feat: enhance document upload process with batch settings and improved chunk handling

astrbot/core/db/vec_db/base.py

@@ -24,6 +24,21 @@ class BaseVecDB:
         """
         ...
 
+    @abc.abstractmethod
+    async def insert_batch(
+        self,
+        contents: list[str],
+        metadatas: list[dict] | None = None,
+        ids: list[str] | None = None,
+        batch_size: int = 32,
+        tasks_limit: int = 3,
+        max_retries: int = 3,
+    ) -> int:
+        """
+        批量插入文本和其对应向量，自动生成 ID 并保持一致性。
+        """
+        ...
+
     @abc.abstractmethod
     async def retrieve(
         self,

astrbot/core/db/vec_db/faiss_impl/embedding_storage.py

@@ -36,6 +36,22 @@ class EmbeddingStorage:
         self.index.add_with_ids(vector.reshape(1, -1), np.array([id]))
         await self.save_index()
 
+    async def insert_batch(self, vectors: np.ndarray, ids: list[int]):
+        """批量插入向量
+
+        Args:
+            vectors (np.ndarray): 要插入的向量数组
+            ids (list[int]): 向量的ID列表
+        Raises:
+            ValueError: 如果向量的维度与存储的维度不匹配
+        """
+        assert self.index is not None, "FAISS index is not initialized."
+        if vectors.shape[1] != self.dimension:
+            raise ValueError(
+                f"向量维度不匹配, 期望: {self.dimension}, 实际: {vectors.shape[1]}"
+            )
+        self.index.add_with_ids(vectors, np.array(ids))
+
     async def search(self, vector: np.ndarray, k: int) -> tuple:
         """搜索最相似的向量
 

astrbot/core/db/vec_db/faiss_impl/vec_db.py

@@ -1,11 +1,13 @@
 import uuid
 import json
+import time
 import numpy as np
 from .document_storage import DocumentStorage
 from .embedding_storage import EmbeddingStorage
 from ..base import Result, BaseVecDB
 from astrbot.core.provider.provider import EmbeddingProvider
 from astrbot.core.provider.provider import RerankProvider
+from astrbot import logger
 
 
 class FaissVecDB(BaseVecDB):
@@ -60,6 +62,55 @@ class FaissVecDB(BaseVecDB):
             await self.embedding_storage.insert(vector, int_id)
             return int_id
 
+    async def insert_batch(
+        self,
+        contents: list[str],
+        metadatas: list[dict] | None = None,
+        ids: list[str] | None = None,
+        batch_size: int = 32,
+        tasks_limit: int = 3,
+        max_retries: int = 3,
+    ) -> list[int]:
+        """
+        批量插入文本和其对应向量，自动生成 ID 并保持一致性。
+        """
+        assert self.document_storage.connection is not None, (
+            "Database connection is not initialized."
+        )
+        metadatas = metadatas or [{} for _ in contents]
+        ids = ids or [str(uuid.uuid4()) for _ in contents]
+
+        start = time.time()
+        logger.debug(f"Generating embeddings for {len(contents)} contents...")
+        vectors = await self.embedding_provider.get_embeddings_batch(
+            contents,
+            batch_size=batch_size,
+            tasks_limit=tasks_limit,
+            max_retries=max_retries,
+        )
+        end = time.time()
+        logger.debug(
+            f"Generated embeddings for {len(contents)} contents in {end - start:.2f} seconds."
+        )
+
+        int_ids = []
+        async with self.document_storage.connection.cursor() as cursor:
+            for str_id, content, metadata in zip(ids, contents, metadatas):
+                await cursor.execute(
+                    "INSERT INTO documents (doc_id, text, metadata) VALUES (?, ?, ?)",
+                    (str_id, content, json.dumps(metadata)),
+                )
+            await self.document_storage.connection.commit()
+
+            for str_id in ids:
+                result = await self.document_storage.get_document_by_doc_id(str_id)
+                int_ids.append(result["id"])
+
+            # 批量插入向量到 FAISS
+            vectors_array = np.array(vectors).astype("float32")
+            await self.embedding_storage.insert_batch(vectors_array, int_ids)
+            return int_ids
+
     async def retrieve(
         self,
         query: str,

astrbot/core/knowledge_base/chunking/base.py

@@ -13,7 +13,7 @@ class BaseChunker(ABC):
     """
 
     @abstractmethod
-    async def chunk(self, text: str) -> list[str]:
+    async def chunk(self, text: str, **kwargs) -> list[str]:
         """将文本分块
 
         Args:

astrbot/core/knowledge_base/chunking/fixed_size.py

@@ -22,7 +22,7 @@ class FixedSizeChunker(BaseChunker):
         self.chunk_size = chunk_size
         self.chunk_overlap = chunk_overlap
 
-    async def chunk(self, text: str) -> list[str]:
+    async def chunk(self, text: str, **kwargs) -> list[str]:
         """固定大小分块
 
         Args:
@@ -31,22 +31,25 @@ class FixedSizeChunker(BaseChunker):
         Returns:
             list[str]: 分块后的文本列表
         """
+        chunk_size = kwargs.get("chunk_size", self.chunk_size)
+        chunk_overlap = kwargs.get("chunk_overlap", self.chunk_overlap)
+
         chunks = []
         start = 0
         text_len = len(text)
 
         while start < text_len:
-            end = start + self.chunk_size
+            end = start + chunk_size
             chunk = text[start:end]
 
             if chunk:
                 chunks.append(chunk)
 
             # 移动窗口,保留重叠部分
-            start = end - self.chunk_overlap
+            start = end - chunk_overlap
 
             # 防止无限循环: 如果重叠过大,直接移到end
-            if start >= end or self.chunk_overlap >= self.chunk_size:
+            if start >= end or chunk_overlap >= chunk_size:
                 start = end
 
         return chunks

astrbot/core/knowledge_base/kb_helper.py

@@ -101,13 +101,18 @@ class KBHelper:
         file_name: str,
         file_content: bytes,
         file_type: str,
+        chunk_size: int = 512,
+        chunk_overlap: int = 50,
+        batch_size: int = 32,
+        tasks_limit: int = 3,
+        max_retries: int = 3,
     ) -> KBDocument:
         """上传并处理文档（带原子性保证和失败清理）
 
         流程:
         1. 保存原始文件
         2. 解析文档内容
-        3. 提取多媒体资源 (TODO)
+        3. 提取多媒体资源
         4. 分块处理
         5. 生成向量并存储
         6. 保存元数据（事务）
@@ -116,7 +121,6 @@ class KBHelper:
         await self._ensure_vec_db()
         doc_id = str(uuid.uuid4())
         media_paths: list[Path] = []
-        vec_doc_ids = []
 
         # file_path = self.kb_files_dir / f"{doc_id}.{file_type}"
         # async with aiofiles.open(file_path, "wb") as f:
@@ -144,18 +148,27 @@ class KBHelper:
                 media_paths.append(Path(media.file_path))
 
             # 分块并生成向量
-            # saved_chunks = []
-            chunks_text = await self.chunker.chunk(text_content)
+            chunks_text = await self.chunker.chunk(
+                text_content, chunk_size=chunk_size, chunk_overlap=chunk_overlap
+            )
+            contents = []
+            metadatas = []
             for idx, chunk_text in enumerate(chunks_text):
-                vec_doc_id = await self.vec_db.insert(
-                    content=chunk_text,
-                    metadata={
+                contents.append(chunk_text)
+                metadatas.append(
+                    {
                         "kb_id": self.kb.kb_id,
                         "doc_id": doc_id,
                         "chunk_index": idx,
-                    },
+                    }
                 )
-                vec_doc_ids.append(str(vec_doc_id))
+            await self.vec_db.insert_batch(
+                contents=contents,
+                metadatas=metadatas,
+                batch_size=batch_size,
+                tasks_limit=tasks_limit,
+                max_retries=max_retries,
+            )
 
             # 保存文档的元数据
             doc = KBDocument(

astrbot/core/provider/provider.py

@@ -1,4 +1,5 @@
 import abc
+import asyncio
 from typing import List
 from typing import AsyncGenerator
 from astrbot.core.agent.tool import ToolSet
@@ -203,6 +204,64 @@ class EmbeddingProvider(AbstractProvider):
         """获取向量的维度"""
         ...
 
+    async def get_embeddings_batch(
+        self,
+        texts: list[str],
+        batch_size: int = 16,
+        tasks_limit: int = 3,
+        max_retries: int = 3,
+    ) -> list[list[float]]:
+        """批量获取文本的向量，分批处理以节省内存
+
+        Args:
+            texts: 文本列表
+            batch_size: 每批处理的文本数量
+            tasks_limit: 并发任务数量限制
+            max_retries: 失败时的最大重试次数
+
+        Returns:
+            向量列表
+        """
+        semaphore = asyncio.Semaphore(tasks_limit)
+        all_embeddings: list[list[float]] = []
+        failed_batches: list[tuple[int, list[str]]] = []
+
+        async def process_batch(batch_idx: int, batch_texts: list[str]):
+            async with semaphore:
+                for attempt in range(max_retries):
+                    try:
+                        batch_embeddings = await self.get_embeddings(batch_texts)
+                        all_embeddings.extend(batch_embeddings)
+                        return
+                    except Exception as e:
+                        if attempt == max_retries - 1:
+                            # 最后一次重试失败，记录失败的批次
+                            failed_batches.append((batch_idx, batch_texts))
+                            raise Exception(
+                                f"批次 {batch_idx} 处理失败，已重试 {max_retries} 次: {str(e)}"
+                            )
+                        # 等待一段时间后重试，使用指数退避
+                        await asyncio.sleep(2**attempt)
+
+        tasks = []
+        for i in range(0, len(texts), batch_size):
+            batch_texts = texts[i : i + batch_size]
+            batch_idx = i // batch_size
+            tasks.append(process_batch(batch_idx, batch_texts))
+
+        # 收集所有任务的结果，包括失败的任务
+        results = await asyncio.gather(*tasks, return_exceptions=True)
+
+        # 检查是否有失败的任务
+        errors = [r for r in results if isinstance(r, Exception)]
+        if errors:
+            error_msg = (
+                f"有 {len(errors)} 个批次处理失败: {'; '.join(str(e) for e in errors)}"
+            )
+            raise Exception(error_msg)
+
+        return all_embeddings
+
 
 class RerankProvider(AbstractProvider):
     def __init__(self, provider_config: dict, provider_settings: dict) -> None:

astrbot/dashboard/routes/knowledge_base.py

@@ -428,17 +428,18 @@ class KnowledgeBaseRoute(Route):
         """上传文档
 
         支持两种方式:
-        1. multipart/form-data 文件上传
-        2. JSON 格式 base64 编码上传
+        1. multipart/form-data 文件上传（支持多文件，最多10个）
+        2. JSON 格式 base64 编码上传（支持多文件，最多10个）
 
         Form Data (multipart/form-data):
         - kb_id: 知识库 ID (必填)
-        - file: 文件对象 (必填)
+        - file: 文件对象 (必填，可多个，字段名为 file, file1, file2, ... 或 files[])
 
         JSON Body (application/json):
         - kb_id: 知识库 ID (必填)
-        - file_name: 文件名 (必填)
-        - file_content: base64 编码的文件内容 (必填)
+        - files: 文件数组 (必填)
+          - file_name: 文件名 (必填)
+          - file_content: base64 编码的文件内容 (必填)
         """
         try:
             kb_manager = self._get_kb_manager()
@@ -446,20 +447,46 @@ class KnowledgeBaseRoute(Route):
             # 检查 Content-Type
             content_type = request.content_type
             kb_id = None
+            chunk_size = None
+            chunk_overlap = None
+            batch_size = 32
+            tasks_limit = 3
+            max_retries = 3
+            files_to_upload = []  # 存储待上传的文件信息列表
 
-            if content_type and "multipart/form-data" in content_type:
-                # 方式 1: multipart/form-data
-                form_data = await request.form
-                files = await request.files
+            if content_type and "multipart/form-data" not in content_type:
+                return (
+                    Response().error("Content-Type 须为 multipart/form-data").__dict__
+                )
+            form_data = await request.form
+            files = await request.files
 
-                kb_id = form_data.get("kb_id")
-                if not kb_id:
-                    return Response().error("缺少参数 kb_id").__dict__
+            kb_id = form_data.get("kb_id")
+            chunk_size = int(form_data.get("chunk_size", 512))
+            chunk_overlap = int(form_data.get("chunk_overlap", 50))
+            batch_size = int(form_data.get("batch_size", 32))
+            tasks_limit = int(form_data.get("tasks_limit", 3))
+            max_retries = int(form_data.get("max_retries", 3))
+            if not kb_id:
+                return Response().error("缺少参数 kb_id").__dict__
 
-                if "file" not in files:
-                    return Response().error("缺少文件").__dict__
+            # 收集所有文件
+            file_list = []
+            # 支持 file, file1, file2, ... 或 files[] 格式
+            for key in files.keys():
+                if key == "file" or key.startswith("file") or key == "files[]":
+                    file_items = files.getlist(key)
+                    file_list.extend(file_items)
 
-                file = files["file"]
+            if not file_list:
+                return Response().error("缺少文件").__dict__
+
+            # 限制文件数量
+            if len(file_list) > 10:
+                return Response().error("最多只能上传10个文件").__dict__
+
+            # 处理每个文件
+            for file in file_list:
                 file_name = file.filename
 
                 # 保存到临时文件
@@ -470,64 +497,83 @@ class KnowledgeBaseRoute(Route):
                     # 异步读取文件内容
                     async with aiofiles.open(temp_file_path, "rb") as f:
                         file_content = await f.read()
+
+                    # 提取文件类型
+                    file_type = (
+                        file_name.rsplit(".", 1)[-1].lower() if "." in file_name else ""
+                    )
+
+                    files_to_upload.append(
+                        {
+                            "file_name": file_name,
+                            "file_content": file_content,
+                            "file_type": file_type,
+                        }
+                    )
                 finally:
                     # 清理临时文件
                     if os.path.exists(temp_file_path):
                         os.remove(temp_file_path)
 
-            else:
-                # 方式 2: JSON base64
-                import base64
-
-                data = await request.json
-
-                kb_id = data.get("kb_id")
-                file_name = data.get("file_name")
-                file_content_b64 = data.get("file_content")
-
-                if not kb_id or not file_name or not file_content_b64:
-                    return (
-                        Response()
-                        .error("缺少参数 kb_id, file_name 或 file_content")
-                        .__dict__
-                    )
-
-                try:
-                    file_content = base64.b64decode(file_content_b64)
-                except Exception:
-                    return (
-                        Response()
-                        .error("file_content 必须是有效的 base64 编码")
-                        .__dict__
-                    )
-
-            # 提取文件类型
-            file_type = file_name.rsplit(".", 1)[-1].lower() if "." in file_name else ""
-
+            # 获取知识库
             kb_helper = await kb_manager.get_kb(kb_id)
             if not kb_helper:
                 return Response().error("知识库不存在").__dict__
 
-            # 上传文档
-            doc = await kb_helper.upload_document(
-                file_name=file_name,
-                file_content=file_content,
-                file_type=file_type,
-            )
+            # 上传所有文档
+            uploaded_docs = []
+            failed_docs = []
 
-            doc_dict = {
-                "doc_id": doc.doc_id,
-                "kb_id": doc.kb_id,
-                "doc_name": doc.doc_name,
-                "file_type": doc.file_type,
-                "file_size": doc.file_size,
-                "chunk_count": doc.chunk_count,
-                "media_count": doc.media_count,
-                "created_at": doc.created_at.isoformat(),
-                "updated_at": doc.updated_at.isoformat(),
+            for file_info in files_to_upload:
+                try:
+                    doc = await kb_helper.upload_document(
+                        file_name=file_info["file_name"],
+                        file_content=file_info["file_content"],
+                        file_type=file_info["file_type"],
+                        chunk_size=chunk_size,
+                        chunk_overlap=chunk_overlap,
+                        batch_size=batch_size,
+                        tasks_limit=tasks_limit,
+                        max_retries=max_retries,
+                    )
+
+                    doc_dict = {
+                        "doc_id": doc.doc_id,
+                        "kb_id": doc.kb_id,
+                        "doc_name": doc.doc_name,
+                        "file_type": doc.file_type,
+                        "file_size": doc.file_size,
+                        "chunk_count": doc.chunk_count,
+                        "media_count": doc.media_count,
+                        "created_at": doc.created_at.isoformat(),
+                        "updated_at": doc.updated_at.isoformat(),
+                    }
+                    uploaded_docs.append(doc_dict)
+                except Exception as e:
+                    logger.error(f"上传文档 {file_info['file_name']} 失败: {e}")
+                    failed_docs.append(
+                        {"file_name": file_info["file_name"], "error": str(e)}
+                    )
+
+            # 返回结果
+            result = {
+                "uploaded": uploaded_docs,
+                "failed": failed_docs,
+                "total": len(files_to_upload),
+                "success_count": len(uploaded_docs),
+                "failed_count": len(failed_docs),
             }
 
-            return Response().ok(doc_dict, "上传文档成功").__dict__
+            if failed_docs:
+                message = (
+                    f"部分文档上传成功 ({len(uploaded_docs)}/{len(files_to_upload)})"
+                )
+            else:
+                message = (
+                    f"所有文档上传成功 ({len(uploaded_docs)}/{len(files_to_upload)})"
+                )
+
+            return Response().ok(result, message).__dict__
 
         except ValueError as e:
             return Response().error(str(e)).__dict__

dashboard/src/i18n/locales/en-US/features/knowledge-base/detail.json

@@ -46,12 +46,19 @@
     "selectFile": "Select File",
     "dropzone": "Drop files here or click to select",
     "supportedFormats": "Supported formats: TXT, PDF, Markdown",
-    "maxSize": "Max file size: 50MB",
+    "maxSize": "Max file size: 128MB",
     "chunkSettings": "Chunk Settings",
+    "batchSettings": "Batch Settings",
     "chunkSize": "Chunk Size",
     "chunkSizeHint": "Number of characters per chunk (default: 512)",
     "chunkOverlap": "Chunk Overlap",
     "chunkOverlapHint": "Overlapping characters between chunks (default: 50)",
+    "batchSize": "Batch Size",
+    "batchSizeHint": "Number of chunks to process in each batch (default: 32)",
+    "tasksLimit": "Concurrent Tasks Limit",
+    "tasksLimitHint": "Maximum number of concurrent upload tasks (default: 3)",
+    "maxRetries": "Max Retries",
+    "maxRetriesHint": "Number of times to retry a failed upload task (default: 3)",
     "cancel": "Cancel",
     "submit": "Upload",
     "fileRequired": "Please select a file to upload"

dashboard/src/i18n/locales/zh-CN/features/knowledge-base/detail.json

@@ -47,12 +47,19 @@
     "selectFile": "选择文件",
     "dropzone": "拖放文件到这里或点击选择",
     "supportedFormats": "支持的格式: TXT, PDF, Markdown",
-    "maxSize": "最大文件大小: 50MB",
+    "maxSize": "最大文件大小: 128MB",
     "chunkSettings": "分块设置",
+    "batchSettings": "批处理设置",
     "chunkSize": "分块大小",
     "chunkSizeHint": "每个文本块的字符数 (默认: 512)",
     "chunkOverlap": "分块重叠",
     "chunkOverlapHint": "相邻文本块之间的重叠字符数 (默认: 50)",
+    "batchSize": "批处理大小",
+    "batchSizeHint": "每批处理的文本块数量 (默认: 32)",
+    "tasksLimit": "并发任务限制",
+    "tasksLimitHint": "最大并发上传任务数 (默认: 3)",
+    "maxRetries": "最大重试次数",
+    "maxRetriesHint": "上传失败任务的重试次数 (默认: 3)",
     "cancel": "取消",
     "submit": "上传",
     "fileRequired": "请选择要上传的文件"

dashboard/src/views/knowledge-base/components/DocumentsTab.vue

@@ -2,35 +2,16 @@
   <div class="documents-tab">
     <!-- 操作栏 -->
     <div class="action-bar mb-4">
-      <v-btn
-        prepend-icon="mdi-upload"
-        color="primary"
-        variant="elevated"
-        @click="showUploadDialog = true"
-      >
+      <v-btn prepend-icon="mdi-upload" color="primary" variant="elevated" @click="showUploadDialog = true">
         {{ t('documents.upload') }}
       </v-btn>
-      <v-text-field
-        v-model="searchQuery"
-        prepend-inner-icon="mdi-magnify"
-        :placeholder="'搜索文档...'"
-        variant="outlined"
-        density="compact"
-        hide-details
-        clearable
-        style="max-width: 300px"
-      />
+      <v-text-field v-model="searchQuery" prepend-inner-icon="mdi-magnify" :placeholder="'搜索文档...'" variant="outlined"
+        density="compact" hide-details clearable style="max-width: 300px" />
     </div>
 
     <!-- 文档列表 -->
     <v-card elevation="2">
-      <v-data-table
-        :headers="headers"
-        :items="documents"
-        :loading="loading"
-        :search="searchQuery"
-        :items-per-page="10"
-      >
+      <v-data-table :headers="headers" :items="documents" :loading="loading" :search="searchQuery" :items-per-page="10">
         <template #item.doc_name="{ item }">
           <div class="d-flex align-center gap-2">
             <v-icon :color="getFileColor(item.file_type)">
@@ -49,20 +30,8 @@
         </template>
 
         <template #item.actions="{ item }">
-          <v-btn
-            icon="mdi-eye"
-            variant="text"
-            size="small"
-            color="info"
-            @click="viewDocument(item)"
-          />
-          <v-btn
-            icon="mdi-delete"
-            variant="text"
-            size="small"
-            color="error"
-            @click="confirmDelete(item)"
-          />
+          <v-btn icon="mdi-eye" variant="text" size="small" color="info" @click="viewDocument(item)" />
+          <v-btn icon="mdi-delete" variant="text" size="small" color="error" @click="confirmDelete(item)" />
         </template>
 
         <template #no-data>
@@ -75,9 +44,9 @@
     </v-card>
 
     <!-- 上传对话框 -->
-    <v-dialog v-model="showUploadDialog" max-width="600px" persistent>
+    <v-dialog v-model="showUploadDialog" max-width="600px" persistent @after-enter="initUploadSettings">
       <v-card>
-        <v-card-title class="pa-4">
+        <v-card-title class="pa-4 d-flex align-center">
           <span class="text-h5">{{ t('upload.title') }}</span>
           <v-spacer />
           <v-btn icon="mdi-close" variant="text" @click="closeUploadDialog" />
@@ -87,68 +56,74 @@
 
         <v-card-text class="pa-6">
           <!-- 文件选择 -->
-          <div
-            class="upload-dropzone"
-            :class="{ 'dragover': isDragging }"
-            @drop.prevent="handleDrop"
-            @dragover.prevent="isDragging = true"
-            @dragleave="isDragging = false"
-            @click="$refs.fileInput.click()"
-          >
+          <div class="upload-dropzone" :class="{ 'dragover': isDragging }" @drop.prevent="handleDrop"
+            @dragover.prevent="isDragging = true" @dragleave="isDragging = false" @click="fileInput?.click()">
             <v-icon size="64" color="primary">mdi-cloud-upload</v-icon>
             <p class="mt-4 text-h6">{{ t('upload.dropzone') }}</p>
             <p class="text-caption text-medium-emphasis mt-2">{{ t('upload.supportedFormats') }}</p>
             <p class="text-caption text-medium-emphasis">{{ t('upload.maxSize') }}</p>
-            <input
-              ref="fileInput"
-              type="file"
-              hidden
-              accept=".txt,.md,.pdf"
-              @change="handleFileSelect"
-            />
+            <p class="text-caption text-medium-emphasis">最多可上传 10 个文件</p>
+            <input ref="fileInput" type="file" multiple hidden accept=".txt,.md,.pdf" @change="handleFileSelect" />
           </div>
 
-          <div v-if="selectedFile" class="mt-4 pa-4 rounded bg-surface-variant">
-            <div class="d-flex align-center justify-space-between">
-              <div class="d-flex align-center gap-2">
-                <v-icon>{{ getFileIcon(selectedFile.name) }}</v-icon>
-                <div>
-                  <div class="font-weight-medium">{{ selectedFile.name }}</div>
-                  <div class="text-caption">{{ formatFileSize(selectedFile.size) }}</div>
+          <div v-if="selectedFiles.length > 0" class="mt-4">
+            <div class="d-flex align-center justify-space-between mb-2">
+              <span class="text-subtitle-2">已选择 {{ selectedFiles.length }} 个文件</span>
+              <v-btn variant="text" size="small" @click="selectedFiles = []">清空</v-btn>
+            </div>
+            <div class="files-list">
+              <div v-for="(file, index) in selectedFiles" :key="index" class="file-item pa-3 mb-2 rounded bg-surface-variant">
+                <div class="d-flex align-center justify-space-between">
+                  <div class="d-flex align-center gap-2">
+                    <v-icon>{{ getFileIcon(file.name) }}</v-icon>
+                    <div>
+                      <div class="font-weight-medium">{{ file.name }}</div>
+                      <div class="text-caption">{{ formatFileSize(file.size) }}</div>
+                    </div>
+                  </div>
+                  <v-btn icon="mdi-close" variant="text" size="small" @click="removeFile(index)" />
                 </div>
               </div>
-              <v-btn icon="mdi-close" variant="text" size="small" @click="selectedFile = null" />
             </div>
           </div>
 
           <!-- 分块设置 -->
-          <v-expansion-panels class="mt-4">
-            <v-expansion-panel>
-              <v-expansion-panel-title>
-                <v-icon start>mdi-cog</v-icon>
-                {{ t('upload.chunkSettings') }}
-              </v-expansion-panel-title>
-              <v-expansion-panel-text>
-                <v-text-field
-                  v-model.number="uploadSettings.chunk_size"
-                  :label="t('upload.chunkSize')"
-                  :hint="t('upload.chunkSizeHint')"
-                  type="number"
-                  variant="outlined"
-                  density="compact"
-                  class="mb-2"
-                />
-                <v-text-field
-                  v-model.number="uploadSettings.chunk_overlap"
-                  :label="t('upload.chunkOverlap')"
-                  :hint="t('upload.chunkOverlapHint')"
-                  type="number"
-                  variant="outlined"
-                  density="compact"
-                />
-              </v-expansion-panel-text>
-            </v-expansion-panel>
-          </v-expansion-panels>
+          <div class="mt-6">
+            <div class="d-flex align-center mb-4">
+              <h3 class="text-h6">{{ t('upload.chunkSettings') }}</h3>
+            </div>
+            <v-row>
+              <v-col cols="12" sm="6">
+                <v-text-field v-model.number="uploadSettings.chunk_size" :label="t('upload.chunkSize')"
+                  :hint="t('upload.chunkSizeHint')" persistent-hint type="number" variant="outlined"
+                  density="compact" :placeholder="props.kb?.chunk_size?.toString() || '512'" />
+              </v-col>
+              <v-col cols="12" sm="6">
+                <v-text-field v-model.number="uploadSettings.chunk_overlap" :label="t('upload.chunkOverlap')"
+                  :hint="t('upload.chunkOverlapHint')" persistent-hint type="number" variant="outlined"
+                  density="compact" :placeholder="props.kb?.chunk_overlap?.toString() || '50'" />
+              </v-col>
+            </v-row>
+          </div>
+
+          <div class="mt-2">
+            <h3 class="text-h6 mb-4">{{ t('upload.batchSettings') }}</h3>
+            <v-row>
+              <v-col cols="12" sm="4">
+                <v-text-field v-model.number="uploadSettings.batch_size" :label="t('upload.batchSize')"
+                  hint="每批处理的文本数量" persistent-hint type="number" variant="outlined" density="compact" />
+              </v-col>
+              <v-col cols="12" sm="4">
+                <v-text-field v-model.number="uploadSettings.tasks_limit" :label="t('upload.tasksLimit')"
+                  hint="并发任务数量限制" persistent-hint type="number" variant="outlined" density="compact" />
+              </v-col>
+              <v-col cols="12" sm="4">
+                <v-text-field v-model.number="uploadSettings.max_retries" :label="t('upload.maxRetries')"
+                  hint="失败时的最大重试次数" persistent-hint type="number" variant="outlined" density="compact" />
+              </v-col>
+            </v-row>
+          </div>
+
         </v-card-text>
 
         <v-divider />
@@ -158,13 +133,8 @@
           <v-btn variant="text" @click="closeUploadDialog">
             {{ t('upload.cancel') }}
           </v-btn>
-          <v-btn
-            color="primary"
-            variant="elevated"
-            @click="uploadDocument"
-            :loading="uploading"
-            :disabled="!selectedFile"
-          >
+          <v-btn color="primary" variant="elevated" @click="uploadDocument" :loading="uploading"
+            :disabled="selectedFiles.length === 0">
             {{ t('upload.submit') }}
           </v-btn>
         </v-card-actions>
@@ -224,9 +194,10 @@ const documents = ref<any[]>([])
 const searchQuery = ref('')
 const showUploadDialog = ref(false)
 const showDeleteDialog = ref(false)
-const selectedFile = ref<File | null>(null)
+const selectedFiles = ref<File[]>([])
 const deleteTarget = ref<any>(null)
 const isDragging = ref(false)
+const fileInput = ref<HTMLInputElement | null>(null)
 
 const snackbar = ref({
   show: false,
@@ -242,10 +213,24 @@ const showSnackbar = (text: string, color: string = 'success') => {
 
 // 上传设置
 const uploadSettings = ref({
-  chunk_size: null,
-  chunk_overlap: null
+  chunk_size: null as number | null,
+  chunk_overlap: null as number | null,
+  batch_size: 32,
+  tasks_limit: 3,
+  max_retries: 3
 })
 
+// 初始化上传设置
+const initUploadSettings = () => {
+  uploadSettings.value = {
+    chunk_size: props.kb?.chunk_size || null,
+    chunk_overlap: props.kb?.chunk_overlap || null,
+    batch_size: 32,
+    tasks_limit: 3,
+    max_retries: 3
+  }
+}
+
 // 表格列
 const headers = [
   { title: t('documents.name'), key: 'doc_name', sortable: true },
@@ -253,7 +238,7 @@ const headers = [
   { title: t('documents.size'), key: 'file_size', sortable: true },
   { title: t('documents.chunks'), key: 'chunk_count', sortable: true },
   { title: t('documents.createdAt'), key: 'created_at', sortable: true },
-  { title: t('documents.actions'), key: 'actions', sortable: false, align: 'end' }
+  { title: t('documents.actions'), key: 'actions', sortable: false, align: 'end' as const }
 ]
 
 // 加载文档列表
@@ -277,22 +262,39 @@ const loadDocuments = async () => {
 // 文件选择
 const handleFileSelect = (event: Event) => {
   const target = event.target as HTMLInputElement
-  if (target.files && target.files[0]) {
-    selectedFile.value = target.files[0]
+  if (target.files && target.files.length > 0) {
+    const newFiles = Array.from(target.files)
+    addFiles(newFiles)
   }
 }
 
+// 添加文件（检查数量限制）
+const addFiles = (files: File[]) => {
+  const totalFiles = selectedFiles.value.length + files.length
+  if (totalFiles > 10) {
+    showSnackbar('最多只能选择 10 个文件', 'warning')
+    return
+  }
+  selectedFiles.value.push(...files)
+}
+
+// 移除文件
+const removeFile = (index: number) => {
+  selectedFiles.value.splice(index, 1)
+}
+
 // 拖放上传
 const handleDrop = (event: DragEvent) => {
   isDragging.value = false
-  if (event.dataTransfer?.files && event.dataTransfer.files[0]) {
-    selectedFile.value = event.dataTransfer.files[0]
+  if (event.dataTransfer?.files && event.dataTransfer.files.length > 0) {
+    const newFiles = Array.from(event.dataTransfer.files)
+    addFiles(newFiles)
   }
 }
 
 // 上传文档
 const uploadDocument = async () => {
-  if (!selectedFile.value) {
+  if (selectedFiles.value.length === 0) {
     showSnackbar(t('upload.fileRequired'), 'warning')
     return
   }
@@ -300,7 +302,12 @@ const uploadDocument = async () => {
   uploading.value = true
   try {
     const formData = new FormData()
-    formData.append('file', selectedFile.value)
+    
+    // 添加所有文件
+    selectedFiles.value.forEach((file, index) => {
+      formData.append(`file${index}`, file)
+    })
+    
     formData.append('kb_id', props.kbId)
     if (uploadSettings.value.chunk_size) {
       formData.append('chunk_size', uploadSettings.value.chunk_size.toString())
@@ -308,13 +315,25 @@ const uploadDocument = async () => {
     if (uploadSettings.value.chunk_overlap) {
       formData.append('chunk_overlap', uploadSettings.value.chunk_overlap.toString())
     }
+    formData.append('batch_size', uploadSettings.value.batch_size.toString())
+    formData.append('tasks_limit', uploadSettings.value.tasks_limit.toString())
+    formData.append('max_retries', uploadSettings.value.max_retries.toString())
 
     const response = await axios.post('/api/kb/document/upload', formData, {
       headers: { 'Content-Type': 'multipart/form-data' }
     })
 
     if (response.data.status === 'ok') {
-      showSnackbar(t('documents.uploadSuccess'))
+      const result = response.data.data
+      const successCount = result.success_count || 0
+      const failedCount = result.failed_count || 0
+      
+      if (failedCount === 0) {
+        showSnackbar(`成功上传 ${successCount} 个文档`)
+      } else {
+        showSnackbar(`上传完成: ${successCount} 个成功, ${failedCount} 个失败`, 'warning')
+      }
+      
       closeUploadDialog()
       await loadDocuments()
       emit('refresh')
@@ -332,8 +351,9 @@ const uploadDocument = async () => {
 // 关闭上传对话框
 const closeUploadDialog = () => {
   showUploadDialog.value = false
-  selectedFile.value = null
-  uploadSettings.value = { chunk_size: null, chunk_overlap: null }
+  selectedFiles.value = []
+  // 重置为知识库默认设置
+  initUploadSettings()
 }
 
 // 查看文档
@@ -428,8 +448,13 @@ onMounted(() => {
 }
 
 @keyframes fadeIn {
-  from { opacity: 0; }
-  to { opacity: 1; }
+  from {
+    opacity: 0;
+  }
+
+  to {
+    opacity: 1;
+  }
 }
 
 .action-bar {
@@ -457,13 +482,26 @@ onMounted(() => {
   transform: scale(1.02);
 }
 
+.files-list {
+  max-height: 300px;
+  overflow-y: auto;
+}
+
+.file-item {
+  transition: all 0.2s ease;
+}
+
+.file-item:hover {
+  background: rgba(var(--v-theme-surface-variant), 0.8) !important;
+}
+
 @media (max-width: 768px) {
   .action-bar {
     flex-direction: column;
     align-items: stretch;
   }
 
-  .action-bar > * {
+  .action-bar>* {
     width: 100%;
   }
 }