Merge remote-tracking branch 'origin/main' into agent-snapshot-history-fixes

main/manager-api/src/main/java/xiaozhi/modules/agent/controller/AgentController.java

@@ -28,6 +28,8 @@ import io.swagger.v3.oas.annotations.tags.Tag;
 import jakarta.validation.Valid;
 import lombok.AllArgsConstructor;
 import xiaozhi.common.constant.Constant;
+import xiaozhi.common.exception.ErrorCode;
+import xiaozhi.common.exception.RenException;
 import xiaozhi.common.page.PageData;
 import xiaozhi.common.redis.RedisKeys;
 import xiaozhi.common.redis.RedisUtils;
@@ -52,8 +54,6 @@ import xiaozhi.modules.agent.service.AgentService;
 import xiaozhi.modules.agent.service.AgentTemplateService;
 import xiaozhi.modules.agent.vo.AgentChatHistoryUserVO;
 import xiaozhi.modules.agent.vo.AgentInfoVO;
-import xiaozhi.modules.device.entity.DeviceEntity;
-import xiaozhi.modules.device.service.DeviceService;
 import xiaozhi.modules.security.user.SecurityUser;
 
 @Tag(name = "智能体管理")
@@ -61,15 +61,40 @@ import xiaozhi.modules.security.user.SecurityUser;
 @RestController
 @RequestMapping("/agent")
 public class AgentController {
+    private static final long AUDIO_PLAY_TOKEN_EXPIRE_SECONDS = 300L;
+
     private final AgentService agentService;
     private final AgentTemplateService agentTemplateService;
-    private final DeviceService deviceService;
     private final AgentChatHistoryService agentChatHistoryService;
     private final AgentChatAudioService agentChatAudioService;
     private final AgentChatSummaryService agentChatSummaryService;
     private final RedisUtils redisUtils;
     private final AgentTagService agentTagService;
 
+    private void requireAgentPermission(String agentId) {
+        if (!agentService.checkAgentPermission(agentId, SecurityUser.getUserId())) {
+            throw new RenException(ErrorCode.NO_PERMISSION);
+        }
+    }
+
+    private String requireSessionAgent(String sessionId) {
+        String agentId = agentChatHistoryService.getAgentIdBySessionId(sessionId);
+        if (StringUtils.isBlank(agentId)) {
+            throw new RenException(ErrorCode.AGENT_NOT_FOUND);
+        }
+        agentService.getAgentById(agentId);
+        return agentId;
+    }
+
+    private String requireAudioPermission(String audioId) {
+        String agentId = agentChatHistoryService.getAgentIdByAudioId(audioId);
+        if (StringUtils.isBlank(agentId)) {
+            throw new RenException(ErrorCode.NO_PERMISSION);
+        }
+        requireAgentPermission(agentId);
+        return agentId;
+    }
+
     @GetMapping("/list")
     @Operation(summary = "获取用户智能体列表")
     @RequiresPermissions("sys:role:normal")
@@ -100,7 +125,7 @@ public class AgentController {
     @Operation(summary = "获取智能体详情")
     @RequiresPermissions("sys:role:normal")
     public Result<AgentInfoVO> getAgentById(@PathVariable("id") String id) {
-        AgentInfoVO agent = agentService.getAgentById(id);
+        AgentInfoVO agent = agentService.getAgentById(id, SecurityUser.getUserId());
         return ResultUtils.success(agent);
     }
 
@@ -114,20 +139,16 @@ public class AgentController {
 
     @PutMapping("/saveMemory/{macAddress}")
     @Operation(summary = "根据设备id更新智能体")
+    @RequiresPermissions("sys:role:normal")
     public Result<Void> updateByDeviceId(@PathVariable String macAddress, @RequestBody @Valid AgentMemoryDTO dto) {
-        DeviceEntity device = deviceService.getDeviceByMacAddress(macAddress);
+        agentService.updateAgentMemoryByDeviceMacAddress(macAddress, dto, SecurityUser.getUserId());
-        if (device == null) {
+        return new Result<Void>().ok(null);
-            return new Result<>();
-        }
-        AgentUpdateDTO agentUpdateDTO = new AgentUpdateDTO();
-        agentUpdateDTO.setSummaryMemory(dto.getSummaryMemory());
-        agentService.updateAgentById(device.getAgentId(), agentUpdateDTO, false);
-        return new Result<>();
     }
 
     @PostMapping("/chat-summary/{sessionId}/save")
     @Operation(summary = "根据会话ID生成聊天记录总结并保存（异步执行）")
     public Result<Void> generateAndSaveChatSummary(@PathVariable String sessionId) {
+        requireSessionAgent(sessionId);
         try {
             // 异步执行总结生成任务，立即返回成功响应
             new Thread(() -> {
@@ -149,6 +170,7 @@ public class AgentController {
     @PostMapping("/chat-title/{sessionId}/generate")
     @Operation(summary = "根据会话ID生成聊天标题")
     public Result<Void> generateAndSaveChatTitle(@PathVariable String sessionId) {
+        requireSessionAgent(sessionId);
         agentChatSummaryService.generateAndSaveChatTitle(sessionId);
         return new Result<Void>().ok(null);
     }
@@ -157,7 +179,7 @@ public class AgentController {
     @Operation(summary = "更新智能体")
     @RequiresPermissions("sys:role:normal")
     public Result<Void> update(@PathVariable String id, @RequestBody @Valid AgentUpdateDTO dto) {
-        agentService.updateAgentById(id, dto);
+        agentService.updateAgentById(id, dto, SecurityUser.getUserId());
         return new Result<>();
     }
 
@@ -165,7 +187,7 @@ public class AgentController {
     @Operation(summary = "删除智能体")
     @RequiresPermissions("sys:role:normal")
     public Result<Void> delete(@PathVariable String id) {
-        agentService.deleteAgent(id);
+        agentService.deleteAgentById(id, SecurityUser.getUserId());
         return new Result<>();
     }
 
@@ -188,6 +210,7 @@ public class AgentController {
     public Result<PageData<AgentChatSessionDTO>> getAgentSessions(
             @PathVariable("id") String id,
             @Parameter(hidden = true) @RequestParam Map<String, Object> params) {
+        requireAgentPermission(id);
         params.put("agentId", id);
         PageData<AgentChatSessionDTO> page = agentChatHistoryService.getSessionListByAgentId(params);
         return new Result<PageData<AgentChatSessionDTO>>().ok(page);
@@ -235,6 +258,7 @@ public class AgentController {
     @RequiresPermissions("sys:role:normal")
     public Result<String> getContentByAudioId(
             @PathVariable("id") String id) {
+        requireAudioPermission(id);
         // 查询聊天记录
         String data = agentChatHistoryService.getContentByAudioId(id);
         return new Result<String>().ok(data);
@@ -244,12 +268,13 @@ public class AgentController {
     @Operation(summary = "获取音频下载ID")
     @RequiresPermissions("sys:role:normal")
     public Result<String> getAudioId(@PathVariable("audioId") String audioId) {
+        requireAudioPermission(audioId);
         byte[] audioData = agentChatAudioService.getAudio(audioId);
         if (audioData == null) {
             return new Result<String>().error("音频不存在");
         }
         String uuid = UUID.randomUUID().toString();
-        redisUtils.set(RedisKeys.getAgentAudioIdKey(uuid), audioId);
+        redisUtils.set(RedisKeys.getAgentAudioIdKey(uuid), audioId, AUDIO_PLAY_TOKEN_EXPIRE_SECONDS);
         return new Result<String>().ok(uuid);
     }
 
@@ -305,6 +330,7 @@ public class AgentController {
     @Operation(summary = "获取智能体的标签")
     @RequiresPermissions("sys:role:normal")
     public Result<List<AgentTagDTO>> getAgentTags(@PathVariable String id) {
+        requireAgentPermission(id);
         List<AgentTagDTO> tags = agentTagService.getTagsByAgentId(id);
         return new Result<List<AgentTagDTO>>().ok(tags);
     }
@@ -313,6 +339,7 @@ public class AgentController {
     @Operation(summary = "保存智能体的标签")
     @RequiresPermissions("sys:role:normal")
     public Result<Void> saveAgentTags(@PathVariable String id, @RequestBody Map<String, Object> params) {
+        requireAgentPermission(id);
         List<String> tagIds = (List<String>) params.get("tagIds");
         List<String> tagNames = (List<String>) params.get("tagNames");
         AgentUpdateDTO dto = new AgentUpdateDTO();

main/manager-api/src/main/java/xiaozhi/modules/agent/service/AgentChatHistoryService.java

@@ -37,6 +37,14 @@ public interface AgentChatHistoryService extends IService<AgentChatHistoryEntity
      */
     List<AgentChatHistoryDTO> getChatHistoryBySessionId(String agentId, String sessionId);
 
+    /**
+     * 根据会话ID获取智能体ID
+     *
+     * @param sessionId 会话ID
+     * @return 智能体ID
+     */
+    String getAgentIdBySessionId(String sessionId);
+
     /**
      * 根据智能体ID删除聊天记录
      *
@@ -62,6 +70,14 @@ public interface AgentChatHistoryService extends IService<AgentChatHistoryEntity
      */
     String getContentByAudioId(String audioId);
 
+    /**
+     * 根据音频ID获取智能体ID
+     *
+     * @param audioId 音频ID
+     * @return 智能体ID
+     */
+    String getAgentIdByAudioId(String audioId);
+
 
     /**
      * 查询此音频id是否属于此智能体

main/manager-api/src/main/java/xiaozhi/modules/agent/service/AgentService.java

@@ -7,6 +7,7 @@ import xiaozhi.common.page.PageData;
 import xiaozhi.common.service.BaseService;
 import xiaozhi.modules.agent.dto.AgentCreateDTO;
 import xiaozhi.modules.agent.dto.AgentDTO;
+import xiaozhi.modules.agent.dto.AgentMemoryDTO;
 import xiaozhi.modules.agent.dto.AgentUpdateDTO;
 import xiaozhi.modules.agent.entity.AgentEntity;
 import xiaozhi.modules.agent.vo.AgentInfoVO;
@@ -35,6 +36,15 @@ public interface AgentService extends BaseService<AgentEntity> {
      */
     AgentInfoVO getAgentById(String id);
 
+    /**
+     * 根据ID获取当前用户有权访问的智能体
+     *
+     * @param id     智能体ID
+     * @param userId 当前用户ID
+     * @return 智能体实体
+     */
+    AgentInfoVO getAgentById(String id, Long userId);
+
     /**
      * 插入智能体
      *
@@ -100,6 +110,32 @@ public interface AgentService extends BaseService<AgentEntity> {
      */
     void updateAgentById(String agentId, AgentUpdateDTO dto);
 
+    /**
+     * 更新当前用户有权访问的智能体
+     *
+     * @param agentId 智能体ID
+     * @param dto     更新智能体所需的信息
+     * @param userId  当前用户ID
+     */
+    void updateAgentById(String agentId, AgentUpdateDTO dto, Long userId);
+
+    /**
+     * 根据设备MAC地址更新当前用户有权访问的智能体记忆
+     *
+     * @param macAddress 设备MAC地址
+     * @param dto        智能体记忆
+     * @param userId     当前用户ID
+     */
+    void updateAgentMemoryByDeviceMacAddress(String macAddress, AgentMemoryDTO dto, Long userId);
+
+    /**
+     * 删除当前用户有权访问的智能体
+     *
+     * @param agentId 智能体ID
+     * @param userId  当前用户ID
+     */
+    void deleteAgentById(String agentId, Long userId);
+
     /**
      * 更新智能体
      *

main/manager-api/src/main/java/xiaozhi/modules/agent/service/impl/AgentChatHistoryServiceImpl.java

@@ -88,6 +88,19 @@ public class AgentChatHistoryServiceImpl extends ServiceImpl<AiAgentChatHistoryD
         return ConvertUtils.sourceToTarget(historyList, AgentChatHistoryDTO.class);
     }
 
+    @Override
+    public String getAgentIdBySessionId(String sessionId) {
+        if (sessionId == null || sessionId.isBlank()) {
+            return null;
+        }
+        AgentChatHistoryEntity entity = baseMapper.selectOne(
+                new LambdaQueryWrapper<AgentChatHistoryEntity>()
+                        .select(AgentChatHistoryEntity::getAgentId)
+                        .eq(AgentChatHistoryEntity::getSessionId, sessionId)
+                        .last("LIMIT 1"));
+        return entity == null ? null : entity.getAgentId();
+    }
+
     @Override
     @Transactional(rollbackFor = Exception.class)
     public void deleteByAgentId(String agentId, Boolean deleteAudio, Boolean deleteText) {
@@ -176,6 +189,19 @@ public class AgentChatHistoryServiceImpl extends ServiceImpl<AiAgentChatHistoryD
         return agentChatHistoryEntity == null ? null : agentChatHistoryEntity.getContent();
     }
 
+    @Override
+    public String getAgentIdByAudioId(String audioId) {
+        if (audioId == null || audioId.isBlank()) {
+            return null;
+        }
+        AgentChatHistoryEntity entity = baseMapper.selectOne(
+                new LambdaQueryWrapper<AgentChatHistoryEntity>()
+                        .select(AgentChatHistoryEntity::getAgentId)
+                        .eq(AgentChatHistoryEntity::getAudioId, audioId)
+                        .last("LIMIT 1"));
+        return entity == null ? null : entity.getAgentId();
+    }
+
     @Override
     public boolean isAudioOwnedByAgent(String audioId, String agentId) {
         // 查询是否有指定音频id和智能体id的数据，如果有且只有一条说明此数据属性此智能体

main/manager-api/src/main/java/xiaozhi/modules/agent/service/impl/AgentServiceImpl.java

@@ -33,6 +33,7 @@ import xiaozhi.modules.agent.dao.AgentDao;
 import xiaozhi.modules.agent.dao.AgentTagDao;
 import xiaozhi.modules.agent.dto.AgentCreateDTO;
 import xiaozhi.modules.agent.dto.AgentDTO;
+import xiaozhi.modules.agent.dto.AgentMemoryDTO;
 import xiaozhi.modules.agent.dto.AgentTagDTO;
 import xiaozhi.modules.agent.dto.AgentUpdateDTO;
 import xiaozhi.modules.agent.entity.AgentContextProviderEntity;
@@ -93,6 +94,7 @@ public class AgentServiceImpl extends BaseServiceImpl<AgentDao, AgentEntity> imp
         if (agent == null) {
             throw new RenException(ErrorCode.AGENT_NOT_FOUND);
         }
+        requireCurrentUserPermissionIfPresent(agent);
 
         if (agent.getMemModelId() != null && agent.getMemModelId().equals(Constant.MEMORY_NO_MEM)) {
             agent.setChatHistoryConf(Constant.ChatHistoryConfEnum.IGNORE.getCode());
@@ -116,6 +118,65 @@ public class AgentServiceImpl extends BaseServiceImpl<AgentDao, AgentEntity> imp
         return agent;
     }
 
+    @Override
+    public AgentInfoVO getAgentById(String id, Long userId) {
+        AgentInfoVO agent = getAgentById(id);
+        requireAgentPermission(agent, userId);
+        return agent;
+    }
+
+    private AgentEntity getAgentEntityOrThrow(String agentId) {
+        AgentEntity agent = agentDao.selectById(agentId);
+        if (agent == null) {
+            throw new RenException(ErrorCode.AGENT_NOT_FOUND);
+        }
+        return agent;
+    }
+
+    private boolean isCurrentUserSuperAdmin() {
+        UserDetail user = SecurityUser.getUser();
+        return user != null && Integer.valueOf(SuperAdminEnum.YES.value()).equals(user.getSuperAdmin());
+    }
+
+    private void requireCurrentUserPermissionIfPresent(AgentEntity agent) {
+        Long userId = SecurityUser.getUserId();
+        if (userId != null) {
+            requireAgentPermission(agent, userId);
+        }
+    }
+
+    private boolean hasAgentPermission(AgentEntity agent, Long userId) {
+        if (agent == null) {
+            return false;
+        }
+        if (isCurrentUserSuperAdmin()) {
+            return true;
+        }
+        return userId != null && userId.equals(agent.getUserId());
+    }
+
+    private void requireAgentPermission(AgentEntity agent, Long userId) {
+        if (!hasAgentPermission(agent, userId)) {
+            throw new RenException(ErrorCode.NO_PERMISSION);
+        }
+    }
+
+    private boolean hasDevicePermission(DeviceEntity device, Long userId) {
+        if (device == null) {
+            return false;
+        }
+        if (isCurrentUserSuperAdmin()) {
+            return true;
+        }
+        return userId != null && userId.equals(device.getUserId());
+    }
+
+    private void requireDevicePermission(DeviceEntity device, Long userId) {
+        if (!hasDevicePermission(device, userId)) {
+            throw new RenException(ErrorCode.NO_PERMISSION);
+        }
+    }
+
     @Override
     public boolean insert(AgentEntity entity) {
         // 如果ID为空，自动生成一个UUID作为ID
@@ -275,25 +336,11 @@ public class AgentServiceImpl extends BaseServiceImpl<AgentDao, AgentEntity> imp
     }
 
     @Override
-    public boolean checkAgentPermission(String agentId, Long userId) {
+    public boolean checkAgentPermission(String agentId, Long userId) {
-        if (SecurityUser.getUser() == null || SecurityUser.getUser().getId() == null) {
+        AgentEntity agent = agentDao.selectById(agentId);
-            return false;
+        return hasAgentPermission(agent, userId);
-        }
+    }
-        // 获取智能体信息
+
-        AgentEntity agent = getAgentById(agentId);
-        if (agent == null) {
-            return false;
-        }
-
-        // 如果是超级管理员，直接返回true
-        if (SecurityUser.getUser().getSuperAdmin() == SuperAdminEnum.YES.value()) {
-            return true;
-        }
-
-        // 检查是否是智能体的所有者
-        return userId.equals(agent.getUserId());
-    }
-
     // 根据id更新智能体信息
     @Override
     @Transactional(rollbackFor = Exception.class)
@@ -301,13 +348,29 @@ public class AgentServiceImpl extends BaseServiceImpl<AgentDao, AgentEntity> imp
         updateAgentById(agentId, dto, true);
     }
 
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public void updateAgentById(String agentId, AgentUpdateDTO dto, Long userId) {
+        updateAgentById(agentId, dto, userId, true);
+    }
+
     // 根据id更新智能体信息
     @Override
     @Transactional(rollbackFor = Exception.class)
     public void updateAgentById(String agentId, AgentUpdateDTO dto, boolean createSnapshot) {
-        if (agentDao.selectByIdForUpdate(agentId) == null) {
+        updateAgentById(agentId, dto, null, createSnapshot);
+    }
+
+    private void updateAgentById(String agentId, AgentUpdateDTO dto, Long userId, boolean createSnapshot) {
+        AgentEntity lockedAgent = agentDao.selectByIdForUpdate(agentId);
+        if (lockedAgent == null) {
             throw new RenException(ErrorCode.AGENT_NOT_FOUND);
         }
+        if (userId == null) {
+            requireCurrentUserPermissionIfPresent(lockedAgent);
+        } else {
+            requireAgentPermission(lockedAgent, userId);
+        }
 
         // 锁定后查询现有实体和关联配置
         AgentEntity existingEntity = this.getAgentById(agentId);
@@ -320,9 +383,9 @@ public class AgentServiceImpl extends BaseServiceImpl<AgentDao, AgentEntity> imp
         if (dto.getAgentName() != null) {
             existingEntity.setAgentName(dto.getAgentName());
         }
-        if (dto.getAgentCode() != null) {
+        if (dto.getAgentCode() != null) {
-            existingEntity.setAgentCode(dto.getAgentCode());
+            existingEntity.setAgentCode(dto.getAgentCode());
-        }
+        }
         if (dto.getAsrModelId() != null) {
             existingEntity.setAsrModelId(dto.getAsrModelId());
         }
@@ -479,6 +542,29 @@ public class AgentServiceImpl extends BaseServiceImpl<AgentDao, AgentEntity> imp
         }
     }
 
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public void updateAgentMemoryByDeviceMacAddress(String macAddress, AgentMemoryDTO dto, Long userId) {
+        DeviceEntity device = deviceService.getDeviceByMacAddress(macAddress);
+        if (device == null || StringUtils.isBlank(device.getAgentId()) || dto == null) {
+            return;
+        }
+
+        requireDevicePermission(device, userId);
+
+        AgentUpdateDTO agentUpdateDTO = new AgentUpdateDTO();
+        agentUpdateDTO.setSummaryMemory(dto.getSummaryMemory());
+        updateAgentById(device.getAgentId(), agentUpdateDTO, userId, false);
+    }
+
+    @Override
+    @Transactional(rollbackFor = Exception.class)
+    public void deleteAgentById(String agentId, Long userId) {
+        AgentEntity agent = getAgentEntityOrThrow(agentId);
+        requireAgentPermission(agent, userId);
+        deleteAgent(agentId);
+    }
+
     /**
      * 验证大语言模型和意图识别的参数是否符合匹配
      * 

main/xiaozhi-server/agent-base-prompt.txt

@@ -65,7 +65,7 @@ Calling unnecessarily is also wrong: User asks ”明天天气” → you call g
 
 <speaker_recognition>
 For the input format `{"speaker":"...", "content":"..."}` (speaker = speaker name, content = text):
-1. [Identity known] When `speaker` is a specific name, identity has been recognized. On the first turn you must address them naturally and adjust your response style based on their history.
+1. [Identity known] When `speaker` is a specific name, identity has been recognized — remember who they are. Do not start every reply by addressing them by name; get straight to the content (a natural greeting by name once, at the very start of a conversation, is fine). If they ask about their identity (e.g. "你知道我是谁吗", "还记得我吗"), answer truthfully using their recognized name.
 2. [Identity unknown] When `speaker` is `未知说话人`, the system failed to identify the voice. You must NEVER mention the data inside the `speakers_info` tag to the user. Judge from context whether the speaker is the owner or the owner's friend, and keep the conversation natural.
 </speaker_recognition>
 

main/xiaozhi-server/core/connection.py

@@ -159,6 +159,8 @@ class ConnectionHandler:
         self.asr_audio = []  # 存储PCM帧列表，供VAD和ASR共享
         self.asr_audio_queue = queue.Queue()
         self.current_speaker = None  # 存储当前说话人
+        self.introduced_speakers = set()  # 已"首次引入"的说话人，控制只在首轮带名字
+        self.system_introduced_speakers = set()  # 已在 system 注入过身份的说话人，控制 system 身份只首轮出现
 
         # llm相关变量
         self.dialogue = Dialogue()
@@ -1120,12 +1122,20 @@ class ConnectionHandler:
                 )
                 memory_str = future.result()
 
+            # 仅在该说话人首次出现时把身份注入 system，之后靠对话历史首轮保留，
+            # 避免每轮在 system 重复出现名字诱导模型反复称呼
+            speaker_for_system = None
+            cs = (self.current_speaker or "").strip()
+            if cs and cs != "未知说话人" and cs not in self.system_introduced_speakers:
+                self.system_introduced_speakers.add(cs)
+                speaker_for_system = cs
+
             if self.intent_type == "function_call" and functions is not None:
                 # 使用支持functions的streaming接口
                 llm_responses = self.llm.response_with_functions(
                     self.session_id,
                     self.dialogue.get_llm_dialogue_with_memory(
-                        memory_str, self.config.get("voiceprint", {})
+                        memory_str, self.config.get("voiceprint", {}), speaker_for_system
                     ),
                     functions=functions,
                 )
@@ -1133,7 +1143,7 @@ class ConnectionHandler:
                 llm_responses = self.llm.response(
                     self.session_id,
                     self.dialogue.get_llm_dialogue_with_memory(
-                        memory_str, self.config.get("voiceprint", {})
+                        memory_str, self.config.get("voiceprint", {}), speaker_for_system
                     ),
                 )
         except Exception as e:

main/xiaozhi-server/core/handle/receiveAudioHandle.py

@@ -43,7 +43,6 @@ async def resume_vad_detection(conn: "ConnectionHandler"):
 async def startToChat(conn: "ConnectionHandler", text):
     # 检查输入是否是JSON格式（包含说话人信息）
     speaker_name = None
-    language_tag = None
     actual_text = text
 
     try:
@@ -52,12 +51,16 @@ async def startToChat(conn: "ConnectionHandler", text):
             data = json.loads(text)
             if "speaker" in data and "content" in data:
                 speaker_name = data["speaker"]
-                language_tag = data["language"]
+                actual_content = data["content"]
-                actual_text = data["content"]
                 conn.logger.bind(tag=TAG).info(f"解析到说话人信息: {speaker_name}")
 
-                # 直接使用JSON格式的文本，不解析
+                # 仅在该说话人首次出现时保留 {"speaker":...} JSON，让模型自然称呼一次；
-                actual_text = text
+                # 后续轮降为纯文本，避免每轮重复出现名字诱导模型反复称呼
+                if speaker_name not in conn.introduced_speakers:
+                    conn.introduced_speakers.add(speaker_name)
+                    actual_text = text
+                else:
+                    actual_text = actual_content
     except (json.JSONDecodeError, KeyError):
         # 如果解析失败，继续使用原始文本
         pass

main/xiaozhi-server/core/providers/asr/doubao_stream.py

@@ -428,4 +428,3 @@ class ASRProvider(ASRProviderBase):
                 pass
             self.forward_task = None
         self.is_processing = False
-

main/xiaozhi-server/core/providers/memory/powermem/powermem.py

@@ -43,11 +43,8 @@ class MemoryProvider(MemoryProviderBase):
         self.memory_client = None
         self.enable_user_profile = False
         self.last_profile_content = ""  # Cache for user profile from UserMemory
-
         try:
-            # Check if user profile mode is enabled
+            self.enable_user_profile = str(config.get("enable_user_profile", False)).lower() == 'true'
-            self.enable_user_profile = config.get("enable_user_profile", False)
-            
             # Get configuration parameters
             database_provider = config.get("database_provider", "sqlite")
             llm_provider = config.get("llm_provider", "qwen")

main/xiaozhi-server/core/utils/dialogue.py

@@ -92,7 +92,8 @@ class Dialogue:
         return result
 
     def get_llm_dialogue_with_memory(
-            self, memory_str: str = None, voiceprint_config: dict = None
+            self, memory_str: str = None, voiceprint_config: dict = None,
+            current_speaker: str = None,
     ) -> List[Dict[str, str]]:
         # 构建对话
         dialogue = []
@@ -103,50 +104,31 @@ class Dialogue:
         )
 
         if system_message:
-            # 以 <context> 为分界点，拆分静态 system prompt 和动态上下文
-            # 静态部分（规则、身份等）保持不变，可命中前缀缓存
-            # 动态部分（时间、天气、记忆等）作为第二条 system 消息，保持 system 权威性
             full_prompt = system_message.content
-            context_match = re.search(r"<context>", full_prompt)
-            if context_match:
-                static_part = full_prompt[:context_match.start()]
-                dynamic_part = full_prompt[context_match.start():]
-            else:
-                static_part = full_prompt
-                dynamic_part = ""
 
-            # 第一段：静态 system prompt（前缀缓存可命中）
-            dialogue.append({"role": "system", "content": static_part})
-
-        # 第二段：few-shot 示例（会话内不变，也是缓存前缀的一部分）
-        non_system_messages = [m for m in self.dialogue if m.role != "system"]
-        fewshot_messages = [m for m in non_system_messages if m.is_temporary]
-        complete_fewshot = self._ensure_tool_calls_complete(fewshot_messages)
-        for m in complete_fewshot:
-            self.getMessages(m, dialogue)
-
-        # 第三段：动态上下文 system prompt（时间、记忆、说话人等）
-        # 保持 system 角色以确保模型权威性，不降级为 user
-        if system_message and dynamic_part:
             # 替换时间占位符
-            dynamic_part = dynamic_part.replace(
+            full_prompt = full_prompt.replace(
                 "{{current_time}}", datetime.now().strftime("%H:%M")
             )
 
             # 填充记忆
             if memory_str is not None:
-                dynamic_part = re.sub(
+                full_prompt = re.sub(
                     r"<memory>.*?</memory>",
                     f"<memory>\n{memory_str}\n</memory>",
-                    dynamic_part,
+                    full_prompt,
                     flags=re.DOTALL,
                 )
 
             # 追加说话人信息
             try:
-                speakers = voiceprint_config.get("speakers", [])
+                current_speaker_name = (current_speaker or "").strip()
-                if speakers:
+                # 仅在本轮注入了有效身份时才输出 speakers_info，避免列表里的名字每轮
-                    dynamic_part += "\n<speakers_info>"
+                # 重复出现诱导模型反复称呼；后续轮不再注入身份，靠对话历史首轮保留
+                if current_speaker_name and current_speaker_name != "未知说话人":
+                    speakers = voiceprint_config.get("speakers", [])
+                    speakers_info = "\n<speakers_info>"
+                    speakers_info += f"\n当前说话人：{current_speaker_name}"
                     for speaker_str in speakers:
                         try:
                             parts = speaker_str.split(",", 2)
@@ -155,16 +137,24 @@ class Dialogue:
                                 description = (
                                     parts[2].strip() if len(parts) >= 3 else ""
                                 )
-                                dynamic_part += f"\n- {name}：{description}"
+                                speakers_info += f"\n- {name}：{description}"
                         except:
                             pass
-                    dynamic_part += "\n</speakers_info>"
+                    speakers_info += "\n</speakers_info>"
+                    full_prompt += speakers_info
             except:
                 pass
 
-            dialogue.append({"role": "system", "content": dynamic_part})
+            dialogue.append({"role": "system", "content": full_prompt})
 
-        # 第四段：实际对话历史（不含 few-shot）
+        # 第二段：few-shot 示例（会话内不变）
+        non_system_messages = [m for m in self.dialogue if m.role != "system"]
+        fewshot_messages = [m for m in non_system_messages if m.is_temporary]
+        complete_fewshot = self._ensure_tool_calls_complete(fewshot_messages)
+        for m in complete_fewshot:
+            self.getMessages(m, dialogue)
+
+        # 第三段：实际对话历史（不含 few-shot）
         actual_messages = [m for m in non_system_messages if not m.is_temporary]
         complete_actual = self._ensure_tool_calls_complete(actual_messages)
         for m in complete_actual: