当前用户的替换词文件列表（不分页，用于下拉选择）
+     *
+     * @return 文件列表
+     */
+    List<CorrectWordFileVO> listAllFiles();
+
+    /**
+     * 获取文件原始内容（用于下载）
+     *
+     * @param fileId 文件ID
+     * @return 文件实体
+     */
+    CorrectWordFileVO getFileContent(String fileId);
+
+    /**
+     * 删除替换词文件及其所有词条和关联记录
+     *
+     * @param fileId 文件ID
+     */
+    void deleteFile(String fileId);
+
+    /**
+     * 删除智能体关联的替换词文件关联记录（不删文件本身）
+     *
+     * @param agentId 智能体ID
+     */
+    void deleteMappingsByAgentId(String agentId);
+
+    /**
+     * 获取智能体的所有替换词条（精简版，供设备端使用）
+     *
+     * @param agentId 智能体ID
+     * @return 替换词列表
+     */
+    List<CorrectWordSimpleVO> getAllItemsByAgentId(String agentId);
+
+    /**
+     * 获取智能体关联的替换词文件ID列表
+     *
+     * @param agentId 智能体ID
+     * @return 文件ID列表
+     */
+    List<String> getAgentCorrectWordFileIds(String agentId);
+
+    /**
+     * 保存智能体关联的替换词文件（全量替换）
+     *
+     * @param agentId 智能体ID
+     * @param fileIds 文件ID列表
+     */
+    void saveAgentCorrectWords(String agentId, List<String> fileIds);
+
+    /**
+     * 批量删除替换词文件
+     *
+     * @param fileIds 文件ID列表
+     */
+    void batchDeleteFiles(List<String> fileIds);
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/correctword/service/impl/CorrectWordFileServiceImpl.java b/main/manager-api/src/main/java/xiaozhi/modules/correctword/service/impl/CorrectWordFileServiceImpl.java
new file mode 100644
index 00000000..b35a0e09
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/correctword/service/impl/CorrectWordFileServiceImpl.java
@@ -0,0 +1,296 @@
+package xiaozhi.modules.correctword.service.impl;
+
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.Date;
+import java.util.List;
+import java.util.Map;
+import java.util.stream.Collectors;
+
+import org.springframework.stereotype.Service;
+import org.springframework.transaction.annotation.Transactional;
+
+import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
+import com.baomidou.mybatisplus.core.metadata.IPage;
+
+import lombok.AllArgsConstructor;
+import xiaozhi.common.exception.ErrorCode;
+import xiaozhi.common.exception.RenException;
+import xiaozhi.common.page.PageData;
+import xiaozhi.common.service.impl.BaseServiceImpl;
+import xiaozhi.common.utils.ConvertUtils;
+import xiaozhi.modules.agent.dao.AgentCorrectWordMappingDao;
+import xiaozhi.modules.correctword.dao.CorrectWordFileDao;
+import xiaozhi.modules.correctword.dao.CorrectWordItemDao;
+import xiaozhi.modules.correctword.dto.CorrectWordFileCreateDTO;
+import xiaozhi.modules.agent.entity.AgentCorrectWordMappingEntity;
+import xiaozhi.modules.correctword.entity.CorrectWordFileEntity;
+import xiaozhi.modules.correctword.entity.CorrectWordItemEntity;
+import xiaozhi.modules.correctword.service.CorrectWordFileService;
+import xiaozhi.modules.correctword.vo.CorrectWordFileVO;
+import xiaozhi.modules.correctword.vo.CorrectWordSimpleVO;
+import xiaozhi.modules.security.user.SecurityUser;
+
+@Service
+@AllArgsConstructor
+public class CorrectWordFileServiceImpl extends BaseServiceImpl<CorrectWordFileDao, CorrectWordFileEntity>
+        implements CorrectWordFileService {
+
+    private final CorrectWordFileDao correctWordFileDao;
+    private final CorrectWordItemDao correctWordItemDao;
+    private final AgentCorrectWordMappingDao agentCorrectWordMappingDao;
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public CorrectWordFileVO createFile(CorrectWordFileCreateDTO dto) {
+        // 校验文件大小不能超过1MB
+        if (dto.getFileSize() != null && dto.getFileSize() > 1024 * 1024) {
+            throw new RenException(ErrorCode.FILE_SIZE_OVER_LIMIT);
+        }
+
+        // 校验文件名是否重复
+        Long userId = SecurityUser.getUserId();
+        LambdaQueryWrapper<CorrectWordFileEntity> nameWrapper = new LambdaQueryWrapper<>();
+        nameWrapper.eq(CorrectWordFileEntity::getCreator, userId)
+                .eq(CorrectWordFileEntity::getFileName, dto.getFileName());
+        if (correctWordFileDao.selectCount(nameWrapper) > 0) {
+            throw new RenException(ErrorCode.CORRECT_WORD_FILE_NAME_EXISTS);
+        }
+
+        List<CorrectWordItemEntity> items = parseContent(dto.getContent());
+
+        // 保存文件记录
+        CorrectWordFileEntity fileEntity = new CorrectWordFileEntity();
+        fileEntity.setFileName(dto.getFileName());
+        fileEntity.setWordCount(items.size());
+        fileEntity.setContent(String.join("\n", dto.getContent()));
+        fileEntity.setCreator(SecurityUser.getUserId());
+        fileEntity.setCreatedAt(new Date());
+        correctWordFileDao.insert(fileEntity);
+
+        // 设置fileId并批量保存词条
+        String fileId = fileEntity.getId();
+        for (CorrectWordItemEntity item : items) {
+            item.setFileId(fileId);
+        }
+        if (!items.isEmpty()) {
+            correctWordItemDao.batchInsert(items);
+        }
+
+        return toVO(fileEntity);
+    }
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public void updateFile(String fileId, CorrectWordFileCreateDTO dto) {
+        CorrectWordFileEntity fileEntity = correctWordFileDao.selectById(fileId);
+        if (fileEntity == null) {
+            return;
+        }
+
+        // 校验文件名是否重复（排除自身）
+        Long userId = SecurityUser.getUserId();
+        LambdaQueryWrapper<CorrectWordFileEntity> nameWrapper = new LambdaQueryWrapper<>();
+        nameWrapper.eq(CorrectWordFileEntity::getCreator, userId)
+                .eq(CorrectWordFileEntity::getFileName, dto.getFileName())
+                .ne(CorrectWordFileEntity::getId, fileId);
+        if (correctWordFileDao.selectCount(nameWrapper) > 0) {
+            throw new RenException("文件名已存在：" + dto.getFileName());
+        }
+
+        // 先删除旧词条
+        LambdaQueryWrapper<CorrectWordItemEntity> deleteWrapper = new LambdaQueryWrapper<>();
+        deleteWrapper.eq(CorrectWordItemEntity::getFileId, fileId);
+        correctWordItemDao.delete(deleteWrapper);
+
+        // 解析新词条并批量保存
+        List<CorrectWordItemEntity> items = parseContent(dto.getContent());
+        if (!items.isEmpty()) {
+            for (CorrectWordItemEntity item : items) {
+                item.setFileId(fileId);
+            }
+            correctWordItemDao.batchInsert(items);
+        }
+
+        // 更新文件记录
+        fileEntity.setFileName(dto.getFileName());
+        fileEntity.setWordCount(items.size());
+        fileEntity.setContent(String.join("\n", dto.getContent()));
+        fileEntity.setUpdater(SecurityUser.getUserId());
+        fileEntity.setUpdatedAt(new Date());
+        correctWordFileDao.updateById(fileEntity);
+    }
+
+    @Override
+    public PageData<CorrectWordFileVO> listFiles(Map<String, Object> params) {
+        Long userId = SecurityUser.getUserId();
+        IPage<CorrectWordFileEntity> page = getPage(params, "created_at", false);
+        LambdaQueryWrapper<CorrectWordFileEntity> wrapper = new LambdaQueryWrapper<>();
+        wrapper.eq(CorrectWordFileEntity::getCreator, userId)
+                .orderByDesc(CorrectWordFileEntity::getCreatedAt);
+        correctWordFileDao.selectPage(page, wrapper);
+        List<CorrectWordFileVO> voList = toVOList(page.getRecords());
+        return new PageData<>(voList, page.getTotal());
+    }
+
+    @Override
+    public List<CorrectWordFileVO> listAllFiles() {
+        Long userId = SecurityUser.getUserId();
+        LambdaQueryWrapper<CorrectWordFileEntity> wrapper = new LambdaQueryWrapper<>();
+        wrapper.eq(CorrectWordFileEntity::getCreator, userId)
+                .orderByDesc(CorrectWordFileEntity::getCreatedAt);
+        List<CorrectWordFileEntity> entities = correctWordFileDao.selectList(wrapper);
+        return toVOList(entities);
+    }
+
+    @Override
+    public CorrectWordFileVO getFileContent(String fileId) {
+        CorrectWordFileEntity entity = correctWordFileDao.selectById(fileId);
+        return toVO(entity);
+    }
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public void deleteFile(String fileId) {
+        if (fileId == null || fileId.trim().isEmpty()) {
+            return;
+        }
+        // 先删除关联表记录
+        agentCorrectWordMappingDao.deleteByFileId(fileId);
+        // 删除词条
+        LambdaQueryWrapper<CorrectWordItemEntity> wrapper = new LambdaQueryWrapper<>();
+        wrapper.eq(CorrectWordItemEntity::getFileId, fileId);
+        correctWordItemDao.delete(wrapper);
+        // 删除文件
+        correctWordFileDao.deleteById(fileId);
+    }
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public void deleteMappingsByAgentId(String agentId) {
+        agentCorrectWordMappingDao.deleteByAgentId(agentId);
+    }
+
+    @Override
+    public List<CorrectWordSimpleVO> getAllItemsByAgentId(String agentId) {
+        // 通过关联表获取文件ID列表
+        List<AgentCorrectWordMappingEntity> mappings = agentCorrectWordMappingDao.selectByAgentId(agentId);
+        if (mappings == null || mappings.isEmpty()) {
+            return new ArrayList<>();
+        }
+        List<String> fileIds = mappings.stream()
+                .map(AgentCorrectWordMappingEntity::getFileId)
+                .collect(Collectors.toList());
+
+        // 根据文件ID列表查询词条
+        LambdaQueryWrapper<CorrectWordItemEntity> wrapper = new LambdaQueryWrapper<>();
+        wrapper.in(CorrectWordItemEntity::getFileId, fileIds);
+        List<CorrectWordItemEntity> entities = correctWordItemDao.selectList(wrapper);
+        return ConvertUtils.sourceToTarget(entities, CorrectWordSimpleVO.class);
+    }
+
+    @Override
+    public List<String> getAgentCorrectWordFileIds(String agentId) {
+        List<AgentCorrectWordMappingEntity> mappings = agentCorrectWordMappingDao.selectByAgentId(agentId);
+        if (mappings == null || mappings.isEmpty()) {
+            return new ArrayList<>();
+        }
+        return mappings.stream()
+                .map(AgentCorrectWordMappingEntity::getFileId)
+                .collect(Collectors.toList());
+    }
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public void saveAgentCorrectWords(String agentId, List<String> fileIds) {
+        // 先删除旧的关联记录
+        agentCorrectWordMappingDao.deleteByAgentId(agentId);
+
+        if (fileIds == null || fileIds.isEmpty()) {
+            return;
+        }
+
+        // 批量插入新的关联记录
+        Long userId = SecurityUser.getUserId();
+        Date now = new Date();
+        List<AgentCorrectWordMappingEntity> mappings = new ArrayList<>();
+        for (String fileId : fileIds) {
+            AgentCorrectWordMappingEntity mapping = new AgentCorrectWordMappingEntity();
+            mapping.setAgentId(agentId);
+            mapping.setFileId(fileId);
+            mapping.setCreator(userId);
+            mapping.setCreatedAt(now);
+            mapping.setUpdater(userId);
+            mapping.setUpdatedAt(now);
+            mappings.add(mapping);
+        }
+        agentCorrectWordMappingDao.batchInsertMapping(mappings);
+    }
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public void batchDeleteFiles(List<String> fileIds) {
+        if (fileIds == null || fileIds.isEmpty()) {
+            return;
+        }
+        for (String fileId : fileIds) {
+            if (fileId == null || fileId.trim().isEmpty()) {
+                continue;
+            }
+            deleteFile(fileId.trim());
+        }
+    }
+
+    /**
+     * 解析替换词内容，每条格式：原词|替换词
+     */
+    private List<CorrectWordItemEntity> parseContent(List<String> lines) {
+        List<CorrectWordItemEntity> items = new ArrayList<>();
+        if (lines == null || lines.isEmpty()) {
+            return items;
+        }
+        for (String line : lines) {
+            line = line.trim();
+            if (line.isEmpty()) {
+                continue;
+            }
+            int idx = line.indexOf('|');
+            if (idx <= 0 || idx >= line.length() - 1) {
+                continue;
+            }
+            String sourceWord = line.substring(0, idx).trim();
+            String targetWord = line.substring(idx + 1).trim();
+            if (sourceWord.isEmpty() || targetWord.isEmpty()) {
+                continue;
+            }
+            CorrectWordItemEntity item = new CorrectWordItemEntity();
+            item.setSourceWord(sourceWord);
+            item.setTargetWord(targetWord);
+            items.add(item);
+        }
+        return items;
+    }
+
+    private CorrectWordFileVO toVO(CorrectWordFileEntity entity) {
+        if (entity == null) {
+            return null;
+        }
+        CorrectWordFileVO vo = new CorrectWordFileVO();
+        vo.setId(entity.getId());
+        vo.setFileName(entity.getFileName());
+        vo.setWordCount(entity.getWordCount());
+        vo.setContent(entity.getContent() != null
+                ? Arrays.asList(entity.getContent().split("\n"))
+                : new ArrayList<>());
+        vo.setCreatedAt(entity.getCreatedAt());
+        vo.setUpdatedAt(entity.getUpdatedAt());
+        return vo;
+    }
+
+    private List<CorrectWordFileVO> toVOList(List<CorrectWordFileEntity> entities) {
+        if (entities == null || entities.isEmpty()) {
+            return new ArrayList<>();
+        }
+        return entities.stream().map(this::toVO).collect(Collectors.toList());
+    }
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/correctword/vo/CorrectWordFileVO.java b/main/manager-api/src/main/java/xiaozhi/modules/correctword/vo/CorrectWordFileVO.java
new file mode 100644
index 00000000..a881b8d5
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/correctword/vo/CorrectWordFileVO.java
@@ -0,0 +1,30 @@
+package xiaozhi.modules.correctword.vo;
+
+import java.util.Date;
+import java.util.List;
+
+import io.swagger.v3.oas.annotations.media.Schema;
+import lombok.Data;
+
+@Data
+@Schema(description = "替换词文件列表VO")
+public class CorrectWordFileVO {
+
+    @Schema(description = "替换词文件ID")
+    private String id;
+
+    @Schema(description = "原始文件名")
+    private String fileName;
+
+    @Schema(description = "替换词数量")
+    private Integer wordCount;
+
+    @Schema(description = "替换词内容，每行一条")
+    private List<String> content;
+
+    @Schema(description = "创建时间")
+    private Date createdAt;
+
+    @Schema(description = "更新时间")
+    private Date updatedAt;
+}
diff --git a/main/manager-api/src/main/java/xiaozhi/modules/correctword/vo/CorrectWordSimpleVO.java b/main/manager-api/src/main/java/xiaozhi/modules/correctword/vo/CorrectWordSimpleVO.java
new file mode 100644
index 00000000..289b8a4e
--- /dev/null
+++ b/main/manager-api/src/main/java/xiaozhi/modules/correctword/vo/CorrectWordSimpleVO.java
@@ -0,0 +1,15 @@
+package xiaozhi.modules.correctword.vo;
+
+import io.swagger.v3.oas.annotations.media.Schema;
+import lombok.Data;
+
+@Data
+@Schema(description = "替换词精简VO（设备端使用）")
+public class CorrectWordSimpleVO {
+
+    @Schema(description = "原词")
+    private String sourceWord;
+
+    @Schema(description = "替换词")
+    private String targetWord;
+}
diff --git a/main/manager-api/src/main/resources/db/changelog/202604211515.sql b/main/manager-api/src/main/resources/db/changelog/202604211515.sql
new file mode 100644
index 00000000..39c4f966
--- /dev/null
+++ b/main/manager-api/src/main/resources/db/changelog/202604211515.sql
@@ -0,0 +1,38 @@
+-- 智能体替换词文件表
+CREATE TABLE IF NOT EXISTS `ai_agent_correct_word_file` (
+    `id`          VARCHAR(32)  NOT NULL,
+    `file_name`   VARCHAR(256) NOT NULL COMMENT '原始文件名',
+    `word_count`  INT          NOT NULL DEFAULT 0 COMMENT '替换词数量',
+    `content`     TEXT         COMMENT '文件原始内容',
+    `creator`     BIGINT       DEFAULT NULL,
+    `created_at`  DATETIME     DEFAULT CURRENT_TIMESTAMP,
+    `updater`     BIGINT       DEFAULT NULL,
+    `updated_at`  DATETIME     DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
+    PRIMARY KEY (`id`),
+    INDEX `idx_creator` (`creator`)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COMMENT='替换词文件';
+
+-- 替换词词条表
+CREATE TABLE IF NOT EXISTS `ai_agent_correct_word_item` (
+    `id`          VARCHAR(32)  NOT NULL,
+    `file_id`     VARCHAR(32)  NOT NULL COMMENT '所属文件ID',
+    `source_word` VARCHAR(128) NOT NULL COMMENT '原词',
+    `target_word` VARCHAR(128) NOT NULL COMMENT '替换词',
+    PRIMARY KEY (`id`),
+    INDEX `idx_file_id` (`file_id`)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COMMENT='替换词词条';
+
+-- 智能体替换词文件关联表
+CREATE TABLE IF NOT EXISTS `ai_agent_correct_word_mapping` (
+    `id`          VARCHAR(32)  NOT NULL,
+    `agent_id`    VARCHAR(32)  NOT NULL,
+    `file_id`     VARCHAR(32)  NOT NULL,
+    `creator`     BIGINT       DEFAULT NULL,
+    `created_at`  DATETIME     DEFAULT CURRENT_TIMESTAMP,
+    `updater`     BIGINT       DEFAULT NULL,
+    `updated_at`  DATETIME     DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
+    PRIMARY KEY (`id`),
+    UNIQUE INDEX `uk_agent_file` (`agent_id`, `file_id`),
+    INDEX `idx_agent_id` (`agent_id`),
+    INDEX `idx_file_id` (`file_id`)
+) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COMMENT='智能体替换词文件关联';
diff --git a/main/manager-api/src/main/resources/db/changelog/db.changelog-master.yaml b/main/manager-api/src/main/resources/db/changelog/db.changelog-master.yaml
index dff9e0ea..725831cb 100755
--- a/main/manager-api/src/main/resources/db/changelog/db.changelog-master.yaml
+++ b/main/manager-api/src/main/resources/db/changelog/db.changelog-master.yaml
@@ -620,6 +620,13 @@ databaseChangeLog:
         - sqlFile:
             encoding: utf8
             path: classpath:db/changelog/202604201719.sql
+  - changeSet:
+      id: 202604211515  
+      author: rainv123
+      changes:
+        - sqlFile:
+            encoding: utf8
+            path: classpath:db/changelog/202604211515.sql
   - changeSet:
       id: 202604211700
       author: RanChen
diff --git a/main/manager-api/src/main/resources/i18n/messages.properties b/main/manager-api/src/main/resources/i18n/messages.properties
index 685fdcde..860bd12a 100644
--- a/main/manager-api/src/main/resources/i18n/messages.properties
+++ b/main/manager-api/src/main/resources/i18n/messages.properties
@@ -209,4 +209,5 @@
 10200=\u6CA1\u6709\u6743\u9650\u67E5\u770B\u8BE5\u667A\u80FD\u4F53\u7684MCP\u63A5\u5165\u70B9\u5730\u5740
 10201=\u8BF7\u8054\u7CFB\u7BA1\u7406\u5458\u8FDB\u5165\u53C2\u6570\u7BA1\u7406\u914D\u7F6Emcp\u63A5\u5165\u70B9\u5730\u5740
 10202=\u6CA1\u6709\u6743\u9650\u67E5\u770B\u8BE5\u667A\u80FD\u4F53\u7684MCP\u5DE5\u5177\u5217\u8868
-
+10203=\u6587\u4EF6\u540D\u79F0\u5DF2\u5B58\u5728
+10204=\u6587\u4EF6\u5927\u5C0F\u8D85\u8FC71MB\u9650\u5236
\ No newline at end of file
diff --git a/main/manager-api/src/main/resources/i18n/messages_de_DE.properties b/main/manager-api/src/main/resources/i18n/messages_de_DE.properties
index 67dc3019..cf70b00c 100644
--- a/main/manager-api/src/main/resources/i18n/messages_de_DE.properties
+++ b/main/manager-api/src/main/resources/i18n/messages_de_DE.properties
@@ -209,4 +209,5 @@
 10200=Keine Berechtigung, die MCP-Endpunktadresse dieses Agenten anzuzeigen
 10201=Bitte kontaktieren Sie den Administrator, um die MCP-Endpunktadresse in der Parameterverwaltung zu konfigurieren
 10202=Keine Berechtigung, die MCP-Tool-Liste dieses Agenten anzuzeigen
-
+10203=Dateiname existiert bereits
+10204=Dateigr\u00f6\u00dfe \u00fcberschreitet 1MB Limit
\ No newline at end of file
diff --git a/main/manager-api/src/main/resources/i18n/messages_en_US.properties b/main/manager-api/src/main/resources/i18n/messages_en_US.properties
index bce68ba9..8354dfc8 100644
--- a/main/manager-api/src/main/resources/i18n/messages_en_US.properties
+++ b/main/manager-api/src/main/resources/i18n/messages_en_US.properties
@@ -209,4 +209,6 @@
 10200=No permission to view the MCP endpoint address of this agent
 10201=Please contact the administrator to configure the MCP endpoint address in parameter management
 10202=No permission to view the MCP tool list of this agent
+10203=File name already exists
+10204=File size exceeds 1MB limit
 
diff --git a/main/manager-api/src/main/resources/i18n/messages_pt_BR.properties b/main/manager-api/src/main/resources/i18n/messages_pt_BR.properties
index b679f253..39926674 100644
--- a/main/manager-api/src/main/resources/i18n/messages_pt_BR.properties
+++ b/main/manager-api/src/main/resources/i18n/messages_pt_BR.properties
@@ -21,7 +21,7 @@
 10015=Exclua primeiro os usu\u00E1rios do departamento
 10016=Falha na implanta\u00E7\u00E3o, sem fluxo
 10017=Diagrama incorreto, verifique
-10018=Falha na exporta\u00E7\u00E3o, ID do modelo � {0}
+10018=Falha na exporta\u00E7\u00E3o, ID do modelo � {0}
 
 10019=Por favor, fa\u00E7a upload do arquivo
 10020=Token n\u00E3o pode ser vazio
@@ -57,7 +57,7 @@
 10046=Erro ao salvar impress\u00E3o vocal, entre em contato com o administrador
 10047=Limite di\u00E1rio de envio atingido
 10048=Erro na inser\u00E7\u00E3o da senha antiga
-10049=LLM configurado n\u00E3o � openai ou Collama
+10049=LLM configurado n\u00E3o � openai ou Collama
 10050=Falha na gera\u00E7\u00E3o do token
 10051=Recurso n\u00E3o existe
 10052=Agente padr\u00E3o n\u00E3o encontrado
@@ -72,12 +72,12 @@
 10061=C\u00F3digo de ativa\u00E7\u00E3o n\u00E3o pode ser vazio
 10062=C\u00F3digo de ativa\u00E7\u00E3o incorreto
 10063=Dispositivo j\u00E1 ativado
-10064=Este modelo � o modelo padr\u00E3o, configure primeiro outro modelo como padr\u00E3o
+10064=Este modelo � o modelo padr\u00E3o, configure primeiro outro modelo como padr\u00E3o
 10065=Falha ao adicionar dados
 10066=Falha ao modificar dados
 10067=C\u00F3digo de verifica\u00E7\u00E3o gr\u00E1fico incorreto
 10068=Registro por telefone n\u00E3o habilitado, fun\u00E7\u00E3o de c\u00F3digo de verifica\u00E7\u00E3o por SMS indispon\u00EDvel
-10069=Nome de usu\u00E1rio n\u00E3o � n\u00FAmero de telefone, por favor insira novamente
+10069=Nome de usu\u00E1rio n\u00E3o � n\u00FAmero de telefone, por favor insira novamente
 10070=Este n\u00FAmero de telefone j\u00E1 est\u00E1 registrado
 10071=N\u00FAmero de telefone inserido n\u00E3o est\u00E1 registrado
 10072=Registro de usu\u00E1rio comum n\u00E3o permitido atualmente
@@ -90,7 +90,7 @@
 10079=Par\u00E2metros de sele\u00E7\u00E3o incompat\u00EDveis entre LLM e reconhecimento de inten\u00E7\u00E3o Intent
 10080=A pessoa ({0}) correspondente a esta impress\u00E3o vocal j\u00E1 est\u00E1 registrada, escolha outra voz para registrar
 10081=Erro ao excluir impress\u00E3o vocal
-10082=Esta modifica\u00E7\u00E3o n\u00E3o � permitida, esta voz j\u00E1 est\u00E1 registrada como impress\u00E3o vocal ({0})
+10082=Esta modifica\u00E7\u00E3o n\u00E3o � permitida, esta voz j\u00E1 est\u00E1 registrada como impress\u00E3o vocal ({0})
 10083=Erro ao modificar impress\u00E3o vocal, entre em contato com o administrador
 10084=Erro no endere\u00E7o da interface de impress\u00E3o vocal, acesse o gerenciamento de par\u00E2metros para modificar o endere\u00E7o
 10085=Dados de \u00E1udio n\u00E3o pertencem a este agente
@@ -205,7 +205,9 @@
 10196=Nome da etiqueta j\u00E1 existe
 10197=Nome da etiqueta n\u00E3o pode ser vazio
 10198=Etiqueta n\u00E3o existe
-10199=Esta opera\u00E7\u00E3o n\u00E3o � suportada na an\u00E1lise de arquivos chineses
+10199=Esta opera\u00E7\u00E3o n\u00E3o � suportada na an\u00E1lise de arquivos chineses
 10200=Sem permiss\u00e3o para visualizar o endere\u00e7o do endpoint MCP deste agente
 10201=Por favor, contate o administrador para configurar o endere\u00e7o do endpoint MCP no gerenciamento de par\u00e2metros
-10202=Sem permiss\u00e3o para visualizar a lista de ferramentas MCP deste agente
\ No newline at end of file
+10202=Sem permiss\u00e3o para visualizar a lista de ferramentas MCP deste agente
+10203=Nome do arquivo j\u00E1 existe
+10204=O tamanho do arquivo excede o limite de 1MB
\ No newline at end of file
diff --git a/main/manager-api/src/main/resources/i18n/messages_vi_VN.properties b/main/manager-api/src/main/resources/i18n/messages_vi_VN.properties
index c5973226..228b92b1 100644
--- a/main/manager-api/src/main/resources/i18n/messages_vi_VN.properties
+++ b/main/manager-api/src/main/resources/i18n/messages_vi_VN.properties
@@ -208,4 +208,6 @@
 10199=T\u1EC7p \u0111ang \u0111\u01B0\u1EE3c ph\u00E2n t\u00EDch, thao t\u00E1c n\u00E0y kh\u00F4ng \u0111\u01B0\u1EE3c h\u1ED7 tr\u1EE3
 10200=Kh\u00f4ng c\u00f3 quy\u1ec1n xem \u0111\u1ecba ch\u1ec9 \u0111i\u1ec3m cu\u1ed1i MCP c\u1ee7a \u0111\u1ea1i l\u00fd n\u00e0y
 10201=Vui l\u00f2ng li\u00ean h\u1ec7 qu\u1ea3n tr\u1ecb vi\u00ean \u0111\u1ec3 c\u1ea5u h\u00ecnh \u0111\u1ecba ch\u1ec9 \u0111i\u1ec3m cu\u1ed1i MCP trong qu\u1ea3n l\u00fd tham s\u1ed1
-10202=Kh\u00f4ng c\u00f3 quy\u1ec1n xem danh s\u00e1ch c\u00f4ng c\u1ee5 MCP c\u1ee7a \u0111\u1ea1i l\u00fd n\u00e0y
\ No newline at end of file
+10202=Kh\u00f4ng c\u00f3 quy\u1ec1n xem danh s\u00e1ch c\u00f4ng c\u1ee5 MCP c\u1ee7a \u0111\u1ea1i l\u00fd n\u00e0y
+10203=T\u00EAn t\u1EC7p \u0111\u00E3 t\u1ED3n t\u1EA1i
+10204=K\u00EDch th\u01b0\u1edbc t\u1ec7p v\u01b0\u1ee3t qu\u00e1 1MB
\ No newline at end of file
diff --git a/main/manager-api/src/main/resources/i18n/messages_zh_CN.properties b/main/manager-api/src/main/resources/i18n/messages_zh_CN.properties
index 9350a892..860bd12a 100644
--- a/main/manager-api/src/main/resources/i18n/messages_zh_CN.properties
+++ b/main/manager-api/src/main/resources/i18n/messages_zh_CN.properties
@@ -209,3 +209,5 @@
 10200=\u6CA1\u6709\u6743\u9650\u67E5\u770B\u8BE5\u667A\u80FD\u4F53\u7684MCP\u63A5\u5165\u70B9\u5730\u5740
 10201=\u8BF7\u8054\u7CFB\u7BA1\u7406\u5458\u8FDB\u5165\u53C2\u6570\u7BA1\u7406\u914D\u7F6Emcp\u63A5\u5165\u70B9\u5730\u5740
 10202=\u6CA1\u6709\u6743\u9650\u67E5\u770B\u8BE5\u667A\u80FD\u4F53\u7684MCP\u5DE5\u5177\u5217\u8868
+10203=\u6587\u4EF6\u540D\u79F0\u5DF2\u5B58\u5728
+10204=\u6587\u4EF6\u5927\u5C0F\u8D85\u8FC71MB\u9650\u5236
\ No newline at end of file
diff --git a/main/manager-api/src/main/resources/i18n/messages_zh_TW.properties b/main/manager-api/src/main/resources/i18n/messages_zh_TW.properties
index 4bd8fbe8..15b61351 100644
--- a/main/manager-api/src/main/resources/i18n/messages_zh_TW.properties
+++ b/main/manager-api/src/main/resources/i18n/messages_zh_TW.properties
@@ -209,3 +209,5 @@
 10200=\u6C92\u6709\u6B0A\u9650\u67E5\u770B\u8A72\u667A\u80FD\u9AD4\u7684MCP\u63A5\u5165\u9EDE\u5730\u5740
 10201=\u8ACB\u806F\u7E6B\u7BA1\u7406\u54E1\u9032\u5165\u53C3\u6578\u7BA1\u7406\u914D\u7F6Emcp\u63A5\u5165\u9EDE\u5730\u5740
 10202=\u6C92\u6709\u6B0A\u9650\u67E5\u770B\u8A72\u667A\u80FD\u9AD4\u7684MCP\u5DE5\u5177\u5217\u8868
+10203=\u6A94\u6848\u540D\u7A31\u5DF2\u5B58\u5728
+10204=\u6A94\u6848\u5927\u5C0F\u8D85\u904E1MB\u9650\u5236
\ No newline at end of file
diff --git a/main/manager-api/src/main/resources/mapper/agent/AgentCorrectWordMappingDao.xml b/main/manager-api/src/main/resources/mapper/agent/AgentCorrectWordMappingDao.xml
new file mode 100644
index 00000000..5cc131af
--- /dev/null
+++ b/main/manager-api/src/main/resources/mapper/agent/AgentCorrectWordMappingDao.xml
@@ -0,0 +1,27 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE mapper PUBLIC "-//mybatis.org//DTD Mapper 3.0//EN" "http://mybatis.org/dtd/mybatis-3-mapper.dtd">
+<mapper namespace="xiaozhi.modules.agent.dao.AgentCorrectWordMappingDao">
+
+    <delete id="deleteByAgentId">
+        DELETE FROM ai_agent_correct_word_mapping WHERE agent_id = #{agentId}
+    </delete>
+
+    <delete id="deleteByFileId">
+        DELETE FROM ai_agent_correct_word_mapping WHERE file_id = #{fileId}
+    </delete>
+
+    <insert id="batchInsertMapping" parameterType="java.util.List">
+        INSERT INTO ai_agent_correct_word_mapping (id, agent_id, file_id, creator, created_at, updater, updated_at)
+        VALUES
+        <foreach collection="list" item="item" separator=",">
+            (#{item.id}, #{item.agentId}, #{item.fileId}, #{item.creator}, #{item.createdAt}, #{item.updater}, #{item.updatedAt})
+        </foreach>
+    </insert>
+
+    <select id="selectByAgentId" resultType="xiaozhi.modules.agent.entity.AgentCorrectWordMappingEntity">
+        SELECT id, agent_id, file_id, creator, created_at, updater, updated_at
+        FROM ai_agent_correct_word_mapping
+        WHERE agent_id = #{agentId}
+    </select>
+
+</mapper>
diff --git a/main/manager-api/src/main/resources/mapper/correctword/CorrectWordItemDao.xml b/main/manager-api/src/main/resources/mapper/correctword/CorrectWordItemDao.xml
new file mode 100644
index 00000000..27ea5f40
--- /dev/null
+++ b/main/manager-api/src/main/resources/mapper/correctword/CorrectWordItemDao.xml
@@ -0,0 +1,13 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE mapper PUBLIC "-//mybatis.org//DTD Mapper 3.0//EN" "http://mybatis.org/dtd/mybatis-3-mapper.dtd">
+<mapper namespace="xiaozhi.modules.correctword.dao.CorrectWordItemDao">
+
+    <insert id="batchInsert" parameterType="java.util.List">
+        INSERT INTO ai_agent_correct_word_item (id, file_id, source_word, target_word)
+        VALUES
+        <foreach collection="list" item="item" separator=",">
+            (#{item.id}, #{item.fileId}, #{item.sourceWord}, #{item.targetWord})
+        </foreach>
+    </insert>
+
+</mapper>
\ No newline at end of file
diff --git a/main/xiaozhi-server/config/config_loader.py b/main/xiaozhi-server/config/config_loader.py
index e0220e33..9c093e58 100644
--- a/main/xiaozhi-server/config/config_loader.py
+++ b/main/xiaozhi-server/config/config_loader.py
@@ -1,7 +1,8 @@
 import os
+import asyncio
 import yaml
 from collections.abc import Mapping
-from config.manage_api_client import init_service, get_server_config, get_agent_models
+from config.manage_api_client import init_service, get_server_config, get_agent_models, get_correct_words
 
 
 def get_project_dir():
@@ -87,7 +88,16 @@ async def get_config_from_api_async(config):
 
 async def get_private_config_from_api(config, device_id, client_id):
     """从Java API获取私有配置"""
-    return await get_agent_models(device_id, client_id, config["selected_module"])
+    results = await asyncio.gather(
+        get_agent_models(device_id, client_id, config["selected_module"]),
+        get_correct_words(device_id),
+        return_exceptions=True
+    )
+    private_config = results[0] if not isinstance(results[0], Exception) else {}
+    correct_words = results[1] if not isinstance(results[1], Exception) else None
+    if correct_words:
+        private_config["correct_words"] = correct_words
+    return private_config
 
 
 def ensure_directories(config):
diff --git a/main/xiaozhi-server/config/manage_api_client.py b/main/xiaozhi-server/config/manage_api_client.py
index e6f7c918..7028f975 100644
--- a/main/xiaozhi-server/config/manage_api_client.py
+++ b/main/xiaozhi-server/config/manage_api_client.py
@@ -183,6 +183,18 @@ async def get_agent_models(
     )
 
 
+async def get_correct_words(mac_address: str) -> Optional[Dict]:
+    """获取智能体替换词"""
+    try:
+        return await ManageApiClient._instance._execute_async_request(
+            "POST", "/config/correct-words",
+            json={"macAddress": mac_address}
+        )
+    except Exception as e:
+        print(f"获取替换词失败: {e}")
+        return None
+
+
 async def generate_and_save_chat_summary(session_id: str) -> Optional[Dict]:
     """生成并保存聊天记录总结"""
     try:
diff --git a/main/xiaozhi-server/core/connection.py b/main/xiaozhi-server/core/connection.py
index 73e5cba5..4503b825 100644
--- a/main/xiaozhi-server/core/connection.py
+++ b/main/xiaozhi-server/core/connection.py
@@ -766,6 +766,13 @@ class ConnectionHandler:
         if private_config.get("context_providers", None) is not None:
             self.config["context_providers"] = private_config["context_providers"]
 
+        # 注入替换词到 TTS 模块配置
+        if private_config.get("correct_words", None) is not None:
+            select_tts_module = self.config["selected_module"]["TTS"]
+            self.config["TTS"][select_tts_module]["correct_words"] = private_config[
+                "correct_words"
+            ]
+
         # 使用 run_in_executor 在线程池中执行 initialize_modules，避免阻塞主循环
         try:
             modules = await self.loop.run_in_executor(
diff --git a/main/xiaozhi-server/core/providers/tts/alibl_stream.py b/main/xiaozhi-server/core/providers/tts/alibl_stream.py
index f3032b5e..4ab30593 100644
--- a/main/xiaozhi-server/core/providers/tts/alibl_stream.py
+++ b/main/xiaozhi-server/core/providers/tts/alibl_stream.py
@@ -194,6 +194,8 @@ class TTSProvider(TTSProviderBase):
 
             # 过滤Markdown
             filtered_text = MarkdownCleaner.clean_markdown(text)
+            if self._correct_words_pattern:
+                filtered_text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], filtered_text)
 
             if filtered_text:
                 # 发送continue-task消息
@@ -480,6 +482,8 @@ class TTSProvider(TTSProviderBase):
 
                     # 发送文本
                     filtered_text = MarkdownCleaner.clean_markdown(text)
+                    if self._correct_words_pattern:
+                        filtered_text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], filtered_text)
                     # 发送continue-task消息
                     continue_task_message = {
                         "header": {
diff --git a/main/xiaozhi-server/core/providers/tts/aliyun_stream.py b/main/xiaozhi-server/core/providers/tts/aliyun_stream.py
index f9581ee8..66a0df9c 100644
--- a/main/xiaozhi-server/core/providers/tts/aliyun_stream.py
+++ b/main/xiaozhi-server/core/providers/tts/aliyun_stream.py
@@ -295,6 +295,8 @@ class TTSProvider(TTSProviderBase):
                 logger.bind(tag=TAG).warning(f"WebSocket连接不存在，终止发送文本")
                 return
             filtered_text = MarkdownCleaner.clean_markdown(text)
+            if self._correct_words_pattern:
+                filtered_text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], filtered_text)
             if filtered_text:
                 run_request = {
                     "header": {
@@ -564,6 +566,8 @@ class TTSProvider(TTSProviderBase):
 
                     # 发送文本合成请求
                     filtered_text = MarkdownCleaner.clean_markdown(text)
+                    if self._correct_words_pattern:
+                        filtered_text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], filtered_text)
                     run_request = {
                         "header": {
                             "message_id": uuid.uuid4().hex,
diff --git a/main/xiaozhi-server/core/providers/tts/base.py b/main/xiaozhi-server/core/providers/tts/base.py
index 4c3dd6be..c2333db3 100644
--- a/main/xiaozhi-server/core/providers/tts/base.py
+++ b/main/xiaozhi-server/core/providers/tts/base.py
@@ -45,6 +45,21 @@ class TTSProviderBase(ABC):
         self.report_on_last = False
         # sentence_id 到文本的映射，用于流式TTS获取正确的字幕文本
         self._sentence_text_map = {}
+        # 加载替换词，用于一次性正则替换
+        raw_words = config.get("correct_words", [])
+        self.correct_words = {}
+        for item in raw_words:
+            parts = item.split("|", 1)
+            if len(parts) == 2:
+                self.correct_words[parts[0]] = parts[1]
+        # 构建正则表达式，使用最长匹配优先（排序后转义拼接）
+        if self.correct_words:
+            # 按key长度降序排列，长的先匹配，避免短词部分干扰
+            sorted_keys = sorted(self.correct_words.keys(), key=len, reverse=True)
+            pattern_str = '|'.join(re.escape(k) for k in sorted_keys)
+            self._correct_words_pattern = re.compile(pattern_str)
+        else:
+            self._correct_words_pattern = None
 
         self.tts_text_buff = []
         self.punctuations = (
@@ -89,7 +104,12 @@ class TTSProviderBase(ABC):
         self.before_stop_play_files.append((file_audio, text))
 
     def to_tts_stream(self, text, opus_handler: Callable[[bytes], None] = None) -> None:
+        # 保留原始文本用于显示/上报
+        original_text = text
         text = MarkdownCleaner.clean_markdown(text)
+        # 使用正则一次性替换，避免重复遍历和部分匹配问题
+        if self._correct_words_pattern:
+            text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], text)
         max_repeat_time = 5
         if self.delete_audio_file:
             # 需要删除文件的直接转为音频数据
@@ -97,7 +117,8 @@ class TTSProviderBase(ABC):
                 try:
                     audio_bytes = asyncio.run(self.text_to_speak(text, None))
                     if audio_bytes:
-                        self.tts_audio_queue.put((SentenceType.FIRST, None, text, getattr(self, 'current_sentence_id', None)))
+                        # 使用原始文本用于显示/上报
+                        self.tts_audio_queue.put((SentenceType.FIRST, None, original_text, getattr(self, 'current_sentence_id', None)))
                         audio_bytes_to_data_stream(
                             audio_bytes,
                             file_type=self.audio_file_type,
@@ -111,16 +132,16 @@ class TTSProviderBase(ABC):
                         max_repeat_time -= 1
                 except Exception as e:
                     logger.bind(tag=TAG).warning(
-                        f"语音生成失败{5 - max_repeat_time + 1}次: {text}，错误: {e}"
+                        f"语音生成失败{5 - max_repeat_time + 1}次: {original_text}，错误: {e}"
                     )
                     max_repeat_time -= 1
             if max_repeat_time > 0:
                 logger.bind(tag=TAG).info(
-                    f"语音生成成功: {text}，重试{5 - max_repeat_time}次"
+                    f"语音生成成功: {original_text}，重试{5 - max_repeat_time}次"
                 )
             else:
                 logger.bind(tag=TAG).error(
-                    f"语音生成失败: {text}，请检查网络或服务是否正常"
+                    f"语音生成失败: {original_text}，请检查网络或服务是否正常"
                 )
             return None
         else:
@@ -131,7 +152,7 @@ class TTSProviderBase(ABC):
                         asyncio.run(self.text_to_speak(text, tmp_file))
                     except Exception as e:
                         logger.bind(tag=TAG).warning(
-                            f"语音生成失败{5 - max_repeat_time + 1}次: {text}，错误: {e}"
+                            f"语音生成失败{5 - max_repeat_time + 1}次: {original_text}，错误: {e}"
                         )
                         # 未执行成功，删除文件
                         if os.path.exists(tmp_file):
@@ -140,20 +161,24 @@ class TTSProviderBase(ABC):
 
                 if max_repeat_time > 0:
                     logger.bind(tag=TAG).info(
-                        f"语音生成成功: {text}:{tmp_file}，重试{5 - max_repeat_time}次"
+                        f"语音生成成功: {original_text}:{tmp_file}，重试{5 - max_repeat_time}次"
                     )
                 else:
                     logger.bind(tag=TAG).error(
-                        f"语音生成失败: {text}，请检查网络或服务是否正常"
+                        f"语音生成失败: {original_text}，请检查网络或服务是否正常"
                     )
-                self.tts_audio_queue.put((SentenceType.FIRST, None, text, getattr(self, 'current_sentence_id', None)))
+                self.tts_audio_queue.put((SentenceType.FIRST, None, original_text, getattr(self, 'current_sentence_id', None)))
                 self._process_audio_file_stream(tmp_file, callback=opus_handler)
             except Exception as e:
                 logger.bind(tag=TAG).error(f"Failed to generate TTS file: {e}")
                 return None
     
     def to_tts(self, text):
+        # 保留原始文本用于日志/显示
+        original_text = text
         text = MarkdownCleaner.clean_markdown(text)
+        if self._correct_words_pattern:
+            text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], text)
         max_repeat_time = 5
         if self.delete_audio_file:
             # 需要删除文件的直接转为音频数据
@@ -174,16 +199,16 @@ class TTSProviderBase(ABC):
                         max_repeat_time -= 1
                 except Exception as e:
                     logger.bind(tag=TAG).warning(
-                        f"语音生成失败{5 - max_repeat_time + 1}次: {text}，错误: {e}"
+                        f"语音生成失败{5 - max_repeat_time + 1}次: {original_text}，错误: {e}"
                     )
                     max_repeat_time -= 1
             if max_repeat_time > 0:
                 logger.bind(tag=TAG).info(
-                    f"语音生成成功: {text}，重试{5 - max_repeat_time}次"
+                    f"语音生成成功: {original_text}，重试{5 - max_repeat_time}次"
                 )
             else:
                 logger.bind(tag=TAG).error(
-                    f"语音生成失败: {text}，请检查网络或服务是否正常"
+                    f"语音生成失败: {original_text}，请检查网络或服务是否正常"
                 )
             return None
         else:
@@ -194,7 +219,7 @@ class TTSProviderBase(ABC):
                         asyncio.run(self.text_to_speak(text, tmp_file))
                     except Exception as e:
                         logger.bind(tag=TAG).warning(
-                            f"语音生成失败{5 - max_repeat_time + 1}次: {text}，错误: {e}"
+                            f"语音生成失败{5 - max_repeat_time + 1}次: {original_text}，错误: {e}"
                         )
                         # 未执行成功，删除文件
                         if os.path.exists(tmp_file):
@@ -203,11 +228,11 @@ class TTSProviderBase(ABC):
 
                 if max_repeat_time > 0:
                     logger.bind(tag=TAG).info(
-                        f"语音生成成功: {text}:{tmp_file}，重试{5 - max_repeat_time}次"
+                        f"语音生成成功: {original_text}:{tmp_file}，重试{5 - max_repeat_time}次"
                     )
                 else:
                     logger.bind(tag=TAG).error(
-                        f"语音生成失败: {text}，请检查网络或服务是否正常"
+                        f"语音生成失败: {original_text}，请检查网络或服务是否正常"
                     )
 
                 return tmp_file
diff --git a/main/xiaozhi-server/core/providers/tts/huoshan_double_stream.py b/main/xiaozhi-server/core/providers/tts/huoshan_double_stream.py
index 0d783754..a6ecc21c 100644
--- a/main/xiaozhi-server/core/providers/tts/huoshan_double_stream.py
+++ b/main/xiaozhi-server/core/providers/tts/huoshan_double_stream.py
@@ -370,6 +370,8 @@ class TTSProvider(TTSProviderBase):
 
             #  过滤Markdown
             filtered_text = MarkdownCleaner.clean_markdown(text)
+            if self._correct_words_pattern:
+                filtered_text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], filtered_text)
 
             if filtered_text:
                 # 发送文本
diff --git a/main/xiaozhi-server/core/providers/tts/index_stream.py b/main/xiaozhi-server/core/providers/tts/index_stream.py
index 72d2d68d..7314ac9c 100644
--- a/main/xiaozhi-server/core/providers/tts/index_stream.py
+++ b/main/xiaozhi-server/core/providers/tts/index_stream.py
@@ -92,22 +92,25 @@ class TTSProvider(TTSProviderBase):
     def to_tts_single_stream(self, text, is_last=False):
         try:
             max_repeat_time = 5
+            original_text = text
             text = MarkdownCleaner.clean_markdown(text)
+            if self._correct_words_pattern:
+                text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], text)
             try:
                 asyncio.run(self.text_to_speak(text, is_last))
             except Exception as e:
                 logger.bind(tag=TAG).warning(
-                    f"语音生成失败{5 - max_repeat_time + 1}次: {text}，错误: {e}"
+                    f"语音生成失败{5 - max_repeat_time + 1}次: {original_text}，错误: {e}"
                 )
                 max_repeat_time -= 1
 
             if max_repeat_time > 0:
                 logger.bind(tag=TAG).info(
-                    f"语音生成成功: {text}，重试{5 - max_repeat_time}次"
+                    f"语音生成成功: {original_text}，重试{5 - max_repeat_time}次"
                 )
             else:
                 logger.bind(tag=TAG).error(
-                    f"语音生成失败: {text}，请检查网络或服务是否正常"
+                    f"语音生成失败: {original_text}，请检查网络或服务是否正常"
                 )
         except Exception as e:
             logger.bind(tag=TAG).error(f"Failed to generate TTS file: {e}")
@@ -203,6 +206,8 @@ class TTSProvider(TTSProviderBase):
         """
         start_time = time.time()
         text = MarkdownCleaner.clean_markdown(text)
+        if self._correct_words_pattern:
+            text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], text)
 
         payload = {"text": text, "character": self.voice}
 
diff --git a/main/xiaozhi-server/core/providers/tts/linkerai.py b/main/xiaozhi-server/core/providers/tts/linkerai.py
index 7de5f3cd..ecfb8b38 100644
--- a/main/xiaozhi-server/core/providers/tts/linkerai.py
+++ b/main/xiaozhi-server/core/providers/tts/linkerai.py
@@ -84,22 +84,25 @@ class TTSProvider(TTSProviderBase):
     def to_tts_single_stream(self, text, is_last=False):
         try:
             max_repeat_time = 5
+            original_text = text
             text = MarkdownCleaner.clean_markdown(text)
+            if self._correct_words_pattern:
+                text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], text)
             try:
                 asyncio.run(self.text_to_speak(text, is_last))
             except Exception as e:
                 logger.bind(tag=TAG).warning(
-                    f"语音生成失败{5 - max_repeat_time + 1}次: {text}，错误: {e}"
+                    f"语音生成失败{5 - max_repeat_time + 1}次: {original_text}，错误: {e}"
                 )
                 max_repeat_time -= 1
 
             if max_repeat_time > 0:
                 logger.bind(tag=TAG).info(
-                    f"语音生成成功: {text}，重试{5 - max_repeat_time}次"
+                    f"语音生成成功: {original_text}，重试{5 - max_repeat_time}次"
                 )
             else:
                 logger.bind(tag=TAG).error(
-                    f"语音生成失败: {text}，请检查网络或服务是否正常"
+                    f"语音生成失败: {original_text}，请检查网络或服务是否正常"
                 )
         except Exception as e:
             logger.bind(tag=TAG).error(f"Failed to generate TTS file: {e}")
@@ -202,6 +205,8 @@ class TTSProvider(TTSProviderBase):
         """
         start_time = time.time()
         text = MarkdownCleaner.clean_markdown(text)
+        if self._correct_words_pattern:
+            text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], text)
 
         params = {
             "tts_text": text,
diff --git a/main/xiaozhi-server/core/providers/tts/minimax_httpstream.py b/main/xiaozhi-server/core/providers/tts/minimax_httpstream.py
index b5b0cd37..f620adfb 100644
--- a/main/xiaozhi-server/core/providers/tts/minimax_httpstream.py
+++ b/main/xiaozhi-server/core/providers/tts/minimax_httpstream.py
@@ -148,22 +148,25 @@ class TTSProvider(TTSProviderBase):
     def to_tts_single_stream(self, text, is_last=False):
         try:
             max_repeat_time = 5
+            original_text = text
             text = MarkdownCleaner.clean_markdown(text)
+            if self._correct_words_pattern:
+                text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], text)
             try:
                 asyncio.run(self.text_to_speak(text, is_last))
             except Exception as e:
                 logger.bind(tag=TAG).warning(
-                    f"语音生成失败{5 - max_repeat_time + 1}次: {text}，错误: {e}"
+                    f"语音生成失败{5 - max_repeat_time + 1}次: {original_text}，错误: {e}"
                 )
                 max_repeat_time -= 1
 
             if max_repeat_time > 0:
                 logger.bind(tag=TAG).info(
-                    f"语音生成成功: {text}，重试{5 - max_repeat_time}次"
+                    f"语音生成成功: {original_text}，重试{5 - max_repeat_time}次"
                 )
             else:
                 logger.bind(tag=TAG).error(
-                    f"语音生成失败: {text}，请检查网络或服务是否正常"
+                    f"语音生成失败: {original_text}，请检查网络或服务是否正常"
                 )
         except Exception as e:
             logger.bind(tag=TAG).error(f"Failed to generate TTS file: {e}")
@@ -298,6 +301,8 @@ class TTSProvider(TTSProviderBase):
         """
         start_time = time.time()
         text = MarkdownCleaner.clean_markdown(text)
+        if self._correct_words_pattern:
+            text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], text)
 
         payload = {
             "model": self.model,
diff --git a/main/xiaozhi-server/core/providers/tts/xunfei_stream.py b/main/xiaozhi-server/core/providers/tts/xunfei_stream.py
index 6cdb8e19..812c282b 100644
--- a/main/xiaozhi-server/core/providers/tts/xunfei_stream.py
+++ b/main/xiaozhi-server/core/providers/tts/xunfei_stream.py
@@ -245,6 +245,8 @@ class TTSProvider(TTSProviderBase):
                 return
 
             filtered_text = MarkdownCleaner.clean_markdown(text)
+            if self._correct_words_pattern:
+                filtered_text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], filtered_text)
             if filtered_text:
                 # 发送文本合成请求
                 run_request = self._build_base_request(status=1,text=filtered_text)
@@ -441,6 +443,8 @@ class TTSProvider(TTSProviderBase):
 
                 try:
                     filtered_text = MarkdownCleaner.clean_markdown(text)
+                    if self._correct_words_pattern:
+                        filtered_text = self._correct_words_pattern.sub(lambda m: self.correct_words[m.group(0)], filtered_text)
 
                     text_request = self._build_base_request(status=2,text=filtered_text)