fix: 修复 doc 解析的问题，并增加文件类型支持检查和相关API

- 在文件上传功能中添加对不支持文件类型的检查，返回400错误。
- 新增获取支持文件类型的API接口。
- 更新前端文件上传组件，动态加载支持的文件类型并提供用户提示。

docs/changelog/roadmap.md

@@ -6,8 +6,8 @@
 
 
 🐛**BUGs**
-- [ ] 部分 doc 格式的文件支持有问题
-- [ ] 当出现不支持的文件类型的时候，前端没有限制
+- [x] 部分 doc 格式的文件支持有问题
+- [x] 当出现不支持的文件类型的时候，前端没有限制
 - [ ] 当消息生成的时候有报错的时候，前端无显示
 - [ ] 另外一个智能体的历史对话无法显示
 - [ ] 调用统计的统计结果有问题（Token 计算方法可能也不对）

server/routers/knowledge_router.py

@@ -9,7 +9,7 @@ from starlette.responses import FileResponse as StarletteFileResponse
 from src.storage.db.models import User
 from server.utils.auth_middleware import get_admin_user
 from src import config, knowledge_base
-from src.knowledge.indexing import process_file_to_markdown
+from src.knowledge.indexing import SUPPORTED_FILE_EXTENSIONS, is_supported_file_extension, process_file_to_markdown
 from src.models.embed import test_embedding_model_status, test_all_embedding_models_status
 from src.utils import hashstr, logger
 
@@ -522,6 +522,10 @@ async def upload_file(
 
     logger.debug(f"Received upload file with filename: {file.filename}")
 
+    if not is_supported_file_extension(file.filename):
+        ext = os.path.splitext(file.filename)[1].lower()
+        raise HTTPException(status_code=400, detail=f"Unsupported file type: {ext}")
+
     # 根据db_id获取上传路径，如果db_id为None则使用默认路径
     if db_id:
         upload_dir = knowledge_base.get_db_upload_path(db_id)
@@ -539,6 +543,12 @@ async def upload_file(
     return {"message": "File successfully uploaded", "file_path": file_path, "db_id": db_id}
 
 
+@knowledge.get("/files/supported-types")
+async def get_supported_file_types(current_user: User = Depends(get_admin_user)):
+    """获取当前支持的文件类型"""
+    return {"message": "success", "file_types": sorted(SUPPORTED_FILE_EXTENSIONS)}
+
+
 @knowledge.post("/files/markdown")
 async def mark_it_down(file: UploadFile = File(...), current_user: User = Depends(get_admin_user)):
     """调用 src.knowledge.indexing 下面的 process_file_to_markdown 解析为 markdown，参数是文件，需要管理员权限"""

src/knowledge/indexing.py

@@ -5,17 +5,69 @@ from pathlib import Path
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain_community.document_loaders import (
     CSVLoader,
-    Docx2txtLoader,
     JSONLoader,
     PyPDFLoader,
     TextLoader,
     UnstructuredHTMLLoader,
     UnstructuredMarkdownLoader,
+    UnstructuredWordDocumentLoader,
 )
 
 from src.utils import logger
 
 
+SUPPORTED_FILE_EXTENSIONS: tuple[str, ...] = (
+    ".txt",
+    ".md",
+    ".doc",
+    ".docx",
+    ".html",
+    ".htm",
+    ".json",
+    ".csv",
+    ".xls",
+    ".xlsx",
+    ".pdf",
+    ".jpg",
+    ".jpeg",
+    ".png",
+    ".bmp",
+    ".tiff",
+    ".tif",
+)
+
+
+def is_supported_file_extension(file_name: str | os.PathLike[str]) -> bool:
+    """Check whether the given file path has a supported extension."""
+    return Path(file_name).suffix.lower() in SUPPORTED_FILE_EXTENSIONS
+
+
+def _extract_word_text(file_path: Path) -> str:
+    """
+    Parse Word documents (.doc/.docx) into plain text.
+
+    Try python-docx first for docx files and fall back to the unstructured
+    loader so legacy .doc files are still parsed when possible.
+    """
+    try:
+        from docx import Document  # type: ignore
+
+        doc = Document(file_path)
+        text = "\n".join(paragraph.text for paragraph in doc.paragraphs).strip()
+        if text:
+            return text
+    except Exception as docx_error:  # noqa: BLE001
+        logger.warning(f"python-docx failed to parse {file_path.name}: {docx_error}")
+
+    try:
+        loader = UnstructuredWordDocumentLoader(str(file_path))
+        docs = loader.load()
+        return "\n".join(doc.page_content for doc in docs).strip()
+    except Exception as unstructured_error:  # noqa: BLE001
+        logger.error(f"Unstructured failed to parse {file_path.name}: {unstructured_error}")
+        raise ValueError(f"无法解析 Word 文档: {file_path.name}") from unstructured_error
+
+
 def chunk_with_parser(file_path, params=None):
     """
     使用文件解析器将文件切分成固定大小的块
@@ -38,7 +90,7 @@ def chunk_with_parser(file_path, params=None):
         loader = UnstructuredMarkdownLoader(file_path)
 
     elif file_type in [".docx", ".doc"]:
-        loader = Docx2txtLoader(file_path)
+        loader = UnstructuredWordDocumentLoader(file_path)
 
     elif file_type in [".html", ".htm"]:
         loader = UnstructuredHTMLLoader(file_path)
@@ -251,10 +303,7 @@ async def process_file_to_markdown(file_path: str, params: dict | None = None) -
 
     elif file_ext in [".doc", ".docx"]:
         # 处理 Word 文档
-        from docx import Document  # type: ignore
-
-        doc = Document(file_path_obj)
-        text = "\n".join([para.text for para in doc.paragraphs])
+        text = _extract_word_text(file_path_obj)
         return f"# {file_path_obj.name}\n\n{text}"
 
     elif file_ext in [".jpg", ".jpeg", ".png", ".bmp", ".tiff", ".tif"]:

web/src/apis/knowledge_api.js

@@ -173,6 +173,14 @@ export const fileApi = {
         'Content-Type': 'multipart/form-data'
       }
     })
+  },
+
+  /**
+   * 获取支持的文件类型
+   * @returns {Promise} - 文件类型列表
+   */
+  getSupportedFileTypes: async () => {
+    return apiAdminGet('/api/knowledge/files/supported-types')
   }
 }
 
@@ -220,4 +228,3 @@ export const embeddingApi = {
     return apiAdminGet('/api/knowledge/embedding-models/status')
   }
 }
-

web/src/components/FileUploadModal.vue

@@ -105,6 +105,8 @@
           name="file"
           :multiple="true"
           :disabled="chunkLoading"
+          :accept="acceptedFileTypes"
+          :before-upload="beforeUpload"
           :action="'/api/knowledge/files/upload?db_id=' + databaseId"
           :headers="getAuthHeaders()"
           @change="handleFileUpload"
@@ -112,7 +114,7 @@
         >
           <p class="ant-upload-text">点击或者把文件拖拽到这里上传</p>
           <p class="ant-upload-hint">
-            目前仅支持上传文本、图片文件，如 .pdf, .txt, .md, .docx, png, jpg等。
+            支持的文件类型：{{ uploadHint }}
           </p>
         </a-upload-dragger>
       </div>
@@ -174,11 +176,12 @@
 </template>
 
 <script setup>
-import { ref, reactive, computed, watch } from 'vue';
-import { message } from 'ant-design-vue';
+import { ref, computed, onMounted } from 'vue';
+import { message, Upload } from 'ant-design-vue';
 import { useUserStore } from '@/stores/user';
 import { useDatabaseStore } from '@/stores/database';
 import { ocrApi } from '@/apis/system_api';
+import { fileApi } from '@/apis/knowledge_api';
 import {
   FileOutlined,
   LinkOutlined,
@@ -198,6 +201,83 @@ const emit = defineEmits(['update:visible']);
 
 const store = useDatabaseStore();
 
+const DEFAULT_SUPPORTED_TYPES = [
+  '.txt',
+  '.pdf',
+  '.jpg',
+  '.jpeg',
+  '.md',
+  '.docx',
+  '.doc',
+];
+
+const normalizeExtensions = (extensions) => {
+  if (!Array.isArray(extensions)) {
+    return [];
+  }
+  const normalized = extensions
+    .map((ext) => (typeof ext === 'string' ? ext.trim().toLowerCase() : ''))
+    .filter((ext) => ext.length > 0)
+    .map((ext) => (ext.startsWith('.') ? ext : `.${ext}`));
+
+  return Array.from(new Set(normalized)).sort();
+};
+
+const supportedFileTypes = ref(normalizeExtensions(DEFAULT_SUPPORTED_TYPES));
+
+const applySupportedFileTypes = (extensions) => {
+  const normalized = normalizeExtensions(extensions);
+  if (normalized.length > 0) {
+    supportedFileTypes.value = normalized;
+  } else {
+    supportedFileTypes.value = normalizeExtensions(DEFAULT_SUPPORTED_TYPES);
+  }
+};
+
+const acceptedFileTypes = computed(() => {
+  if (!supportedFileTypes.value.length) {
+    return '';
+  }
+  return supportedFileTypes.value.join(',');
+});
+
+const uploadHint = computed(() => {
+  if (!supportedFileTypes.value.length) {
+    return '加载中...';
+  }
+  return supportedFileTypes.value.join(', ');
+});
+
+const isSupportedExtension = (fileName) => {
+  if (!fileName) {
+    return true;
+  }
+  if (!supportedFileTypes.value.length) {
+    return true;
+  }
+  const lastDotIndex = fileName.lastIndexOf('.');
+  if (lastDotIndex === -1) {
+    return false;
+  }
+  const ext = fileName.slice(lastDotIndex).toLowerCase();
+  return supportedFileTypes.value.includes(ext);
+};
+
+const loadSupportedFileTypes = async () => {
+  try {
+    const data = await fileApi.getSupportedFileTypes();
+    applySupportedFileTypes(data?.file_types);
+  } catch (error) {
+    console.error('获取支持的文件类型失败:', error);
+    message.warning('获取支持的文件类型失败，已使用默认配置');
+    applySupportedFileTypes(DEFAULT_SUPPORTED_TYPES);
+  }
+};
+
+onMounted(() => {
+  loadSupportedFileTypes();
+});
+
 const visible = computed({
   get: () => props.visible,
   set: (value) => emit('update:visible', value)
@@ -370,14 +450,24 @@ const handleCancel = () => {
   emit('update:visible', false);
 };
 
-const handleFileUpload = (event) => {
-  console.log(event);
+const beforeUpload = (file) => {
+  if (!isSupportedExtension(file?.name)) {
+    message.error(`不支持的文件类型：${file?.name || '未知文件'}`);
+    return Upload.LIST_IGNORE;
+  }
+  return true;
 };
 
-const handleDrop = (event) => {
-  console.log(event);
+const handleFileUpload = (info) => {
+  if (info?.file?.status === 'error') {
+    const errorMessage = info.file?.response?.detail || `文件上传失败：${info.file.name}`;
+    message.error(errorMessage);
+  }
+  fileList.value = info?.fileList ?? [];
 };
 
+const handleDrop = () => {};
+
 const showChunkConfigModal = () => {
   tempChunkParams.value = {
     chunk_size: chunkParams.value.chunk_size,
@@ -547,4 +637,4 @@ const chunkData = async () => {
 .chunk-config-content .params-info {
   margin-bottom: 16px;
 }
-</style>
+</style>