feat: attachment_support

frontend/src/api/agent/index.ts

@@ -39,6 +39,8 @@ export interface CustomAgentConfig {
   image_upload_enabled?: boolean;    // 是否启用图片上传（默认: false）
   vlm_model_id?: string;            // VLM模型ID（图片分析用）
   image_storage_provider?: string;   // 图片存储提供商
+  audio_upload_enabled?: boolean;    // 是否启用音频上传/ASR转录（默认: false）
+  asr_model_id?: string;            // ASR模型ID（音频转录用）
 
   // ===== 文件类型限制 =====
   // 支持的文件类型（如 ["csv", "xlsx", "xls"]）

frontend/src/api/chat/streame.ts

@@ -29,7 +29,7 @@ export function useStream() {
   let renderTimer: number | null = null
 
   // 启动流式请求
-  const startStream = async (params: { session_id: any; query: any; knowledge_base_ids?: string[]; knowledge_ids?: string[]; agent_enabled?: boolean; agent_id?: string; web_search_enabled?: boolean; enable_memory?: boolean; summary_model_id?: string; mcp_service_ids?: string[]; mentioned_items?: Array<{id: string; name: string; type: string; kb_type?: string}>; images?: Array<{data: string}>; method: string; url: string }) => {
+  const startStream = async (params: { session_id: any; query: any; knowledge_base_ids?: string[]; knowledge_ids?: string[]; agent_enabled?: boolean; agent_id?: string; web_search_enabled?: boolean; enable_memory?: boolean; summary_model_id?: string; mcp_service_ids?: string[]; mentioned_items?: Array<{id: string; name: string; type: string; kb_type?: string}>; images?: Array<{data: string}>; attachment_uploads?: Array<{data: string; file_name: string; file_size: number}>; method: string; url: string }) => {
     // 重置状态
     output.value = '';
     error.value = null;
@@ -118,6 +118,10 @@ export function useStream() {
       if (params.images !== undefined && params.images.length > 0) {
         postBody.images = params.images;
       }
+      // Include attachment_uploads if provided (documents, audio, etc.)
+      if (params.attachment_uploads !== undefined && params.attachment_uploads.length > 0) {
+        postBody.attachment_uploads = params.attachment_uploads;
+      }
       postBody.channel = "web";
       
       await fetchEventSource(url, {

frontend/src/components/AttachmentUpload.vue

@@ -0,0 +1,228 @@
+<script setup lang="ts">
+import { ref, computed } from 'vue';
+import { MessagePlugin } from 'tdesign-vue-next';
+import { useI18n } from 'vue-i18n';
+
+const { t } = useI18n();
+
+export interface AttachmentFile {
+  file: File;
+  id: string;
+  name: string;
+  size: number;
+  type: string;
+  preview?: string;
+}
+
+const props = defineProps<{
+  maxFiles?: number;
+  maxSize?: number; // in MB
+  disabled?: boolean;
+}>();
+
+const emit = defineEmits<{
+  (e: 'update:files', files: AttachmentFile[]): void;
+  (e: 'remove', id: string): void;
+}>();
+
+const attachments = ref<AttachmentFile[]>([]);
+const fileInputRef = ref<HTMLInputElement>();
+
+// Supported file types (matching backend)
+const SUPPORTED_TYPES = [
+  // Documents
+  '.pdf', '.doc', '.docx', '.xls', '.xlsx', '.ppt', '.pptx',
+  // Text
+  '.txt', '.md', '.csv', '.json', '.xml', '.html',
+  // Audio
+  '.mp3', '.wav', '.m4a', '.flac', '.ogg', '.aac',
+];
+
+const maxFiles = computed(() => props.maxFiles || 5);
+const maxSize = computed(() => (props.maxSize || 20) * 1024 * 1024); // Convert MB to bytes
+
+const triggerFileSelect = () => {
+  if (props.disabled) return;
+  fileInputRef.value?.click();
+};
+
+const handleFileSelect = async (event: Event) => {
+  const input = event.target as HTMLInputElement;
+  if (!input.files) return;
+  
+  await addFiles(Array.from(input.files));
+  input.value = ''; // Reset input
+};
+
+const addFiles = async (files: File[]) => {
+  if (props.disabled) return;
+  
+  for (const file of files) {
+    // Check max files limit
+    if (attachments.value.length >= maxFiles.value) {
+      MessagePlugin.warning(t('chat.attachmentTooMany', { max: maxFiles.value }));
+      break;
+    }
+    
+    // Check file size
+    if (file.size > maxSize.value) {
+      MessagePlugin.warning(t('chat.attachmentTooLarge', { name: file.name, max: props.maxSize || 20 }));
+      continue;
+    }
+    
+    // Check file type
+    const ext = '.' + file.name.split('.').pop()?.toLowerCase();
+    if (!SUPPORTED_TYPES.includes(ext)) {
+      MessagePlugin.warning(t('chat.attachmentTypeNotSupported', { name: file.name }));
+      continue;
+    }
+    
+    const attachment: AttachmentFile = {
+      file,
+      id: `${Date.now()}-${Math.random()}`,
+      name: file.name,
+      size: file.size,
+      type: file.type || ext,
+    };
+    
+    attachments.value.push(attachment);
+  }
+  
+  emit('update:files', attachments.value);
+};
+
+const removeAttachment = (id: string) => {
+  const index = attachments.value.findIndex(a => a.id === id);
+  if (index !== -1) {
+    attachments.value.splice(index, 1);
+    emit('update:files', attachments.value);
+    emit('remove', id);
+  }
+};
+
+const formatFileSize = (bytes: number): string => {
+  if (bytes < 1024) return bytes + ' B';
+  if (bytes < 1024 * 1024) return (bytes / 1024).toFixed(1) + ' KB';
+  return (bytes / (1024 * 1024)).toFixed(1) + ' MB';
+};
+
+const getFileIcon = (fileName: string): string => {
+  const ext = fileName.split('.').pop()?.toLowerCase();
+  if (['pdf'].includes(ext || '')) return 'file-pdf';
+  if (['doc', 'docx'].includes(ext || '')) return 'file-word';
+  if (['xls', 'xlsx'].includes(ext || '')) return 'file-excel';
+  if (['ppt', 'pptx'].includes(ext || '')) return 'file-powerpoint';
+  if (['txt', 'md'].includes(ext || '')) return 'file-text';
+  if (['mp3', 'wav', 'm4a', 'flac', 'ogg', 'aac'].includes(ext || '')) return 'sound';
+  return 'file';
+};
+
+defineExpose({
+  attachments,
+  triggerFileSelect,
+  clear: () => {
+    attachments.value = [];
+    emit('update:files', []);
+  }
+});
+</script>
+
+<template>
+  <div class="attachment-upload">
+    <!-- Hidden file input -->
+    <input
+      ref="fileInputRef"
+      type="file"
+      :accept="SUPPORTED_TYPES.join(',')"
+      multiple
+      style="display: none"
+      @change="handleFileSelect"
+    />
+    
+    <!-- Attachment list -->
+    <div v-if="attachments.length > 0" class="attachment-preview-bar">
+      <div
+        v-for="attachment in attachments"
+        :key="attachment.id"
+        class="attachment-preview-item"
+      >
+        <div class="attachment-preview-icon">
+          <t-icon :name="getFileIcon(attachment.name)" />
+        </div>
+        <div class="attachment-preview-name">{{ attachment.name }}</div>
+        <span class="attachment-preview-remove" @click="removeAttachment(attachment.id)" :aria-label="$t('common.remove')">×</span>
+      </div>
+    </div>
+    
+    <!-- Upload button (shown in control bar) -->
+    <slot name="trigger" :trigger="triggerFileSelect" :count="attachments.length" />
+  </div>
+</template>
+
+<style scoped lang="less">
+.attachment-upload {
+  width: 100%;
+}
+
+.attachment-preview-bar {
+  display: flex;
+  gap: 8px;
+  padding: 8px 12px 4px;
+  flex-wrap: wrap;
+}
+
+.attachment-preview-item {
+  position: relative;
+  width: 60px;
+  height: 60px;
+  border-radius: 8px;
+  border: 1px solid var(--td-border-level-1-color, #e7e7e7);
+  background: var(--td-bg-color-secondarycontainer, #f5f5f5);
+  display: flex;
+  flex-direction: column;
+  align-items: center;
+  justify-content: center;
+  gap: 4px;
+  overflow: hidden;
+  cursor: default;
+
+  .attachment-preview-icon {
+    font-size: 22px;
+    color: var(--td-brand-color, #07C05F);
+    line-height: 1;
+  }
+
+  .attachment-preview-name {
+    font-size: 10px;
+    color: var(--td-text-color-secondary, #666);
+    text-align: center;
+    width: 100%;
+    padding: 0 4px;
+    overflow: hidden;
+    text-overflow: ellipsis;
+    white-space: nowrap;
+    box-sizing: border-box;
+  }
+
+  .attachment-preview-remove {
+    position: absolute;
+    top: 2px;
+    right: 2px;
+    width: 16px;
+    height: 16px;
+    background: rgba(0, 0, 0, 0.5);
+    color: #fff;
+    border-radius: 50%;
+    display: flex;
+    align-items: center;
+    justify-content: center;
+    font-size: 12px;
+    cursor: pointer;
+    line-height: 1;
+
+    &:hover {
+      background: rgba(0, 0, 0, 0.7);
+    }
+  }
+}
+</style>

frontend/src/components/Input-field.vue

@@ -18,6 +18,7 @@ import { listAgents, type CustomAgent, BUILTIN_QUICK_ANSWER_ID, BUILTIN_SMART_RE
 import { getTenantWebSearchConfig } from '@/api/web-search';
 import { getConversationConfig, updateConversationConfig, type ConversationConfig } from '@/api/system';
 import { useI18n } from 'vue-i18n';
+import AttachmentUpload, { type AttachmentFile } from './AttachmentUpload.vue';
 
 const route = useRoute();
 const router = useRouter();
@@ -35,6 +36,10 @@ const uploadedImages = ref<Array<{ file: File; preview: string }>>([]);
 const imageInputRef = ref<HTMLInputElement>();
 const imageUploading = ref(false);
 
+// Attachment upload state
+const attachmentUploadRef = ref<InstanceType<typeof AttachmentUpload>>();
+const uploadedAttachments = ref<AttachmentFile[]>([]);
+
 const handleImageSelect = (event: Event) => {
   const input = event.target as HTMLInputElement;
   if (!input.files) return;
@@ -1338,7 +1343,10 @@ watch([selectedKbIds, selectedFileIds], ([kbIds, fileIds]) => {
   }
 }, { deep: true });
 
-const emit = defineEmits(['send-msg', 'stop-generation']);
+const emit = defineEmits<{
+  (e: 'send-msg', query: string, modelId: string, mentionedItems: any[], imageFiles: File[], attachmentFiles: AttachmentFile[]): void;
+  (e: 'stop-generation'): void;
+}>();
 
 const createSession = async (val: string) => {
   if (!val.trim()) {
@@ -1375,15 +1383,23 @@ const createSession = async (val: string) => {
     kb_type: item.type === 'kb' ? (item.kbType || 'document') : undefined
   }));
   const imageFiles = uploadedImages.value.map(img => img.file);
+  const attachmentFiles = uploadedAttachments.value;
+  
   // Blur the textarea BEFORE emitting, so that when the parent navigates away
   // and Vue unmounts this component, TDesign's blur handler won't fire on a
   // detached DOM element (which causes getComputedStyle to throw).
   const textarea = getTextareaEl();
   if (textarea) textarea.blur();
-  emit('send-msg', val, selectedModelId.value, mentionedItems, imageFiles);
+  emit('send-msg', val, selectedModelId.value, mentionedItems, imageFiles, attachmentFiles);
+  
   // Clean up image previews
   uploadedImages.value.forEach(img => URL.revokeObjectURL(img.preview));
   uploadedImages.value = [];
+  
+  // Clean up attachments
+  attachmentUploadRef.value?.clear();
+  uploadedAttachments.value = [];
+  
   clearvalue();
 }
 
@@ -1854,6 +1870,15 @@ defineExpose({
           <span class="image-preview-remove" @click="removeImage(idx)">×</span>
         </div>
       </div>
+      
+      <!-- 附件列表区域 (由 AttachmentUpload 组件渲染) -->
+      <AttachmentUpload
+        ref="attachmentUploadRef"
+        :max-files="5"
+        :max-size="20"
+        @update:files="uploadedAttachments = $event"
+      />
+      
         <!-- 选中的知识库和文件标签（显示在输入框内顶部） -->
       <div v-if="allSelectedItems.length > 0" class="selected-tags-inline">
         <span 
@@ -2013,6 +2038,24 @@ defineExpose({
           </div>
         </t-tooltip>
 
+        <!-- 附件上传按钮 -->
+        <t-tooltip placement="top" theme="light" :popupProps="{ overlayClassName: 'input-field-tooltip' }">
+          <template #content>
+            <span>{{ uploadedAttachments.length > 0 ? $t('chat.attachmentWithCount', { count: uploadedAttachments.length }) : $t('chat.attachmentUploadTooltip') }}</span>
+          </template>
+          <div
+            class="control-btn attachment-upload-btn"
+            :class="{ 'active': uploadedAttachments.length > 0 }"
+            @click.stop="attachmentUploadRef?.triggerFileSelect()"
+          >
+            <!-- 回形针图标 -->
+            <svg width="18" height="18" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round" class="control-icon">
+              <path d="M21.44 11.05l-9.19 9.19a6 6 0 0 1-8.49-8.49l9.19-9.19a4 4 0 0 1 5.66 5.66l-9.2 9.19a2 2 0 0 1-2.83-2.83l8.49-8.48"/>
+            </svg>
+            <span v-if="uploadedAttachments.length > 0" class="attachment-count">{{ uploadedAttachments.length }}</span>
+          </div>
+        </t-tooltip>
+
         <!-- @ 知识库/文件选择按钮 -->
         <t-tooltip placement="top" theme="light" :popupProps="{ overlayClassName: 'input-field-tooltip' }">
           <template #content>
@@ -2564,6 +2607,45 @@ const getImgSrc = (url: string) => {
   }
 }
 
+/* Attachment upload */
+.attachment-upload-btn {
+  width: 28px;
+  height: 28px;
+  padding: 0;
+  min-width: auto;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  position: relative;
+  color: var(--td-text-color-secondary, #666);
+
+  &:hover {
+    background: var(--td-bg-color-secondarycontainer-hover, #f0f0f0);
+    color: var(--td-text-color-primary, #333);
+  }
+
+  &.active {
+    background: rgba(16, 185, 129, 0.1);
+    color: #07C05F;
+  }
+
+  .attachment-count {
+    position: absolute;
+    top: -2px;
+    right: -2px;
+    background: #07C05F;
+    color: #fff;
+    font-size: 10px;
+    width: 14px;
+    height: 14px;
+    border-radius: 50%;
+    display: flex;
+    align-items: center;
+    justify-content: center;
+    line-height: 1;
+  }
+}
+
 .image-preview-bar {
   display: flex;
   gap: 8px;

frontend/src/i18n/locales/en-US.ts

@@ -1864,6 +1864,11 @@ export default {
     imageTooMany: 'Maximum 5 images allowed',
     imageTypeSizeError: 'Only JPG/PNG/GIF/WEBP under 10MB supported',
     imageUploadTooltip: 'Upload image (paste/drop supported)',
+    attachmentUploadTooltip: 'Upload attachment (documents, audio, etc.)',
+    attachmentWithCount: '{count} attachment(s) uploaded',
+    attachmentTooMany: 'Maximum {max} attachments allowed',
+    attachmentTooLarge: 'File {name} exceeds {max}MB limit',
+    attachmentTypeNotSupported: 'Unsupported file type: {name}',
   },
   tenant: {
     title: 'Tenant Information',
@@ -3342,6 +3347,13 @@ export default {
       notConfigured: 'Not Configured',
       goStorageSettings: 'Go to Storage Settings',
     },
+    audioUpload: {
+      label: 'Audio Upload',
+      desc: 'When enabled, users can upload audio files in conversations. The system will automatically transcribe them using the ASR model.',
+      asrModel: 'ASR Model',
+      asrModelDesc: 'Speech recognition model for audio transcription. If not configured, audio files will be passed as placeholders.',
+      asrModelPlaceholder: 'Select ASR Model',
+    },
     faq: {
       title: 'FAQ Priority Strategy',
       tooltip: 'When the knowledge base contains FAQ (Q&A pairs), enable this strategy to prioritize FAQ answers over regular documents',

frontend/src/i18n/locales/ko-KR.ts

@@ -413,6 +413,11 @@ export default {
     imageTooMany: "최대 5장까지 업로드 가능합니다",
     imageTypeSizeError: "JPG/PNG/GIF/WEBP 형식만 지원하며, 단일 파일 10MB 이하",
     imageUploadTooltip: "이미지 업로드 (붙여넣기/드래그 지원)",
+    attachmentUploadTooltip: "첨부 파일 업로드 (문서, 오디오 등)",
+    attachmentWithCount: "{count}개 파일 업로드됨",
+    attachmentTooMany: "최대 {max}개 파일까지 업로드 가능합니다",
+    attachmentTooLarge: "파일 {name}이(가) {max}MB 제한을 초과합니다",
+    attachmentTypeNotSupported: "지원하지 않는 파일 형식: {name}",
   },
   settings: {
     title: "설정",
@@ -3393,6 +3398,13 @@ export default {
       notConfigured: '미설정',
       goStorageSettings: '저장소 설정으로 이동',
     },
+    audioUpload: {
+      label: '음성 업로드',
+      desc: '활성화하면 사용자가 대화에서 오디오 파일을 업로드할 수 있으며, ASR 모델로 자동 변환됩니다',
+      asrModel: 'ASR 모델',
+      asrModelDesc: '음성 인식 모델입니다. 설정하지 않으면 오디오 파일이 플레이스홀더로 전달됩니다',
+      asrModelPlaceholder: 'ASR 모델 선택',
+    },
     faq: {
       title: 'FAQ 우선 전략',
       tooltip: '지식베이스에 FAQ(질문-답변 쌍)가 포함된 경우, 이 전략을 활성화하면 FAQ 답변이 일반 문서보다 우선됩니다',

frontend/src/i18n/locales/ru-RU.ts

@@ -2032,6 +2032,11 @@ export default {
     imageTooMany: 'Максимум 5 изображений',
     imageTypeSizeError: 'Поддерживаются только JPG/PNG/GIF/WEBP до 10 МБ',
     imageUploadTooltip: 'Загрузить изображение (вставка/перетаскивание)',
+    attachmentUploadTooltip: 'Загрузить вложение (документы, аудио и т.д.)',
+    attachmentWithCount: 'Загружено файлов: {count}',
+    attachmentTooMany: 'Максимум {max} файлов',
+    attachmentTooLarge: 'Файл {name} превышает лимит {max} МБ',
+    attachmentTypeNotSupported: 'Неподдерживаемый тип файла: {name}',
     thinkingAlt: 'Обдумывание...',
     deepThoughtCompleted: 'Глубокий анализ завершён',
     deepThoughtAlt: 'Глубокий анализ'
@@ -3009,6 +3014,13 @@ export default {
       notConfigured: 'Не настроено',
       goStorageSettings: 'Перейти к настройкам хранилища'
     },
+    audioUpload: {
+      label: 'Загрузка аудио',
+      desc: 'Позволяет пользователям загружать аудиофайлы в чате. Система автоматически транскрибирует их с помощью ASR-модели.',
+      asrModel: 'ASR-модель',
+      asrModelDesc: 'Модель распознавания речи. Если не настроена, аудиофайлы передаются как заглушки.',
+      asrModelPlaceholder: 'Выберите ASR-модель',
+    },
     faq: {
       title: 'Стратегия приоритета FAQ',
       tooltip: 'Если база знаний содержит FAQ (пары вопрос-ответ), включите эту стратегию для приоритета ответов FAQ над обычными документами',

frontend/src/i18n/locales/zh-CN.ts

@@ -410,6 +410,11 @@ export default {
     imageTooMany: "最多上传5张图片",
     imageTypeSizeError: "仅支持 JPG/PNG/GIF/WEBP 格式，单张不超过 10MB",
     imageUploadTooltip: "上传图片（支持粘贴/拖拽）",
+    attachmentUploadTooltip: "上传附件（文档、音频等）",
+    attachmentWithCount: "已上传 {count} 个附件",
+    attachmentTooMany: "最多上传 {max} 个附件",
+    attachmentTooLarge: "文件 {name} 超过 {max}MB 限制",
+    attachmentTypeNotSupported: "不支持的文件类型：{name}",
   },
   settings: {
     title: "设置",
@@ -3341,6 +3346,13 @@ export default {
       notConfigured: "未配置",
       goStorageSettings: "去存储设置中配置",
     },
+    audioUpload: {
+      label: "语音上传",
+      desc: "启用后用户可在对话中上传音频文件，系统将使用 ASR 模型自动转录为文字",
+      asrModel: "ASR 模型",
+      asrModelDesc: "用于音频转录的语音识别模型，未配置时音频文件将以占位符形式传递",
+      asrModelPlaceholder: "请选择 ASR 模型",
+    },
     faq: {
       title: "FAQ 优先策略",
       tooltip: "当知识库中包含 FAQ（问答对）时，可以启用此策略让 FAQ 答案优先于普通文档",

frontend/src/stores/menu.ts

@@ -38,6 +38,7 @@ export const useMenuStore = defineStore('menuStore', () => {
   const firstMentionedItems = ref<any[]>([])
   const firstModelId = ref('')
   const firstImageFiles = ref<any[]>([])
+  const firstAttachmentFiles = ref<any[]>([])
   const prefillQuery = ref('')
 
   const applyMenuTranslations = () => {
@@ -99,11 +100,12 @@ export const useMenuStore = defineStore('menuStore', () => {
     isFirstSession.value = payload
   }
 
-  const changeFirstQuery = (payload: string, mentionedItems: any[] = [], modelId: string = '', imageFiles: any[] = []) => {
+  const changeFirstQuery = (payload: string, mentionedItems: any[] = [], modelId: string = '', imageFiles: any[] = [], attachmentFiles: any[] = []) => {
     firstQuery.value = payload
     firstMentionedItems.value = mentionedItems
     firstModelId.value = modelId
     firstImageFiles.value = imageFiles
+    firstAttachmentFiles.value = attachmentFiles
   }
 
   const setPrefillQuery = (q: string) => {
@@ -123,6 +125,7 @@ export const useMenuStore = defineStore('menuStore', () => {
     firstMentionedItems,
     firstModelId,
     firstImageFiles,
+    firstAttachmentFiles,
     prefillQuery,
     clearMenuArr,
     updatemenuArr,

frontend/src/views/agent/AgentEditorModal.vue

@@ -394,6 +394,35 @@
                         </a>
                       </div>
                     </div>
+
+                    <!-- 音频上传开关 -->
+                    <div class="setting-row">
+                      <div class="setting-info">
+                        <label>{{ $t('agentEditor.audioUpload.label') }}</label>
+                        <p class="desc">{{ $t('agentEditor.audioUpload.desc') }}</p>
+                      </div>
+                      <div class="setting-control">
+                        <t-switch v-model="formData.config.audio_upload_enabled" />
+                      </div>
+                    </div>
+
+                    <!-- ASR模型（音频上传启用时） -->
+                    <div v-if="formData.config.audio_upload_enabled" class="setting-row">
+                      <div class="setting-info">
+                        <label>{{ $t('agentEditor.audioUpload.asrModel') }}</label>
+                        <p class="desc">{{ $t('agentEditor.audioUpload.asrModelDesc') }}</p>
+                      </div>
+                      <div class="setting-control">
+                        <ModelSelector
+                          model-type="ASR"
+                          :selected-model-id="formData.config.asr_model_id"
+                          :all-models="allModels"
+                          @update:selected-model-id="(val: string) => formData.config.asr_model_id = val"
+                          @add-model="handleAddModel('asr')"
+                          :placeholder="$t('agentEditor.audioUpload.asrModelPlaceholder')"
+                        />
+                      </div>
+                    </div>
                   </div>
                 </div>
 

frontend/src/views/chat/components/usermsg.vue

@@ -27,6 +27,15 @@
                 @click="previewImage($event)"
             />
         </div>
+        <!-- 显示上传的附件 -->
+        <div v-if="hasAttachments" class="user_attachments">
+            <div v-for="(att, idx) in props.attachments" :key="idx" class="user_attachment_card">
+                <div class="attachment_card_icon">
+                    <t-icon :name="getAttachmentIcon(att.file_name || att.file_type)" />
+                </div>
+                <div class="attachment_card_name">{{ att.file_name }}</div>
+            </div>
+        </div>
         <div class="user_msg">
             {{ content }}
         </div>
@@ -56,6 +65,11 @@ const props = defineProps({
         required: false,
         default: () => []
     },
+    attachments: {
+        type: Array,
+        required: false,
+        default: () => []
+    },
     channel: {
         type: String,
         required: false,
@@ -79,6 +93,25 @@ const channelClass = computed(() => props.channel ? `channel-${props.channel}` :
 
 const containerRef = ref(null);
 const hasImages = computed(() => props.images && props.images.length > 0);
+const hasAttachments = computed(() => props.attachments && props.attachments.length > 0);
+
+const getAttachmentIcon = (fileNameOrType) => {
+    const ext = (fileNameOrType || '').split('.').pop()?.toLowerCase();
+    if (['pdf'].includes(ext)) return 'file-pdf';
+    if (['doc', 'docx'].includes(ext)) return 'file-word';
+    if (['xls', 'xlsx'].includes(ext)) return 'file-excel';
+    if (['ppt', 'pptx'].includes(ext)) return 'file-powerpoint';
+    if (['txt', 'md'].includes(ext)) return 'file-text';
+    if (['mp3', 'wav', 'm4a', 'flac', 'ogg', 'aac'].includes(ext)) return 'sound';
+    return 'file';
+};
+
+const formatFileSize = (bytes) => {
+    if (!bytes) return '';
+    if (bytes < 1024) return bytes + ' B';
+    if (bytes < 1024 * 1024) return (bytes / 1024).toFixed(1) + ' KB';
+    return (bytes / (1024 * 1024)).toFixed(1) + ' MB';
+};
 
 const hydrateImages = async () => {
     await nextTick();
@@ -198,6 +231,47 @@ const closePreImg = () => {
     max-width: 100%;
 }
 
+.user_attachments {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 8px;
+    justify-content: flex-end;
+    max-width: 100%;
+}
+
+.user_attachment_card {
+    width: 60px;
+    height: 60px;
+    border-radius: 8px;
+    border: 1px solid var(--td-border-level-1-color, #e7e7e7);
+    background: var(--td-bg-color-secondarycontainer, #f5f5f5);
+    display: flex;
+    flex-direction: column;
+    align-items: center;
+    justify-content: center;
+    gap: 4px;
+    overflow: hidden;
+    cursor: default;
+
+    .attachment_card_icon {
+        font-size: 22px;
+        color: var(--td-brand-color, #07C05F);
+        line-height: 1;
+    }
+
+    .attachment_card_name {
+        font-size: 10px;
+        color: var(--td-text-color-secondary, #666);
+        text-align: center;
+        width: 100%;
+        padding: 0 4px;
+        overflow: hidden;
+        text-overflow: ellipsis;
+        white-space: nowrap;
+        box-sizing: border-box;
+    }
+}
+
 .user_image_thumb {
     width: 120px;
     height: 120px;

frontend/src/views/chat/index.vue

@@ -23,7 +23,7 @@
                 </div>
                 <div v-for="(session, id) in messagesList" :key='id'>
                     <div v-if="session.role == 'user'">
-                        <usermsg :content="session.content" :mentioned_items="session.mentioned_items" :images="session.images"></usermsg>
+                        <usermsg :content="session.content" :mentioned_items="session.mentioned_items" :images="session.images" :attachments="session.attachments"></usermsg>
                     </div>
                     <div v-if="session.role == 'assistant'">
                         <botmsg :content="session.content" :session="session" :user-query="getUserQuery(id)" @scroll-bottom="scrollToBottom"
@@ -43,7 +43,7 @@
         <div style="min-height: 115px; margin: 16px auto 4px;width: 100%;max-width: 800px;">
             <InputField
                 ref="inputFieldRef"
-                @send-msg="(query, modelId, mentionedItems, imageFiles) => sendMsg(query, modelId, mentionedItems, imageFiles)"
+                @send-msg="(query, modelId, mentionedItems, imageFiles, attachmentFiles) => sendMsg(query, modelId, mentionedItems, imageFiles, attachmentFiles)"
                 @stop-generation="handleStopGeneration"
                 :isReplying="isReplying"
                 :sessionId="session_id"
@@ -82,7 +82,7 @@ const useSettingsStoreInstance = useSettingsStore();
 const uiStore = useUIStore();
 const { navigateToKnowledgeBaseList } = useKnowledgeBaseCreationNavigation();
 const { t } = useI18n();
-const { menuArr, isFirstSession, firstQuery, firstMentionedItems, firstModelId, firstImageFiles } = storeToRefs(usemenuStore);
+const { menuArr, isFirstSession, firstQuery, firstMentionedItems, firstModelId, firstImageFiles, firstAttachmentFiles } = storeToRefs(usemenuStore);
 const { output, onChunk, isStreaming, isLoading, error, startStream, stopStream } = useStream();
 const route = useRoute();
 const router = useRouter();
@@ -418,7 +418,7 @@ const handleStopGeneration = () => {
     // API 调用成功后，后端的 stop 事件会清空它
 };
 
-const sendMsg = async (value, modelId = '', mentionedItems = [], imageFiles = []) => {
+const sendMsg = async (value, modelId = '', mentionedItems = [], imageFiles = [], attachmentFiles = []) => {
     userquery.value = value;
     isReplying.value = true;
     loading.value = true;
@@ -441,8 +441,39 @@ const sendMsg = async (value, modelId = '', mentionedItems = [], imageFiles = []
         }
     }
 
+    // Convert attachment files to base64 for backend processing
+    let attachmentUploads = [];
+    if (attachmentFiles && attachmentFiles.length > 0) {
+        try {
+            for (const attachment of attachmentFiles) {
+                const reader = new FileReader();
+                const base64Promise = new Promise((resolve, reject) => {
+                    reader.onload = () => {
+                        const result = reader.result;
+                        // Extract base64 content (remove data:...;base64, prefix)
+                        const base64 = result.split(',')[1];
+                        resolve(base64);
+                    };
+                    reader.onerror = reject;
+                    reader.readAsDataURL(attachment.file);
+                });
+                const base64Data = await base64Promise;
+                attachmentUploads.push({
+                    data: base64Data,
+                    file_name: attachment.name,
+                    file_size: attachment.size
+                });
+            }
+        } catch (e) {
+            console.error('[Attachment] Failed to read attachments:', e);
+            loading.value = false;
+            isReplying.value = false;
+            return;
+        }
+    }
+
     // 将@提及的知识库和文件信息存入用户消息
-    messagesList.push({ content: value, role: 'user', mentioned_items: mentionedItems, images: userImages, channel: 'web' });
+    messagesList.push({ content: value, role: 'user', mentioned_items: mentionedItems, images: userImages, attachments: attachmentFiles.map(a => ({ file_name: a.name, file_size: a.size, file_type: '.' + a.name.split('.').pop()?.toLowerCase() })), channel: 'web' });
     scrollToBottom();
     
     // Get agent mode status from settings store
@@ -492,6 +523,7 @@ const sendMsg = async (value, modelId = '', mentionedItems = [], imageFiles = []
         mcp_service_ids: mcpServiceIds,
         mentioned_items: mentionedItems,
         images: imageAttachments.length > 0 ? imageAttachments : undefined,
+        attachment_uploads: attachmentUploads.length > 0 ? attachmentUploads : undefined,
         query: value, 
         method: 'POST', 
         url: endpoint
@@ -1078,8 +1110,8 @@ onMounted(async () => {
     checkmenuTitle(session_id.value)
     if (firstQuery.value) {
         scrollLock.value = true;
-        sendMsg(firstQuery.value, firstModelId.value || '', firstMentionedItems.value || [], firstImageFiles.value || []);
-        usemenuStore.changeFirstQuery('', [], '', []);
+        sendMsg(firstQuery.value, firstModelId.value || '', firstMentionedItems.value || [], firstImageFiles.value || [], firstAttachmentFiles.value || []);
+        usemenuStore.changeFirstQuery('', [], '', [], []);
     } else {
         scrollLock.value = false;
         let data = {

frontend/src/views/creatChat/creatChat.vue

@@ -162,11 +162,11 @@ const handleSuggestedQuestionClick = (question: string) => {
     inputFieldRef.value?.triggerSend(question);
 };
 
-const sendMsg = (value: string, modelId: string, mentionedItems: any[], imageFiles: any[] = []) => {
-    createNewSession(value, modelId, mentionedItems, imageFiles);
+const sendMsg = (value: string, modelId: string, mentionedItems: any[], imageFiles: any[] = [], attachmentFiles: any[] = []) => {
+    createNewSession(value, modelId, mentionedItems, imageFiles, attachmentFiles);
 }
 
-async function createNewSession(value: string, modelId: string, mentionedItems: any[] = [], imageFiles: any[] = []) {
+async function createNewSession(value: string, modelId: string, mentionedItems: any[] = [], imageFiles: any[] = [], attachmentFiles: any[] = []) {
     const selectedKbs = settingsStore.settings.selectedKnowledgeBases || [];
     const selectedFiles = settingsStore.settings.selectedFiles || [];
 
@@ -186,7 +186,7 @@ async function createNewSession(value: string, modelId: string, mentionedItems:
     try {
         const res = await createSessions(sessionData);
         if (res.data && res.data.id) {
-            await navigateToSession(res.data.id, value, modelId, mentionedItems, imageFiles);
+            await navigateToSession(res.data.id, value, modelId, mentionedItems, imageFiles, attachmentFiles);
         } else {
             console.error('[createChat] Failed to create session');
             MessagePlugin.error(t('createChat.messages.createFailed'));
@@ -197,7 +197,7 @@ async function createNewSession(value: string, modelId: string, mentionedItems:
     }
 }
 
-const navigateToSession = async (sessionId: string, value: string, modelId: string, mentionedItems: any[], imageFiles: any[] = []) => {
+const navigateToSession = async (sessionId: string, value: string, modelId: string, mentionedItems: any[], imageFiles: any[] = [], attachmentFiles: any[] = []) => {
     const now = new Date().toISOString();
     let obj = { 
         title: t('createChat.newSessionTitle'), 
@@ -210,7 +210,7 @@ const navigateToSession = async (sessionId: string, value: string, modelId: stri
     };
     usemenuStore.updataMenuChildren(obj);
     usemenuStore.changeIsFirstSession(true);
-    usemenuStore.changeFirstQuery(value, mentionedItems, modelId, imageFiles);
+    usemenuStore.changeFirstQuery(value, mentionedItems, modelId, imageFiles, attachmentFiles);
     router.push(`/platform/chat/${sessionId}`);
 }
 

internal/application/service/chat_pipeline/into_chat_message.go

@@ -95,6 +95,10 @@ func (p *PluginIntoChatMessage) OnEvent(ctx context.Context,
 		if chatManage.QuotedContext != "" {
 			userContent += "\n\n" + chatManage.QuotedContext
 		}
+		// Inject attachment content (documents, audio transcripts, etc.)
+		if len(chatManage.Attachments) > 0 {
+			userContent += chatManage.Attachments.BuildPrompt()
+		}
 
 		if tpl := chatManage.SummaryConfig.ContextTemplate; tpl != "" {
 			chatManage.UserContent = types.RenderPromptPlaceholders(tpl, types.PlaceholderValues{
@@ -180,6 +184,10 @@ func (p *PluginIntoChatMessage) OnEvent(ctx context.Context,
 	if chatManage.QuotedContext != "" {
 		userContent += "\n\n" + chatManage.QuotedContext
 	}
+	// Inject attachment content (documents, audio transcripts, etc.)
+	if len(chatManage.Attachments) > 0 {
+		userContent += chatManage.Attachments.BuildPrompt()
+	}
 
 	// Set formatted content back to chat management
 	chatManage.UserContent = userContent

internal/application/service/session_knowledge_qa.go

@@ -121,6 +121,7 @@ func (s *sessionService) KnowledgeQA(
 			Images:                  req.ImageURLs,
 			VLMModelID:              vlmModelID,
 			ChatModelSupportsVision: chatModelSupportsVision,
+			Attachments:             req.Attachments,
 			Language:                types.LanguageNameFromContext(ctx),
 		},
 		PipelineState: types.PipelineState{
@@ -154,6 +155,10 @@ func (s *sessionService) KnowledgeQA(
 		if req.QuotedContext != "" {
 			userContent += "\n\n" + req.QuotedContext
 		}
+		// Inject attachment content for pure-chat path (RAG path handles this in INTO_CHAT_MESSAGE).
+		if len(req.Attachments) > 0 {
+			userContent += req.Attachments.BuildPrompt()
+		}
 		chatManage.UserContent = userContent
 
 		pipeline = types.NewPipelineBuilder().

internal/handler/session/attachment_processor.go

@@ -0,0 +1,320 @@
+package session
+
+import (
+	"bufio"
+	"bytes"
+	"context"
+	"encoding/base64"
+	"fmt"
+	"path/filepath"
+	"strings"
+
+	"github.com/Tencent/WeKnora/internal/infrastructure/docparser"
+	"github.com/Tencent/WeKnora/internal/logger"
+	"github.com/Tencent/WeKnora/internal/types"
+	"github.com/Tencent/WeKnora/internal/types/interfaces"
+	secutils "github.com/Tencent/WeKnora/internal/utils"
+	"github.com/google/uuid"
+)
+
+const (
+	// maxTextFileLines is the line limit for inline text content; excess lines are truncated.
+	maxTextFileLines = 500
+	// textFileExtensions lists plain-text extensions handled by the line-based reader.
+	textFileExtensions = ".txt,.md,.markdown,.json,.xml,.yaml,.yml,.csv,.log"
+)
+
+// AttachmentProcessor saves uploaded file attachments and extracts their text content
+// for injection into the LLM prompt.
+type AttachmentProcessor struct {
+	fileService    interfaces.FileService
+	documentReader interfaces.DocumentReader
+	imageResolver  *docparser.ImageResolver
+	modelService   interfaces.ModelService // used to obtain the ASR model
+}
+
+// NewAttachmentProcessor creates an AttachmentProcessor with the given dependencies.
+func NewAttachmentProcessor(
+	fileService interfaces.FileService,
+	documentReader interfaces.DocumentReader,
+	imageResolver *docparser.ImageResolver,
+	modelService interfaces.ModelService,
+) *AttachmentProcessor {
+	return &AttachmentProcessor{
+		fileService:    fileService,
+		documentReader: documentReader,
+		imageResolver:  imageResolver,
+		modelService:   modelService,
+	}
+}
+
+// ProcessAttachment validates, saves, and extracts content from a single uploaded file.
+// Content extraction is attempted for all supported types; errors are non-fatal (logged as warnings).
+func (p *AttachmentProcessor) ProcessAttachment(
+	ctx context.Context,
+	data []byte,
+	fileName string,
+	fileSize int64,
+	tenantID uint64,
+	asrModelID string, // optional; enables audio transcription when set
+) (*types.MessageAttachment, error) {
+	logger.Infof(ctx, "processing attachment: fileName=%s, fileSize=%d", secutils.SanitizeForLog(fileName), fileSize)
+
+	// Validate filename (injection / path-traversal checks)
+	safeFileName, isValid := secutils.ValidateInput(fileName)
+	if !isValid {
+		return nil, fmt.Errorf("invalid characters in file name")
+	}
+
+	baseName, err := secutils.SafeFileName(safeFileName)
+	if err != nil {
+		return nil, fmt.Errorf("unsafe file name: %w", err)
+	}
+
+	ext := strings.ToLower(filepath.Ext(baseName))
+	if ext == "" {
+		ext = ".txt"
+	}
+
+	if !isValidFileType(baseName) {
+		return nil, fmt.Errorf("unsupported file type: %s", ext)
+	}
+
+	uniqueFileName := fmt.Sprintf("attachment_%s%s", uuid.New().String()[:12], ext)
+
+	storageURL, err := p.fileService.SaveBytes(ctx, data, tenantID, uniqueFileName, false)
+	if err != nil {
+		return nil, fmt.Errorf("failed to save attachment: %w", err)
+	}
+
+	attachment := &types.MessageAttachment{
+		URL:      storageURL,
+		FileName: baseName,
+		FileType: ext,
+		FileSize: fileSize,
+	}
+
+	// Extract text content based on file type; errors are non-fatal.
+	if p.isTextFile(ext) {
+		if err := p.processTextFile(ctx, data, attachment); err != nil {
+			logger.Warnf(ctx, "text file processing failed: %v", err)
+		}
+	} else if docparser.IsAudioFormat(ext) {
+		if err := p.processAudioFile(ctx, data, baseName, attachment, asrModelID); err != nil {
+			logger.Warnf(ctx, "audio transcription failed: %v, keeping placeholder", err)
+		}
+	} else if docparser.IsSimpleFormat(ext) {
+		if err := p.processWithDocParser(ctx, data, baseName, ext, attachment, tenantID); err != nil {
+			logger.Warnf(ctx, "SimpleFormatReader failed: %v", err)
+		}
+	} else {
+		if err := p.processWithDocumentReader(ctx, data, baseName, ext, attachment, tenantID); err != nil {
+			logger.Warnf(ctx, "DocumentReader failed: %v, keeping metadata only", err)
+		}
+	}
+
+	logger.Infof(ctx, "attachment processed: fileName=%s, truncated=%v, contentLen=%d",
+		secutils.SanitizeForLog(baseName), attachment.IsTruncated, len(attachment.Content))
+
+	return attachment, nil
+}
+
+// isTextFile reports whether ext is a plain-text extension handled line-by-line.
+func (p *AttachmentProcessor) isTextFile(ext string) bool {
+	return strings.Contains(textFileExtensions, ext)
+}
+
+// processTextFile reads plain-text content line by line, truncating at maxTextFileLines.
+func (p *AttachmentProcessor) processTextFile(ctx context.Context, data []byte, attachment *types.MessageAttachment) error {
+	scanner := bufio.NewScanner(bytes.NewReader(data))
+	var lines []string
+	lineCount := 0
+
+	for scanner.Scan() {
+		lineCount++
+		if lineCount <= maxTextFileLines {
+			lines = append(lines, scanner.Text())
+		}
+	}
+
+	if err := scanner.Err(); err != nil {
+		return fmt.Errorf("failed to read file content: %w", err)
+	}
+
+	attachment.LineCount = lineCount
+	attachment.Content = strings.Join(lines, "\n")
+
+	if lineCount > maxTextFileLines {
+		attachment.IsTruncated = true
+		logger.Infof(ctx, "text file truncated: total=%d, kept=%d", lineCount, maxTextFileLines)
+	}
+
+	return nil
+}
+
+// processWithDocParser extracts content via SimpleFormatReader (md, csv, json, images, etc.).
+func (p *AttachmentProcessor) processWithDocParser(
+	ctx context.Context,
+	data []byte,
+	fileName string,
+	fileType string,
+	attachment *types.MessageAttachment,
+	tenantID uint64,
+) error {
+	reader := &docparser.SimpleFormatReader{}
+	result, err := reader.Read(ctx, &types.ReadRequest{
+		FileContent: data,
+		FileName:    fileName,
+		FileType:    fileType,
+	})
+	if err != nil {
+		return fmt.Errorf("SimpleFormatReader failed: %w", err)
+	}
+
+	// Resolve embedded image refs to storage URLs.
+	if len(result.ImageRefs) > 0 && p.imageResolver != nil {
+		updatedMarkdown, _, err := p.imageResolver.ResolveAndStore(ctx, result, p.fileService, tenantID)
+		if err != nil {
+			logger.Warnf(ctx, "image resolution failed: %v", err)
+		} else {
+			result.MarkdownContent = updatedMarkdown
+		}
+	}
+
+	p.applyLineTruncation(ctx, result.MarkdownContent, attachment)
+	return nil
+}
+
+// processAudioFile transcribes audio via ASR. Falls back to a placeholder when no ASR model is configured.
+func (p *AttachmentProcessor) processAudioFile(
+	ctx context.Context,
+	data []byte,
+	fileName string,
+	attachment *types.MessageAttachment,
+	asrModelID string,
+) error {
+	if asrModelID == "" || p.modelService == nil {
+		attachment.Content = fmt.Sprintf("[audio file: %s]", fileName)
+		logger.Infof(ctx, "no ASR model configured, keeping audio placeholder")
+		return nil
+	}
+
+	asrInstance, err := p.modelService.GetASRModel(ctx, asrModelID)
+	if err != nil {
+		return fmt.Errorf("failed to get ASR model: %w", err)
+	}
+
+	logger.Infof(ctx, "starting audio transcription: fileName=%s, size=%d", fileName, len(data))
+	transcript, err := asrInstance.Transcribe(ctx, data, fileName)
+	if err != nil {
+		return fmt.Errorf("audio transcription failed: %w", err)
+	}
+
+	p.applyLineTruncation(ctx, transcript, attachment)
+	logger.Infof(ctx, "audio transcription done: textLen=%d", len(transcript))
+	return nil
+}
+
+// processWithDocumentReader extracts content from complex formats (pdf, docx, xlsx, etc.).
+func (p *AttachmentProcessor) processWithDocumentReader(
+	ctx context.Context,
+	data []byte,
+	fileName string,
+	fileType string,
+	attachment *types.MessageAttachment,
+	tenantID uint64,
+) error {
+	if p.documentReader == nil {
+		return fmt.Errorf("DocumentReader not configured")
+	}
+
+	result, err := p.documentReader.Read(ctx, &types.ReadRequest{
+		FileContent: data,
+		FileName:    fileName,
+		FileType:    fileType,
+	})
+	if err != nil {
+		return fmt.Errorf("DocumentReader failed: %w", err)
+	}
+
+	// Resolve embedded image refs to storage URLs.
+	if len(result.ImageRefs) > 0 && p.imageResolver != nil {
+		updatedMarkdown, _, err := p.imageResolver.ResolveAndStore(ctx, result, p.fileService, tenantID)
+		if err != nil {
+			logger.Warnf(ctx, "image resolution failed: %v", err)
+		} else {
+			result.MarkdownContent = updatedMarkdown
+		}
+	}
+
+	p.applyLineTruncation(ctx, result.MarkdownContent, attachment)
+	return nil
+}
+
+// applyLineTruncation stores content into attachment, truncating at maxTextFileLines if needed.
+func (p *AttachmentProcessor) applyLineTruncation(ctx context.Context, content string, attachment *types.MessageAttachment) {
+	lines := strings.Split(content, "\n")
+	lineCount := len(lines)
+	attachment.LineCount = lineCount
+
+	if lineCount > maxTextFileLines {
+		attachment.Content = strings.Join(lines[:maxTextFileLines], "\n")
+		attachment.IsTruncated = true
+		logger.Infof(ctx, "content truncated: total=%d, kept=%d", lineCount, maxTextFileLines)
+	} else {
+		attachment.Content = content
+	}
+}
+
+// isValidFileType reports whether fileName has a supported extension.
+// Kept in sync with the frontend SUPPORTED_TYPES list in AttachmentUpload.vue.
+func isValidFileType(fileName string) bool {
+	ext := strings.ToLower(filepath.Ext(fileName))
+	if ext == "" {
+		return false
+	}
+	ext = strings.TrimPrefix(ext, ".")
+
+	supportedTypes := []string{
+		// documents
+		"docx", "doc", "pdf", "ppt", "pptx",
+		// spreadsheets
+		"xlsx", "xls",
+		// text / markup
+		"md", "markdown", "txt", "csv", "json", "xml", "yaml", "yml", "log", "html",
+		// images
+		"jpg", "jpeg", "png", "gif", "bmp", "tiff", "webp",
+		// audio
+		"mp3", "wav", "m4a", "flac", "ogg", "aac",
+	}
+
+	for _, t := range supportedTypes {
+		if ext == t {
+			return true
+		}
+	}
+	return false
+}
+
+// DecodeBase64Attachment decodes a base64 attachment payload, stripping any data URI prefix.
+// Tries Std, URL, RawStd, and RawURL encodings in order.
+func DecodeBase64Attachment(data string) ([]byte, error) {
+	// Strip data URI prefix (e.g. "data:application/pdf;base64,")
+	if idx := strings.Index(data, ","); idx != -1 {
+		data = data[idx+1:]
+	}
+	data = strings.TrimSpace(data)
+
+	for _, enc := range []struct{ e *base64.Encoding }{
+		{base64.StdEncoding},
+		{base64.URLEncoding},
+		{base64.RawStdEncoding},
+		{base64.RawURLEncoding},
+	} {
+		if decoded, err := enc.e.DecodeString(data); err == nil {
+			return decoded, nil
+		}
+	}
+
+	return nil, fmt.Errorf("base64 decode failed: unrecognised encoding")
+}

internal/handler/session/handler.go

@@ -5,6 +5,7 @@ import (
 
 	"github.com/Tencent/WeKnora/internal/config"
 	"github.com/Tencent/WeKnora/internal/errors"
+	"github.com/Tencent/WeKnora/internal/infrastructure/docparser"
 	"github.com/Tencent/WeKnora/internal/logger"
 	"github.com/Tencent/WeKnora/internal/types"
 	"github.com/Tencent/WeKnora/internal/types/interfaces"
@@ -24,6 +25,7 @@ type Handler struct {
 	agentShareService    interfaces.AgentShareService    // Service for resolving shared agents (KB scope in retrieval)
 	fileService          interfaces.FileService          // Service for file storage (image uploads)
 	modelService         interfaces.ModelService         // Service for model management (VLM access)
+	attachmentProcessor  *AttachmentProcessor            // Processor for file attachments
 }
 
 // NewHandler creates a new instance of Handler with all necessary dependencies
@@ -38,6 +40,8 @@ func NewHandler(
 	agentShareService interfaces.AgentShareService,
 	fileService interfaces.FileService,
 	modelService interfaces.ModelService,
+	documentReader interfaces.DocumentReader,
+	imageResolver *docparser.ImageResolver,
 ) *Handler {
 	return &Handler{
 		sessionService:       sessionService,
@@ -50,6 +54,12 @@ func NewHandler(
 		agentShareService:    agentShareService,
 		fileService:          fileService,
 		modelService:         modelService,
+		attachmentProcessor: NewAttachmentProcessor(
+			fileService,
+			documentReader,
+			imageResolver,
+			modelService,
+		),
 	}
 }
 

internal/handler/session/helpers.go

@@ -166,7 +166,7 @@ func createAgentQueryEvent(sessionID, assistantMessageID string) interfaces.Stre
 }
 
 // createUserMessage creates a user message and returns the created message.
-func (h *Handler) createUserMessage(ctx context.Context, sessionID, query, requestID string, mentionedItems types.MentionedItems, images types.MessageImages, channel string) (*types.Message, error) {
+func (h *Handler) createUserMessage(ctx context.Context, sessionID, query, requestID string, mentionedItems types.MentionedItems, images types.MessageImages, attachments types.MessageAttachments, channel string) (*types.Message, error) {
 	return h.messageService.CreateMessage(ctx, &types.Message{
 		SessionID:      sessionID,
 		Role:           "user",
@@ -176,6 +176,7 @@ func (h *Handler) createUserMessage(ctx context.Context, sessionID, query, reque
 		IsCompleted:    true,
 		MentionedItems: mentionedItems,
 		Images:         images,
+		Attachments:    attachments,
 		Channel:        channel,
 	})
 }

internal/handler/session/qa.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"net/http"
 	"runtime"
+	"sync"
 	"time"
 
 	"github.com/Tencent/WeKnora/internal/errors"
@@ -37,6 +38,7 @@ type qaRequestContext struct {
 	images            []ImageAttachment // Uploaded images with analysis text
 	userMessageID     string            // Created user message ID (populated after createUserMessage)
 	channel           string            // Source channel: "web", "api", "im", etc.
+	attachments       types.MessageAttachments // Processed file attachments
 }
 
 // buildQARequest converts the qaRequestContext into a types.QARequest for service invocation.
@@ -55,6 +57,7 @@ func (rc *qaRequestContext) buildQARequest() *types.QARequest {
 		UserMessageID:      rc.userMessageID,
 		WebSearchEnabled:   rc.webSearchEnabled,
 		EnableMemory:       rc.enableMemory,
+		Attachments:        rc.attachments,
 	}
 }
 
@@ -135,6 +138,67 @@ func (h *Handler) parseQARequest(c *gin.Context, logPrefix string) (*qaRequestCo
 		// - Normal pure-chat mode: runs in the async goroutine with progress events
 	}
 
+	// Process file attachments: decode and save to storage, extract content
+	var processedAttachments types.MessageAttachments
+	if len(request.AttachmentUploads) > 0 {
+		logger.Infof(ctx, "[%s] processing %d attachment(s)", logPrefix, len(request.AttachmentUploads))
+
+		maxSize := secutils.GetMaxFileSize()
+		for i, upload := range request.AttachmentUploads {
+			if upload.FileSize > maxSize {
+				return nil, nil, errors.NewBadRequestError(
+					fmt.Sprintf("attachment %d exceeds size limit of %dMB", i+1, secutils.GetMaxFileSizeMB()))
+			}
+		}
+
+		tenantID := c.GetUint64(types.TenantIDContextKey.String())
+
+		// Use ASR only when the agent has audio upload enabled.
+		asrModelID := ""
+		if customAgent != nil && customAgent.Config.AudioUploadEnabled && customAgent.Config.ASRModelID != "" {
+			asrModelID = customAgent.Config.ASRModelID
+		}
+
+		// Process all attachments concurrently.
+		processedAttachments = make(types.MessageAttachments, len(request.AttachmentUploads))
+		var wg sync.WaitGroup
+		errChan := make(chan error, len(request.AttachmentUploads))
+
+		for i, upload := range request.AttachmentUploads {
+			wg.Add(1)
+			go func(idx int, att AttachmentUpload) {
+				defer wg.Done()
+
+				data, err := DecodeBase64Attachment(att.Data)
+				if err != nil {
+					errChan <- fmt.Errorf("attachment %d decode failed: %w", idx+1, err)
+					return
+				}
+
+				processed, err := h.attachmentProcessor.ProcessAttachment(
+					ctx, data, att.FileName, att.FileSize, tenantID, asrModelID,
+				)
+				if err != nil {
+					errChan <- fmt.Errorf("attachment %d processing failed: %w", idx+1, err)
+					return
+				}
+
+				processedAttachments[idx] = *processed
+			}(i, upload)
+		}
+
+		wg.Wait()
+		close(errChan)
+
+		if len(errChan) > 0 {
+			err := <-errChan
+			logger.Errorf(ctx, "[%s] attachment processing failed: %v", logPrefix, err)
+			return nil, nil, errors.NewBadRequestError(fmt.Sprintf("attachment processing failed: %v", err))
+		}
+
+		logger.Infof(ctx, "[%s] all attachments processed", logPrefix)
+	}
+
 	// Build request context
 	reqCtx := &qaRequestContext{
 		ctx:         ctx,
@@ -160,6 +224,7 @@ func (h *Handler) parseQARequest(c *gin.Context, logPrefix string) (*qaRequestCo
 		effectiveTenantID: effectiveTenantID,
 		images:            request.Images,
 		channel:           request.Channel,
+		attachments:       processedAttachments,
 	}
 
 	return reqCtx, &request, nil
@@ -478,7 +543,7 @@ func (h *Handler) executeQA(reqCtx *qaRequestContext, mode qaMode, generateTitle
 	}
 
 	// Create user message
-	userMsg, err := h.createUserMessage(ctx, sessionID, reqCtx.query, reqCtx.requestID, reqCtx.mentionedItems, convertImageAttachments(reqCtx.images), reqCtx.channel)
+	userMsg, err := h.createUserMessage(ctx, sessionID, reqCtx.query, reqCtx.requestID, reqCtx.mentionedItems, convertImageAttachments(reqCtx.images), reqCtx.attachments, reqCtx.channel)
 	if err != nil {
 		reqCtx.c.Error(errors.NewInternalServerError(err.Error()))
 		return

internal/handler/session/types.go

@@ -49,9 +49,17 @@ type CreateKnowledgeQARequest struct {
 	DisableTitle     bool                   `json:"disable_title"`                         // Whether to disable auto title generation
 	EnableMemory     bool                   `json:"enable_memory"`                         // Whether memory feature is enabled for this request
 	Images           []ImageAttachment      `json:"images"`                                // Attached images for multimodal chat
+	AttachmentUploads []AttachmentUpload    `json:"attachment_uploads,omitempty"`          // Attached files (documents, audio, etc.)
 	Channel          string                 `json:"channel"`                               // Source channel: "web", "api", "im", etc.
 }
 
+// AttachmentUpload represents a file attachment upload from the client
+type AttachmentUpload struct {
+	Data     string `json:"data"`      // Base64-encoded file content
+	FileName string `json:"file_name"` // Original filename
+	FileSize int64  `json:"file_size"` // File size in bytes
+}
+
 // SearchKnowledgeRequest defines the request structure for searching knowledge without LLM summarization
 type SearchKnowledgeRequest struct {
 	Query            string   `json:"query"              binding:"required"` // Query text to search for

internal/types/chat_manage.go

@@ -45,6 +45,9 @@ type PipelineRequest struct {
 	VLMModelID              string   `json:"-"`
 	ChatModelSupportsVision bool     `json:"-"`
 
+	// File attachments support
+	Attachments MessageAttachments `json:"-"`
+
 	// Misc request-scoped config
 	TenantID              uint64 `json:"-"`
 	WebSearchEnabled      bool   `json:"-"`
@@ -185,6 +188,7 @@ func (c *ChatManage) Clone() *ChatManage {
 			Images:                   append([]string(nil), c.Images...),
 			VLMModelID:               c.VLMModelID,
 			ChatModelSupportsVision:  c.ChatModelSupportsVision,
+			Attachments:              append(MessageAttachments(nil), c.Attachments...),
 			TenantID:                 c.TenantID,
 			WebSearchEnabled:         c.WebSearchEnabled,
 			WebSearchProviderID:     c.WebSearchProviderID,

internal/types/custom_agent.go

@@ -120,6 +120,10 @@ type CustomAgentConfig struct {
 	ImageUploadEnabled bool `yaml:"image_upload_enabled" json:"image_upload_enabled"`
 	// VLM model ID for image analysis (optional, falls back to tenant-level VLM)
 	VLMModelID string `yaml:"vlm_model_id" json:"vlm_model_id"`
+	// Whether audio upload (ASR transcription) is enabled for this agent (default: false)
+	AudioUploadEnabled bool `yaml:"audio_upload_enabled" json:"audio_upload_enabled"`
+	// ASR model ID for audio transcription (optional)
+	ASRModelID string `yaml:"asr_model_id" json:"asr_model_id"`
 	// Storage provider for image uploads: "local", "minio", "cos", "tos"
 	// Empty means use the global/tenant default provider.
 	ImageStorageProvider string `yaml:"image_storage_provider" json:"image_storage_provider"`

internal/types/message.go

@@ -4,6 +4,8 @@ package types
 import (
 	"database/sql/driver"
 	"encoding/json"
+	"fmt"
+	"strings"
 	"time"
 
 	"github.com/google/uuid"
@@ -64,6 +66,79 @@ func (m *MessageImages) Scan(value interface{}) error {
 	return json.Unmarshal(b, m)
 }
 
+// MessageAttachment represents a file attachment in a chat message
+type MessageAttachment struct {
+	URL         string `json:"url"`                    // Storage URL (provider://path)
+	FileName    string `json:"file_name"`              // Original filename
+	FileType    string `json:"file_type"`              // File extension (e.g., ".pdf", ".docx")
+	FileSize    int64  `json:"file_size"`              // File size in bytes
+	Content     string `json:"content,omitempty"`      // Extracted text content (for small text files)
+	IsTruncated bool   `json:"is_truncated,omitempty"` // Whether content was truncated
+	LineCount   int    `json:"line_count,omitempty"`   // Total line count (for text files)
+}
+
+// MessageAttachments is a slice of MessageAttachment for database storage
+type MessageAttachments []MessageAttachment
+
+// BuildPrompt returns a formatted prompt section for all attachments,
+// injecting file metadata and extracted content into the LLM context.
+func (attachments MessageAttachments) BuildPrompt() string {
+	if len(attachments) == 0 {
+		return ""
+	}
+
+	var sb strings.Builder
+	sb.WriteString("\n\n## 用户上传的附件\n\n")
+
+	for i, att := range attachments {
+		sb.WriteString(fmt.Sprintf("### 附件 %d: %s\n\n", i+1, att.FileName))
+		sb.WriteString(fmt.Sprintf("- **文件类型**: %s\n", att.FileType))
+		sb.WriteString(fmt.Sprintf("- **文件大小**: %.2f KB\n\n", float64(att.FileSize)/1024))
+
+		if att.Content != "" {
+			sb.WriteString("**文件内容**:\n\n")
+			sb.WriteString(att.Content)
+			sb.WriteString("\n\n")
+
+			if att.IsTruncated {
+				sb.WriteString(fmt.Sprintf("*注意: 此文件共有 %d 行，已截取前 500 行显示。*\n\n",
+					att.LineCount))
+			}
+		} else {
+			sb.WriteString("*此文件内容提取失败或不支持。*\n\n")
+		}
+	}
+
+	return sb.String()
+}
+
+// Value implements the driver.Valuer interface for database serialization
+func (m MessageAttachments) Value() (driver.Value, error) {
+	if m == nil {
+		return json.Marshal([]MessageAttachment{})
+	}
+	return json.Marshal(m)
+}
+
+// Scan implements the sql.Scanner interface for database deserialization
+func (m *MessageAttachments) Scan(value interface{}) error {
+	if value == nil {
+		*m = make(MessageAttachments, 0)
+		return nil
+	}
+	var b []byte
+	switch v := value.(type) {
+	case []byte:
+		b = v
+	case string:
+		b = []byte(v)
+	default:
+		*m = make(MessageAttachments, 0)
+		return nil
+	}
+	return json.Unmarshal(b, m)
+}
+
 // MentionedItems is a slice of MentionedItem for database storage
 type MentionedItems []MentionedItem
 
@@ -119,6 +194,8 @@ type Message struct {
 	MentionedItems MentionedItems `json:"mentioned_items,omitempty" gorm:"type:jsonb,column:mentioned_items"`
 	// Attached images with OCR/Caption text (for user messages)
 	Images MessageImages `json:"images,omitempty" gorm:"type:jsonb;column:images"`
+	// Attached files (documents, audio, etc., for user messages)
+	Attachments MessageAttachments `json:"attachments,omitempty" gorm:"type:jsonb;column:attachments"`
 	// Whether message generation is complete
 	IsCompleted bool `json:"is_completed"`
 	// Whether this response is a fallback (no knowledge base match found)
@@ -194,6 +271,9 @@ func (m *Message) BeforeCreate(tx *gorm.DB) (err error) {
 	if m.Images == nil {
 		m.Images = make(MessageImages, 0)
 	}
+	if m.Attachments == nil {
+		m.Attachments = make(MessageAttachments, 0)
+	}
 	return nil
 }
 

internal/types/qa_request.go

@@ -17,4 +17,5 @@ type QARequest struct {
 	WebSearchEnabled   bool         // Whether web search is enabled for this request
 	EnableMemory       bool         // Whether memory feature is enabled
 	QuotedContext      string       // Quoted message content from IM quote-reply (appended at LLM prompt stage, not used for retrieval)
+	Attachments        MessageAttachments // File attachments (processed and ready for prompt injection)
 }

migrations/versioned/000033_add_attachments.down.sql

@@ -0,0 +1 @@
+ALTER TABLE messages DROP COLUMN IF EXISTS attachments;

migrations/versioned/000033_add_attachments.up.sql

@@ -0,0 +1 @@
+ALTER TABLE messages ADD COLUMN IF NOT EXISTS attachments JSONB DEFAULT '[]'::jsonb;