feat(chat): show reply model attribution in assistant message bubbles

mateclaw-server/src/main/java/vip/mate/channel/web/ChatController.java

@@ -1416,6 +1416,14 @@ public class ChatController {
         payload.put("status", status);
         if (savedAssistant != null && savedAssistant.getId() != null) {
             payload.put("assistantMessageId", savedAssistant.getId());
+            // Surface runtime model attribution so the chat bubble can show
+            // which model produced this reply without waiting for a history reload.
+            if (savedAssistant.getRuntimeModel() != null && !savedAssistant.getRuntimeModel().isBlank()) {
+                payload.put("runtimeModel", savedAssistant.getRuntimeModel());
+            }
+            if (savedAssistant.getRuntimeProvider() != null && !savedAssistant.getRuntimeProvider().isBlank()) {
+                payload.put("runtimeProvider", savedAssistant.getRuntimeProvider());
+            }
         }
         if (promptTokens > 0) payload.put("promptTokens", promptTokens);
         if (completionTokens > 0) payload.put("completionTokens", completionTokens);

mateclaw-server/src/main/java/vip/mate/workspace/conversation/vo/MessageVO.java

@@ -41,6 +41,12 @@ public class MessageVO {
     /** Completion tokens 消耗 */
     private Integer completionTokens;
 
+    /** Model name actually used to produce this message (e.g. "deepseek-chat"). */
+    private String runtimeModel;
+
+    /** Provider id of the runtime model (e.g. "deepseek", "zhipu"). */
+    private String runtimeProvider;
+
     private LocalDateTime createTime;
 
     private LocalDateTime updateTime;
@@ -59,6 +65,8 @@ public class MessageVO {
         vo.setMetadata(parseMetadataToObject(entity.getMetadata()));
         vo.setPromptTokens(entity.getPromptTokens());
         vo.setCompletionTokens(entity.getCompletionTokens());
+        vo.setRuntimeModel(entity.getRuntimeModel());
+        vo.setRuntimeProvider(entity.getRuntimeProvider());
         vo.setCreateTime(entity.getCreateTime());
         vo.setUpdateTime(entity.getUpdateTime());
         vo.setContentParts(contentParts);

mateclaw-ui/src/components/chat/MessageBubble.vue

@@ -337,6 +337,12 @@
           >
             <el-icon><RefreshRight /></el-icon>
           </button>
+          <!-- Reply model attribution (assistant only) -->
+          <span
+            v-if="role === 'assistant' && replyModel"
+            class="action-model"
+            :title="replyModelTitle"
+          >{{ replyModel }}</span>
           <!-- 时间戳（inline） -->
           <span class="action-time">{{ formattedTime }}</span>
         </div>
@@ -710,6 +716,16 @@ const formattedTime = computed(() => {
   })
 })
 
+// Reply model attribution: shows which model produced the assistant message.
+// Empty for streaming (server only emits runtimeModel after persistence) and
+// historical messages prior to MessageVO carrying the field.
+const replyModel = computed(() => props.message.runtimeModel || '')
+const replyModelTitle = computed(() => {
+  const provider = props.message.runtimeProvider
+  const base = t('chat.replyModel', { model: replyModel.value })
+  return provider ? `${base} (${provider})` : base
+})
+
 const formatFileSize = (size: number) => {
   if (size < 1024) return `${size} B`
   if (size < 1024 * 1024) return `${(size / 1024).toFixed(1)} KB`
@@ -1475,6 +1491,18 @@ watch(isGenerating, (generating) => {
   user-select: none;
 }
 
+.action-model {
+  font-size: 11px;
+  color: var(--mc-text-secondary, #64748b);
+  margin-left: 4px;
+  padding: 1px 6px;
+  border-radius: 4px;
+  background: var(--mc-fill-2, rgba(100, 116, 139, 0.08));
+  font-family: var(--mc-mono-font, ui-monospace, "SF Mono", Menlo, monospace);
+  user-select: text;
+  white-space: nowrap;
+}
+
 /* ==================== 主内容区域 ==================== */
 .msg-content {
   position: relative;

mateclaw-ui/src/composables/chat/useChat.ts

@@ -487,6 +487,8 @@ export function useChat(options: UseChatOptions): UseChatReturn {
         const msg = messages.value[msgIndex]
         if (data.promptTokens !== undefined) msg.promptTokens = data.promptTokens
         if (data.completionTokens !== undefined) msg.completionTokens = data.completionTokens
+        if (data.runtimeModel) msg.runtimeModel = data.runtimeModel
+        if (data.runtimeProvider) msg.runtimeProvider = data.runtimeProvider
         // Replace the local temp ID with the backend-persisted ID so reconcile can match by ID
         if (data.assistantMessageId) {
           msg.id = data.assistantMessageId

mateclaw-ui/src/i18n/locales/en-US.ts

@@ -148,6 +148,7 @@ export default {
     copy: 'Copy',
     copied: 'Copied',
     regenerate: 'Regenerate',
+    replyModel: 'Reply model: {model}',
     ttsPlay: 'Read Aloud',
     ttsStop: 'Stop Reading',
     conversations: 'Conversations',

mateclaw-ui/src/i18n/locales/zh-CN.ts

@@ -148,6 +148,7 @@ export default {
     copy: '复制',
     copied: '已复制',
     regenerate: '重新生成',
+    replyModel: '本条回复模型: {model}',
     ttsPlay: '朗读',
     ttsStop: '停止朗读',
     conversations: '会话列表',

mateclaw-ui/src/types/index.ts

@@ -83,6 +83,9 @@ export interface Message {
   // Token 统计
   promptTokens?: number
   completionTokens?: number
+  // Runtime model attribution (assistant messages): the model that actually produced this reply
+  runtimeModel?: string
+  runtimeProvider?: string
   // 前端临时字段
   streaming?: boolean  // 内部动画控制，UI 渲染以 status 为准
   attachments?: ChatAttachment[]

mateclaw-ui/src/views/ChatConsole.vue

@@ -1774,6 +1774,8 @@ function normalizeMessage(raw: Message): Message {
   // 保留后端返回的 token 字段（MessageVO 新增）
   if ((raw as any).promptTokens) msg.promptTokens = (raw as any).promptTokens
   if ((raw as any).completionTokens) msg.completionTokens = (raw as any).completionTokens
+  if ((raw as any).runtimeModel) msg.runtimeModel = (raw as any).runtimeModel
+  if ((raw as any).runtimeProvider) msg.runtimeProvider = (raw as any).runtimeProvider
 
   if (msg.contentParts.length === 0 && msg.content) {
     if (msg.role === 'assistant') {