Merge branch 'xinnan-tech:main' into custom_tts_api

docs/firmware-setting.md

@@ -1,7 +1,7 @@
 # 基于虾哥编译好的固件配置自定义服务器
 
 ## 第1步 确认版本
-烧录虾哥已经编译好的1.6.1版本固件
+烧录虾哥已经编译好的[1.6.1版本以上固件](https://github.com/78/xiaozhi-esp32/releases)
 
 ## 第2步 准备你的ota地址
 如果你按照教程使用的是全模块部署，就应该会有ota地址。
@@ -29,6 +29,7 @@ wss://2662r3426b.vicp.fun/xiaozhi/v1/
 
 ## 第3步 进入配网模式
 进入机器的配网模式，在页面顶部，点击“高级选项”，在里面输入你服务器的`ota`地址，点击保存。重启设备
+![请参考-OTA地址设置](../docs/images/firmware-setting-ota.png)
 
 ## 第4步 唤醒小智，查看日志输出
 

main/manager-api/src/main/java/xiaozhi/common/constant/Constant.java

@@ -200,4 +200,9 @@ public interface Constant {
      * 版本号
      */
     public static final String VERSION = "0.4.3";
+
+    /**
+     * 无效固件URL
+     */
+    String INVALID_FIRMWARE_URL = "http://xiaozhi.server.com:8002/xiaozhi/otaMag/download/NOT_ACTIVATED_FIRMWARE_THIS_IS_A_INVALID_URL";
 }

main/manager-api/src/main/java/xiaozhi/modules/device/service/impl/DeviceServiceImpl.java

@@ -142,7 +142,7 @@ public class DeviceServiceImpl extends BaseServiceImpl<DeviceDao, DeviceEntity>
         if (deviceById == null) {
             DeviceReportRespDTO.Firmware firmware = new DeviceReportRespDTO.Firmware();
             firmware.setVersion(deviceReport.getApplication().getVersion());
-            firmware.setUrl("http://xiaozhi.server.com:8002/xiaozhi/otaMag/download/NOT_ACTIVATED_FIRMWARE_THIS_IS_A_INVALID_URL");
+            firmware.setUrl(Constant.INVALID_FIRMWARE_URL);
             response.setFirmware(firmware);
         } else {
             // 只有在设备已绑定且autoUpdate不为0的情况下才返回固件升级信息
@@ -378,7 +378,7 @@ public class DeviceServiceImpl extends BaseServiceImpl<DeviceDao, DeviceEntity>
         }
 
         firmware.setVersion(ota == null ? currentVersion : ota.getVersion());
-        firmware.setUrl(downloadUrl == null ? "" : downloadUrl);
+        firmware.setUrl(downloadUrl == null ? Constant.INVALID_FIRMWARE_URL : downloadUrl);
         return firmware;
     }
 

main/manager-web/src/components/HeaderBar.vue

@@ -58,7 +58,7 @@
 
       <!-- 右侧元素 -->
       <div class="header-right">
-        <div class="search-container" v-if="$route.path === '/home'">
+        <div class="search-container" v-if="$route.path === '/home' && !(isSuperAdmin && isSmallScreen)">
           <el-input v-model="search" placeholder="输入名称搜索.." class="custom-search-input"
             @keyup.enter.native="handleSearch">
             <i slot="suffix" class="el-icon-search search-icon" @click="handleSearch"></i>
@@ -103,7 +103,8 @@ export default {
       },
       isChangePasswordDialogVisible: false, // 控制修改密码弹窗的显示
       userDropdownVisible: false,
-      paramDropdownVisible: false
+      paramDropdownVisible: false,
+      isSmallScreen: false
     }
   },
   computed: {
@@ -113,7 +114,13 @@ export default {
     }
   },
   mounted() {
-    this.fetchUserInfo()
+    this.fetchUserInfo();
+    this.checkScreenSize();
+    window.addEventListener('resize', this.checkScreenSize);
+  },
+  //移除事件监听器
+  beforeDestroy() {
+    window.removeEventListener('resize', this.checkScreenSize);
   },
   methods: {
     goHome() {
@@ -147,7 +154,9 @@ export default {
         }
       })
     },
-
+    checkScreenSize() {
+      this.isSmallScreen = window.innerWidth <= 1386;
+    },
     // 处理搜索
     handleSearch() {
       const searchValue = this.search.trim();
@@ -354,45 +363,6 @@ export default {
   }
 }
 
-@media (max-width: 1024px) {
-  .search-container {
-    margin-right: 10px;
-    max-width: 150px;
-  }
-
-  .header-right {
-    gap: 5px;
-  }
-}
-
-@media (max-width: 900px) {
-  .header-left {
-    margin-right: auto;
-  }
-
-  .search-container {
-    max-width: 150px;
-  }
-}
-
-@media (max-width: 768px) {
-  .search-container {
-    max-width: 145px;
-  }
-
-  .custom-search-input>>>.el-input__inner {
-    padding-left: 10px;
-    font-size: 11px;
-  }
-}
-
-@media (max-width: 600px) {
-  .search-container {
-    max-width: 120px;
-    min-width: 100px;
-  }
-}
-
 .equipment-management.more-dropdown {
   position: relative;
 }
@@ -411,13 +381,4 @@ export default {
   color: #606266;
   white-space: nowrap;
 }
-
-@media (max-width: 768px) {
-  .equipment-management.more-dropdown .el-dropdown-menu {
-    position: fixed;
-    right: 10px;
-    top: 60px;
-    z-index: 2000;
-  }
-}
 </style>

main/xiaozhi-server/core/connection.py

@@ -231,7 +231,9 @@ class ConnectionHandler:
                         # 创建新事件循环（避免与主循环冲突）
                         loop = asyncio.new_event_loop()
                         asyncio.set_event_loop(loop)
-                        loop.run_until_complete(self.memory.save_memory(self.dialogue.dialogue))
+                        loop.run_until_complete(
+                            self.memory.save_memory(self.dialogue.dialogue)
+                        )
                     except Exception as e:
                         self.logger.bind(tag=TAG).error(f"保存记忆失败: {e}")
                     finally:
@@ -441,10 +443,10 @@ class ConnectionHandler:
     def _initialize_memory(self):
         """初始化记忆模块"""
         self.memory.init_memory(
-            self.device_id,
+            role_id=self.device_id,
-            self.llm,
+            llm=self.llm,
-            self.config["summaryMemory"],
+            summary_memory=self.config.get("summaryMemory", None),
-            not self.read_config_from_api,
+            save_to_file=not self.read_config_from_api,
         )
 
     def _initialize_intent(self):

main/xiaozhi-server/core/providers/intent/intent_llm/intent_llm.py

@@ -215,7 +215,7 @@ class IntentProvider(IntentProviderBase):
 
                 # 记录识别到的function call
                 logger.bind(tag=TAG).info(
-                    f"识别到function call: {function_name}, 参数: {function_args}"
+                    f"llm 识别到意图: {function_name}, 参数: {function_args}"
                 )
 
                 # 添加到缓存

main/xiaozhi-server/core/providers/llm/ollama/ollama.py

@@ -21,27 +21,67 @@ class LLMProvider(LLMProviderBase):
             api_key="ollama"  # Ollama doesn't need an API key but OpenAI client requires one
         )
 
+        # 检查是否是qwen3模型
+        self.is_qwen3 = self.model_name and self.model_name.lower().startswith("qwen3")
+
     def response(self, session_id, dialogue):
         try:
+            # 如果是qwen3模型，在用户最后一条消息中添加/no_think指令
+            if self.is_qwen3:
+                # 复制对话列表，避免修改原始对话
+                dialogue_copy = dialogue.copy()
+
+                # 找到最后一条用户消息
+                for i in range(len(dialogue_copy) - 1, -1, -1):
+                    if dialogue_copy[i]["role"] == "user":
+                        # 在用户消息前添加/no_think指令
+                        dialogue_copy[i]["content"] = "/no_think " + dialogue_copy[i]["content"]
+                        logger.bind(tag=TAG).debug(f"为qwen3模型添加/no_think指令")
+                        break
+
+                # 使用修改后的对话
+                dialogue = dialogue_copy
+
             responses = self.client.chat.completions.create(
                 model=self.model_name,
                 messages=dialogue,
                 stream=True
             )
-            is_active=True
+            is_active = True
+            # 用于处理跨chunk的标签
+            buffer = ""
+
             for chunk in responses:
                 try:
                     delta = chunk.choices[0].delta if getattr(chunk, 'choices', None) else None
                     content = delta.content if hasattr(delta, 'content') else ''
+
                     if content:
-                        if '<think>' in content:
+                        # 将内容添加到缓冲区
+                        buffer += content
+
+                        # 处理缓冲区中的标签
+                        while '<think>' in buffer and '</think>' in buffer:
+                            # 找到完整的<think></think>标签并移除
+                            pre = buffer.split('<think>', 1)[0]
+                            post = buffer.split('</think>', 1)[1]
+                            buffer = pre + post
+
+                        # 处理只有开始标签的情况
+                        if '<think>' in buffer:
                             is_active = False
-                            content = content.split('<think>')[0]
+                            buffer = buffer.split('<think>', 1)[0]
-                        if '</think>' in content:
+
+                        # 处理只有结束标签的情况
+                        if '</think>' in buffer:
                             is_active = True
-                            content = content.split('</think>')[-1]
+                            buffer = buffer.split('</think>', 1)[1]
-                        if is_active:
+
-                            yield content
+                        # 如果当前处于活动状态且缓冲区有内容，则输出
+                        if is_active and buffer:
+                            yield buffer
+                            buffer = ""  # 清空缓冲区
+
                 except Exception as e:
                     logger.bind(tag=TAG).error(f"Error processing chunk: {e}")
 
@@ -51,6 +91,22 @@ class LLMProvider(LLMProviderBase):
 
     def response_with_functions(self, session_id, dialogue, functions=None):
         try:
+            # 如果是qwen3模型，在用户最后一条消息中添加/no_think指令
+            if self.is_qwen3:
+                # 复制对话列表，避免修改原始对话
+                dialogue_copy = dialogue.copy()
+
+                # 找到最后一条用户消息
+                for i in range(len(dialogue_copy) - 1, -1, -1):
+                    if dialogue_copy[i]["role"] == "user":
+                        # 在用户消息前添加/no_think指令
+                        dialogue_copy[i]["content"] = "/no_think " + dialogue_copy[i]["content"]
+                        logger.bind(tag=TAG).debug(f"为qwen3模型添加/no_think指令")
+                        break
+
+                # 使用修改后的对话
+                dialogue = dialogue_copy
+
             stream = self.client.chat.completions.create(
                 model=self.model_name,
                 messages=dialogue,
@@ -58,8 +114,49 @@ class LLMProvider(LLMProviderBase):
                 tools=functions,
             )
 
+            is_active = True
+            buffer = ""
+
             for chunk in stream:
-                yield chunk.choices[0].delta.content, chunk.choices[0].delta.tool_calls
+                try:
+                    delta = chunk.choices[0].delta if getattr(chunk, 'choices', None) else None
+                    content = delta.content if hasattr(delta, 'content') else None
+                    tool_calls = delta.tool_calls if hasattr(delta, 'tool_calls') else None
+
+                    # 如果是工具调用，直接传递
+                    if tool_calls:
+                        yield None, tool_calls
+                        continue
+
+                    # 处理文本内容
+                    if content:
+                        # 将内容添加到缓冲区
+                        buffer += content
+
+                        # 处理缓冲区中的标签
+                        while '<think>' in buffer and '</think>' in buffer:
+                            # 找到完整的<think></think>标签并移除
+                            pre = buffer.split('<think>', 1)[0]
+                            post = buffer.split('</think>', 1)[1]
+                            buffer = pre + post
+
+                        # 处理只有开始标签的情况
+                        if '<think>' in buffer:
+                            is_active = False
+                            buffer = buffer.split('<think>', 1)[0]
+
+                        # 处理只有结束标签的情况
+                        if '</think>' in buffer:
+                            is_active = True
+                            buffer = buffer.split('</think>', 1)[1]
+
+                        # 如果当前处于活动状态且缓冲区有内容，则输出
+                        if is_active and buffer:
+                            yield buffer, None
+                            buffer = ""  # 清空缓冲区
+                except Exception as e:
+                    logger.bind(tag=TAG).error(f"Error processing function chunk: {e}")
+                    continue
 
         except Exception as e:
             logger.bind(tag=TAG).error(f"Error in Ollama function call: {e}")

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

@@ -21,6 +21,6 @@ class MemoryProviderBase(ABC):
         """Query memories for specific role based on similarity"""
         return "please implement query method"
 
-    def init_memory(self, role_id, llm, summary_memory=None):
+    def init_memory(self, role_id, llm, **kwargs):
         self.role_id = role_id
         self.llm = llm

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

@@ -112,8 +112,10 @@ class MemoryProvider(MemoryProviderBase):
         self.memory_path = get_project_dir() + "data/.memory.yaml"
         self.load_memory(summary_memory)
 
-    def init_memory(self, role_id, llm, summary_memory=None, save_to_file=True):
+    def init_memory(
-        super().init_memory(role_id, llm)
+        self, role_id, llm, summary_memory=None, save_to_file=True, **kwargs
+    ):
+        super().init_memory(role_id, llm, **kwargs)
         self.save_to_file = save_to_file
         self.load_memory(summary_memory)
 

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

@@ -319,7 +319,7 @@ def initialize_modules(
         modules["memory"] = memory.create_instance(
             memory_type,
             config["Memory"][select_memory_module],
-            config['summaryMemory'],
+            config.get('summaryMemory', None),
         )
         logger.bind(tag=TAG).info(f"初始化组件: memory成功 {select_memory_module}")
 

main/xiaozhi-server/docker-compose_all.yml

@@ -76,6 +76,7 @@ services:
     expose:
       - 6379
     container_name: xiaozhi-esp32-server-redis
+    restart: always
     healthcheck:
       test: ["CMD", "redis-cli", "ping"]
       interval: 10s

main/xiaozhi-server/requirements.txt

@@ -22,11 +22,12 @@ mem0ai==0.1.62
 bs4==0.0.2
 modelscope==1.23.2
 sherpa_onnx==1.11.0
-mcp==1.7.1
+mcp==1.8.1
 cnlunar==0.2.0
 PySocks==1.7.1
 dashscope==1.23.1
 baidu-aip==4.16.13
 chardet==5.2.0
 aioconsole==0.8.1
-markitdown==0.1.1
+markitdown==0.1.1
+mcp-proxy==0.6.0