feat: 完善重新分块功能，优化代码逻辑，保存请求参数

Author: supreme0597

src/knowledge/base.py

@@ -293,6 +293,7 @@ def get_database_info(self, db_id: str) -> dict | None:
                     "type": file_info.get("file_type", ""),
                     "status": file_info.get("status", "done"),
                     "created_at": created_at,
+                    "processing_params": file_info.get("processing_params", None),
                 }
 
         # 按创建时间倒序排序文件列表

src/knowledge/implementations/chroma.py

@@ -47,10 +47,6 @@ def __init__(self, work_dir: str, **kwargs):
 
         # 存储集合映射 {db_id: collection}
         self.collections: dict[str, Any] = {}
-
-        # 元数据锁
-        self._metadata_lock = asyncio.Lock()
-
         logger.info("ChromaKB initialized")
 
     @property
@@ -290,6 +286,7 @@ async def update_content(self, db_id: str, file_ids: list[str], params: dict | N
 
             try:
                 # 更新状态为处理中
+                self.files_meta[file_id]["processing_params"] = params.copy()
                 self.files_meta[file_id]["status"] = "processing"
                 self._save_metadata()
 
@@ -361,24 +358,6 @@ async def update_content(self, db_id: str, file_ids: list[str], params: dict | N
 
         return processed_items_info
 
-    async def delete_file_chunks_only(self, db_id: str, file_id: str) -> None:
-        """仅删除文件的chunks数据，保留元数据（用于更新操作）"""
-        collection = await self._get_chroma_collection(db_id)
-
-        if collection:
-            try:
-                # 查找所有相关的chunks
-                results = collection.get(where={"full_doc_id": file_id}, include=["metadatas"])
-
-                # 删除所有相关chunks
-                if results and results.get("ids"):
-                    collection.delete(ids=results["ids"])
-                    logger.info(f"Deleted {len(results['ids'])} chunks for file {file_id}")
-
-            except Exception as e:
-                logger.error(f"Error deleting file {file_id} from ChromaDB: {e}")
-        # 注意：这里不删除 files_meta[file_id]，保留元数据用于后续操作
-
     async def aquery(self, query_text: str, db_id: str, **kwargs) -> list[dict]:
         """异步查询知识库"""
         collection = await self._get_chroma_collection(db_id)
@@ -473,16 +452,32 @@ async def aquery(self, query_text: str, db_id: str, **kwargs) -> list[dict]:
             logger.error(f"ChromaDB query error: {e}, {traceback.format_exc()}")
             return []
 
+    async def delete_file_chunks_only(self, db_id: str, file_id: str) -> None:
+        """仅删除文件的chunks数据，保留元数据（用于更新操作）"""
+        collection = await self._get_chroma_collection(db_id)
+        if collection:
+            try:
+                # 查找所有相关的chunks
+                results = collection.get(where={"full_doc_id": file_id}, include=["metadatas"])
+
+                # 删除所有相关chunks
+                if results and results.get("ids"):
+                    collection.delete(ids=results["ids"])
+                    logger.info(f"Deleted {len(results['ids'])} chunks for file {file_id}")
+
+            except Exception as e:
+                logger.error(f"Error deleting file {file_id} from ChromaDB: {e}")
+        # 注意：这里不删除 files_meta[file_id]，保留元数据用于后续操作
+
     async def delete_file(self, db_id: str, file_id: str) -> None:
         """删除文件（包括元数据）"""
         # 先删除 ChromaDB 中的 chunks 数据
         await self.delete_file_chunks_only(db_id, file_id)
 
-        # 使用锁确保元数据操作的原子性
-        async with self._metadata_lock:
-            if file_id in self.files_meta:
-                del self.files_meta[file_id]
-                self._save_metadata()
+        # 删除文件记录
+        if file_id in self.files_meta:
+            del self.files_meta[file_id]
+            self._save_metadata()
 
     async def get_file_basic_info(self, db_id: str, file_id: str) -> dict:
         """获取文件基本信息（仅元数据）"""

src/knowledge/implementations/lightrag.py

@@ -32,9 +32,6 @@ def __init__(self, work_dir: str, **kwargs):
         # 存储 LightRAG 实例映射 {db_id: LightRAG}
         self.instances: dict[str, LightRAG] = {}
 
-        # 元数据锁
-        self._metadata_lock = asyncio.Lock()
-
         # 设置 LightRAG 日志
         log_dir = os.path.join(work_dir, "logs", "lightrag")
         os.makedirs(log_dir, exist_ok=True)
@@ -306,6 +303,7 @@ async def update_content(self, db_id: str, file_ids: list[str], params: dict | N
 
             try:
                 # 更新状态为处理中
+                self.files_meta[file_id]["processing_params"] = params.copy()
                 self.files_meta[file_id]["status"] = "processing"
                 self._save_metadata()
 
@@ -357,19 +355,6 @@ async def update_content(self, db_id: str, file_ids: list[str], params: dict | N
 
         return processed_items_info
 
-    async def delete_file_chunks_only(self, db_id: str, file_id: str) -> None:
-        """仅删除文件的chunks数据，保留元数据（用于更新操作）"""
-        rag = await self._get_lightrag_instance(db_id)
-
-        if rag:
-            try:
-                # 使用 LightRAG 删除文档
-                await rag.adelete_by_doc_id(file_id)
-                logger.info(f"Deleted chunks for file {file_id} from LightRAG")
-            except Exception as e:
-                logger.error(f"Error deleting file {file_id} from LightRAG: {e}")
-        # 注意：这里不删除 files_meta[file_id]，保留元数据用于后续操作
-
     async def aquery(self, query_text: str, db_id: str, **kwargs) -> str:
         """异步查询知识库"""
         rag = await self._get_lightrag_instance(db_id)
@@ -395,16 +380,27 @@ async def aquery(self, query_text: str, db_id: str, **kwargs) -> str:
             logger.error(f"Query error: {e}, {traceback.format_exc()}")
             return ""
 
+    async def delete_file_chunks_only(self, db_id: str, file_id: str) -> None:
+        """仅删除文件的chunks数据，保留元数据（用于更新操作）"""
+        rag = await self._get_lightrag_instance(db_id)
+        if rag:
+            try:
+                # 使用 LightRAG 删除文档
+                await rag.adelete_by_doc_id(file_id)
+                logger.info(f"Deleted chunks for file {file_id} from LightRAG")
+            except Exception as e:
+                logger.error(f"Error deleting file {file_id} from LightRAG: {e}")
+        # 注意：这里不删除 files_meta[file_id]，保留元数据用于后续操作
+
     async def delete_file(self, db_id: str, file_id: str) -> None:
         """删除文件（包括元数据）"""
         # 先删除 LightRAG 中的 chunks 数据
         await self.delete_file_chunks_only(db_id, file_id)
 
-        # 使用锁确保元数据操作的原子性
-        async with self._metadata_lock:
-            if file_id in self.files_meta:
-                del self.files_meta[file_id]
-                self._save_metadata()
+        # 删除文件记录
+        if file_id in self.files_meta:
+            del self.files_meta[file_id]
+            self._save_metadata()
 
     async def get_file_basic_info(self, db_id: str, file_id: str) -> dict:
         """获取文件基本信息（仅元数据）"""

src/knowledge/implementations/milvus.py

@@ -332,6 +332,7 @@ async def update_content(self, db_id: str, file_ids: list[str], params: dict | N
             try:
                 # 更新状态为处理中
                 async with self._metadata_lock:
+                    self.files_meta[file_id]["processing_params"] = params.copy()
                     self.files_meta[file_id]["status"] = "processing"
                     self._save_metadata()
 

web/src/components/FileTable.vue

@@ -203,7 +203,8 @@ const selectedRowKeys = computed({
 
 // 重新分块参数配置相关
 const rechunkModalVisible = ref(false);
-const rechunkModalLoading = ref(false);
+const rechunkModalLoading = computed(() => store.state.chunkLoading);
+
 const rechunkParams = ref({
   chunk_size: 1000,
   chunk_overlap: 200,
@@ -448,18 +449,17 @@ const handleDownloadFile = async (record) => {
 };
 
 const handleRechunkFile = async (record) => {
-  const dbId = store.databaseId;
-  if (!dbId) {
-    console.error('无法获取数据库ID，数据库ID:', store.databaseId, '记录:', record);
-    message.error('无法获取数据库ID，请刷新页面后重试');
-    return;
-  }
-
   try {
     // 设置当前重新分块的文件ID
     currentRechunkFileIds.value = [record.file_id];
     isBatchRechunk.value = false;
 
+    if (record?.processing_params) {
+      rechunkParams.value = {
+        ...record?.processing_params
+      };
+    }
+
     // 显示参数配置模态框
     rechunkModalVisible.value = true;
   } catch (error) {
@@ -470,50 +470,39 @@ const handleRechunkFile = async (record) => {
 
 // 重新分块确认
 const handleRechunkConfirm = async () => {
-  const dbId = store.databaseId;
-  if (!dbId) {
-    console.error('无法获取数据库ID，数据库ID:', store.databaseId);
-    message.error('无法获取数据库ID，请刷新页面后重试');
-    return;
-  }
-
-  if (currentRechunkFileIds.value.length === 0) {
-    message.warning('请选择要重新分块的文件');
-    return;
-  }
-
-  rechunkModalLoading.value = true;
-
   try {
     // 调用 rechunks 接口
-    const result = await documentApi.rechunksDocuments(dbId, currentRechunkFileIds.value, rechunkParams.value);
-
-    if (result.status === 'queued') {
-      message.success('重新分块任务已提交，请在任务中心查看进度');
-      // 刷新文件列表
-      store.getDatabaseInfo(undefined, true);
+    const result = await store.rechunksFiles({fileIds: currentRechunkFileIds.value, params: rechunkParams.value});
+    if (result) {
+      currentRechunkFileIds.value = [];
       // 清空选择
       if (isBatchRechunk.value) {
         selectedRowKeys.value = [];
       }
       // 关闭模态框
       rechunkModalVisible.value = false;
+
+      // 重置参数为默认值
+      rechunkParams.value = {
+        chunk_size: 1000,
+        chunk_overlap: 200,
+        use_qa_split: false,
+        qa_separator: '\n\n\n'
+      };
     } else {
       message.error(`重新分块失败: ${result.message}`);
     }
   } catch (error) {
     console.error('重新分块失败:', error);
     const errorMessage = error.message || '重新分块失败，请稍后重试';
     message.error(errorMessage);
-  } finally {
-    rechunkModalLoading.value = false;
   }
 };
 
 // 重新分块取消
 const handleRechunkCancel = () => {
   rechunkModalVisible.value = false;
-  rechunkModalLoading.value = false;
+  // rechunkModalLoading.value = false;
   currentRechunkFileIds.value = [];
   isBatchRechunk.value = false;
   // 重置参数为默认值

web/src/stores/database.js

@@ -256,6 +256,45 @@ export const useDatabaseStore = defineStore('database', () => {
     }
   }
 
+  async function rechunksFiles({ fileIds, params }) {
+    if (fileIds.length === 0) {
+      message.error('请选择要重新分块的文件！');
+      return;
+    }
+
+    state.chunkLoading = true;
+    try {
+      const data = await documentApi.rechunksDocuments(databaseId.value, fileIds, { ...params });
+      if (data.status === 'success' || data.status === 'queued') {
+        enableAutoRefresh('auto');
+        message.success(data.message || `文档已提交处理，请在任务中心查看进度`);
+        if (data.task_id) {
+          taskerStore.registerQueuedTask({
+            task_id: data.task_id,
+            name: `文档重新分块 (${databaseId.value || ''})`,
+            task_type: 'knowledge_rechunks',
+            message: data.message,
+            payload: {
+              db_id: databaseId.value,
+              count: fileIds.length,
+            }
+          });
+        }
+        await getDatabaseInfo(undefined, true); // Skip query params when adding files
+        return true; // Indicate success
+      } else {
+        message.error(data.message || '处理失败');
+        return false;
+      }
+    } catch (error) {
+      console.error(error);
+      message.error(error.message || '处理请求失败');
+      return false;
+    } finally {
+      state.chunkLoading = false;
+    }
+  }
+
   async function openFileDetail(record) {
     if (record.status !== 'done') {
       message.error('文件未处理完成，请稍后再试');
@@ -379,6 +418,7 @@ export const useDatabaseStore = defineStore('database', () => {
     handleDeleteFile,
     handleBatchDelete,
     addFiles,
+    rechunksFiles,
     openFileDetail,
     loadQueryParams,