feat(annotation): 优化文本标注分段功能实现

- 新增 getEditorTaskSegmentsUsingGet 接口用于获取任务分段信息
- 移除 SegmentInfo 中的 text、start、end 字段，精简数据结构
- 添加 EditorTaskSegmentsResponse 类型定义用于分段摘要响应
- 实现服务端 get_task_segments 方法，支持分段信息查询
- 重构前端组件缓存机制，使用 segmentSummaryFileRef 管理分段状态
- 优化分段构建逻辑，提取 _build_segment_contexts 公共方法
- 调整后端 _build_text_task 方法中的分段处理流程
- 更新 API 类型定义，统一 RequestParams 和 RequestPayload 类型

backend/openapi/specs/data-management.yaml

@@ -470,6 +470,23 @@ paths:
         '200':
           description: 上传成功
 
+  /data-management/datasets/upload/cancel-upload/{reqId}:
+    put:
+      tags: [ DatasetFile ]
+      operationId: cancelUpload
+      summary: 取消上传
+      description: 取消预上传请求并清理临时分片
+      parameters:
+        - name: reqId
+          in: path
+          required: true
+          schema:
+            type: string
+          description: 预上传请求ID
+      responses:
+        '200':
+          description: 取消成功
+
   /data-management/dataset-types:
     get:
       operationId: getDatasetTypes

backend/services/data-management-service/src/main/java/com/datamate/datamanagement/application/DatasetApplicationService.java

@@ -1,5 +1,6 @@
 package com.datamate.datamanagement.application;
 
+import com.baomidou.mybatisplus.core.conditions.update.LambdaUpdateWrapper;
 import com.baomidou.mybatisplus.core.metadata.IPage;
 import com.baomidou.mybatisplus.extension.plugins.pagination.Page;
 import com.datamate.common.domain.utils.ChunksSaver;
@@ -101,6 +102,7 @@ public class DatasetApplicationService {
     public Dataset updateDataset(String datasetId, UpdateDatasetRequest updateDatasetRequest) {
         Dataset dataset = datasetRepository.getById(datasetId);
         BusinessAssert.notNull(dataset, DataManagementErrorCode.DATASET_NOT_FOUND);
+        
         if (StringUtils.hasText(updateDatasetRequest.getName())) {
             dataset.setName(updateDatasetRequest.getName());
         }
@@ -113,14 +115,31 @@ public class DatasetApplicationService {
         if (Objects.nonNull(updateDatasetRequest.getStatus())) {
             dataset.setStatus(updateDatasetRequest.getStatus());
         }
-        // 处理父数据集变更：始终调用 handleParentChange，以支持设置新的关联或清除关联
+        if (updateDatasetRequest.isParentDatasetIdProvided()) {
+            // 保存原始的 parentDatasetId 值，用于比较是否发生了变化
+            String originalParentDatasetId = dataset.getParentDatasetId();
+
+            // 处理父数据集变更：仅当请求显式包含 parentDatasetId 时处理
             // handleParentChange 内部通过 normalizeParentId 方法将空字符串和 null 都转换为 null
             // 这样既支持设置新的父数据集，也支持清除关联
             handleParentChange(dataset, updateDatasetRequest.getParentDatasetId());
+
+            // 检查 parentDatasetId 是否发生了变化
+            if (!Objects.equals(originalParentDatasetId, dataset.getParentDatasetId())) {
+                // 使用 LambdaUpdateWrapper 显式地更新 parentDatasetId 字段
+                // 这样即使值为 null 也能被正确更新到数据库
+                datasetRepository.update(null, new LambdaUpdateWrapper<Dataset>()
+                    .eq(Dataset::getId, datasetId)
+                    .set(Dataset::getParentDatasetId, dataset.getParentDatasetId()));
+            }
+        }
+        
         if (StringUtils.hasText(updateDatasetRequest.getDataSource())) {
             // 数据源id不为空，使用异步线程进行文件扫盘落库
             processDataSourceAsync(dataset.getId(), updateDatasetRequest.getDataSource());
         }
+        
+        // 更新其他字段（不包括 parentDatasetId，因为它已经在上面的代码中更新了）
         datasetRepository.updateById(dataset);
         return dataset;
     }

backend/services/data-management-service/src/main/java/com/datamate/datamanagement/application/DatasetFileApplicationService.java

@@ -505,6 +505,14 @@ public class DatasetFileApplicationService {
         saveFileInfoToDb(uploadResult, datasetId);
     }
 
+    /**
+     * 取消上传
+     */
+    @Transactional
+    public void cancelUpload(String reqId) {
+        fileService.cancelUpload(reqId);
+    }
+
     private void saveFileInfoToDb(FileUploadResult fileUploadResult, String datasetId) {
         if (Objects.isNull(fileUploadResult.getSavedFile())) {
             // 文件切片上传没有完成

backend/services/data-management-service/src/main/java/com/datamate/datamanagement/domain/model/dataset/Dataset.java

@@ -1,6 +1,5 @@
 package com.datamate.datamanagement.domain.model.dataset;
 
-import com.baomidou.mybatisplus.annotation.FieldStrategy;
 import com.baomidou.mybatisplus.annotation.TableField;
 import com.baomidou.mybatisplus.annotation.TableName;
 import com.baomidou.mybatisplus.extension.handlers.JacksonTypeHandler;
@@ -32,7 +31,6 @@ public class Dataset extends BaseEntity<String> {
     /**
      * 父数据集ID
      */
-    @TableField(updateStrategy = FieldStrategy.IGNORED)
     private String parentDatasetId;
     /**
      * 数据集类型

backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/dto/UpdateDatasetRequest.java

@@ -1,8 +1,10 @@
 package com.datamate.datamanagement.interfaces.dto;
 
 import com.datamate.datamanagement.common.enums.DatasetStatusType;
+import com.fasterxml.jackson.annotation.JsonIgnore;
 import jakarta.validation.constraints.NotBlank;
 import jakarta.validation.constraints.Size;
+import lombok.AccessLevel;
 import lombok.Getter;
 import lombok.Setter;
 
@@ -24,9 +26,18 @@ public class UpdateDatasetRequest {
     /** 归集任务id */
     private String dataSource;
     /** 父数据集ID */
+    @Setter(AccessLevel.NONE)
     private String parentDatasetId;
+    @JsonIgnore
+    @Setter(AccessLevel.NONE)
+    private boolean parentDatasetIdProvided;
     /** 标签列表 */
     private List<String> tags;
     /** 数据集状态 */
     private DatasetStatusType status;
+
+    public void setParentDatasetId(String parentDatasetId) {
+        this.parentDatasetIdProvided = true;
+        this.parentDatasetId = parentDatasetId;
+    }
 }

backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/rest/DatasetUploadController.java

@@ -0,0 +1,33 @@
+package com.datamate.datamanagement.interfaces.rest;
+
+import com.datamate.datamanagement.application.DatasetFileApplicationService;
+import lombok.RequiredArgsConstructor;
+import lombok.extern.slf4j.Slf4j;
+import org.springframework.http.ResponseEntity;
+import org.springframework.web.bind.annotation.PathVariable;
+import org.springframework.web.bind.annotation.PutMapping;
+import org.springframework.web.bind.annotation.RequestMapping;
+import org.springframework.web.bind.annotation.RestController;
+
+/**
+ * 数据集上传控制器
+ */
+@Slf4j
+@RestController
+@RequiredArgsConstructor
+@RequestMapping("/data-management/datasets/upload")
+public class DatasetUploadController {
+
+    private final DatasetFileApplicationService datasetFileApplicationService;
+
+    /**
+     * 取消上传
+     *
+     * @param reqId 预上传请求ID
+     */
+    @PutMapping("/cancel-upload/{reqId}")
+    public ResponseEntity<Void> cancelUpload(@PathVariable("reqId") String reqId) {
+        datasetFileApplicationService.cancelUpload(reqId);
+        return ResponseEntity.ok().build();
+    }
+}

backend/shared/domain-common/src/main/java/com/datamate/common/domain/service/FileService.java

@@ -74,6 +74,26 @@ public class FileService {
             .build();
     }
 
+    /**
+     * 取消上传
+     */
+    @Transactional
+    public void cancelUpload(String reqId) {
+        if (reqId == null || reqId.isBlank()) {
+            throw BusinessException.of(CommonErrorCode.PARAM_ERROR);
+        }
+        ChunkUploadPreRequest preRequest = chunkUploadRequestMapper.findById(reqId);
+        if (preRequest == null) {
+            return;
+        }
+        String uploadPath = preRequest.getUploadPath();
+        if (uploadPath != null && !uploadPath.isBlank()) {
+            File tempDir = new File(uploadPath, String.format(ChunksSaver.TEMP_DIR_NAME_FORMAT, preRequest.getId()));
+            ChunksSaver.deleteFolder(tempDir.getPath());
+        }
+        chunkUploadRequestMapper.deleteById(reqId);
+    }
+
     private File uploadFile(ChunkUploadRequest fileUploadRequest, ChunkUploadPreRequest preRequest) {
         File savedFile = ChunksSaver.saveFile(fileUploadRequest, preRequest);
         preRequest.setTimeout(LocalDateTime.now().plusSeconds(DEFAULT_TIMEOUT));

frontend/src/hooks/useSliceUpload.tsx

@@ -1,5 +1,5 @@
 import { TaskItem } from "@/pages/DataManagement/dataset.model";
-import { calculateSHA256, checkIsFilesExist } from "@/utils/file.util";
+import { calculateSHA256, checkIsFilesExist, streamSplitAndUpload, StreamUploadResult } from "@/utils/file.util";
 import { App } from "antd";
 import { useRef, useState } from "react";
 
@@ -9,17 +9,18 @@ export function useFileSliceUpload(
     uploadChunk,
     cancelUpload,
   }: {
-    preUpload: (id: string, params: any) => Promise<{ data: number }>;
+    preUpload: (id: string, params: Record<string, unknown>) => Promise<{ data: number }>;
-    uploadChunk: (id: string, formData: FormData, config: any) => Promise<any>;
+    uploadChunk: (id: string, formData: FormData, config: Record<string, unknown>) => Promise<unknown>;
-    cancelUpload: ((reqId: number) => Promise<any>) | null;
+    cancelUpload: ((reqId: number) => Promise<unknown>) | null;
   },
-  showTaskCenter = true // 上传时是否显示任务中心
+  showTaskCenter = true, // 上传时是否显示任务中心
+  enableStreamUpload = true // 是否启用流式分割上传
 ) {
   const { message } = App.useApp();
   const [taskList, setTaskList] = useState<TaskItem[]>([]);
   const taskListRef = useRef<TaskItem[]>([]); // 用于固定任务顺序
 
-  const createTask = (detail: any = {}) => {
+  const createTask = (detail: Record<string, unknown> = {}) => {
     const { dataset } = detail;
     const title = `上传数据集: ${dataset.name} `;
     const controller = new AbortController();
@@ -68,7 +69,7 @@ export function useFileSliceUpload(
       // 携带前缀信息，便于刷新后仍停留在当前目录
       window.dispatchEvent(
         new CustomEvent(task.updateEvent, {
-          detail: { prefix: (task as any).prefix },
+          detail: { prefix: task.prefix },
         })
       );
     }
@@ -79,7 +80,7 @@ export function useFileSliceUpload(
     }
   };
 
-  async function buildFormData({ file, reqId, i, j }) {
+  async function buildFormData({ file, reqId, i, j }: { file: { slices: Blob[]; name: string; size: number }; reqId: number; i: number; j: number }) {
     const formData = new FormData();
     const { slices, name, size } = file;
     const checkSum = await calculateSHA256(slices[j]);
@@ -94,12 +95,18 @@ export function useFileSliceUpload(
     return formData;
   }
 
-  async function uploadSlice(task: TaskItem, fileInfo) {
+  async function uploadSlice(task: TaskItem, fileInfo: { loaded: number; i: number; j: number; files: { slices: Blob[]; name: string; size: number }[]; totalSize: number }) {
     if (!task) {
       return;
     }
-    const { reqId, key } = task;
+    const { reqId, key, controller } = task;
     const { loaded, i, j, files, totalSize } = fileInfo;
+    
+    // 检查是否已取消
+    if (controller.signal.aborted) {
+      throw new Error("Upload cancelled");
+    }
+    
     const formData = await buildFormData({
       file: files[i],
       i,
@@ -109,6 +116,7 @@ export function useFileSliceUpload(
 
     let newTask = { ...task };
     await uploadChunk(key, formData, {
+      signal: controller.signal,
       onUploadProgress: (e) => {
         const loadedSize = loaded + e.loaded;
         const curPercent = Number((loadedSize / totalSize) * 100).toFixed(2);
@@ -124,7 +132,7 @@ export function useFileSliceUpload(
     });
   }
 
-  async function uploadFile({ task, files, totalSize }) {
+  async function uploadFile({ task, files, totalSize }: { task: TaskItem; files: { slices: Blob[]; name: string; size: number; originFile: Blob }[]; totalSize: number }) {
     console.log('[useSliceUpload] Calling preUpload with prefix:', task.prefix);
     const { data: reqId } = await preUpload(task.key, {
       totalFileNum: files.length,
@@ -140,9 +148,10 @@ export function useFileSliceUpload(
       reqId,
       isCancel: false,
       cancelFn: () => {
-        task.controller.abort();
+        // 使用 newTask 的 controller 确保一致性
+        newTask.controller.abort();
         cancelUpload?.(reqId);
-        if (task.updateEvent) window.dispatchEvent(new Event(task.updateEvent));
+        if (newTask.updateEvent) window.dispatchEvent(new Event(newTask.updateEvent));
       },
     };
     updateTaskList(newTask);
@@ -152,8 +161,16 @@ export function useFileSliceUpload(
 
     let loaded = 0;
     for (let i = 0; i < files.length; i++) {
+      // 检查是否已取消
+      if (newTask.controller.signal.aborted) {
+        throw new Error("Upload cancelled");
+      }
       const { slices } = files[i];
       for (let j = 0; j < slices.length; j++) {
+        // 检查是否已取消
+        if (newTask.controller.signal.aborted) {
+          throw new Error("Upload cancelled");
+        }
         await uploadSlice(newTask, {
           loaded,
           i,
@@ -167,7 +184,7 @@ export function useFileSliceUpload(
     removeTask(newTask);
   }
 
-  const handleUpload = async ({ task, files }) => {
+  const handleUpload = async ({ task, files }: { task: TaskItem; files: { slices: Blob[]; name: string; size: number; originFile: Blob }[] }) => {
     const isErrorFile = await checkIsFilesExist(files);
     if (isErrorFile) {
       message.error("文件被修改或删除，请重新选择文件上传");
@@ -193,10 +210,174 @@ export function useFileSliceUpload(
     }
   };
 
+  /**
+   * 流式分割上传处理
+   * 用于大文件按行分割并立即上传的场景
+   */
+  const handleStreamUpload = async ({ task, files }: { task: TaskItem; files: File[] }) => {
+    try {
+      console.log('[useSliceUpload] Starting stream upload for', files.length, 'files');
+      
+      const totalSize = files.reduce((acc, file) => acc + file.size, 0);
+      
+      // 存储所有文件的 reqId，用于取消上传
+      const reqIds: number[] = [];
+      
+      const newTask: TaskItem = {
+        ...task,
+        reqId: -1,
+        isCancel: false,
+        cancelFn: () => {
+          // 使用 newTask 的 controller 确保一致性
+          newTask.controller.abort();
+          // 取消所有文件的预上传请求
+          reqIds.forEach(id => cancelUpload?.(id));
+          if (newTask.updateEvent) window.dispatchEvent(new Event(newTask.updateEvent));
+        },
+      };
+      updateTaskList(newTask);
+
+      let totalUploadedLines = 0;
+      let totalProcessedBytes = 0;
+      const results: StreamUploadResult[] = [];
+
+      // 逐个处理文件，每个文件单独调用 preUpload
+      for (let i = 0; i < files.length; i++) {
+        // 检查是否已取消
+        if (newTask.controller.signal.aborted) {
+          throw new Error("Upload cancelled");
+        }
+        
+        const file = files[i];
+        console.log(`[useSliceUpload] Processing file ${i + 1}/${files.length}: ${file.name}`);
+        
+        const result = await streamSplitAndUpload(
+          file,
+          (formData, config) => uploadChunk(task.key, formData, {
+            ...config,
+            signal: newTask.controller.signal,
+          }),
+          (currentBytes, totalBytes, uploadedLines) => {
+            // 检查是否已取消
+            if (newTask.controller.signal.aborted) {
+              return;
+            }
+            
+            // 更新进度
+            const overallBytes = totalProcessedBytes + currentBytes;
+            const curPercent = Number((overallBytes / totalSize) * 100).toFixed(2);
+            
+            const updatedTask: TaskItem = {
+              ...newTask,
+              ...taskListRef.current.find((item) => item.key === task.key),
+              size: overallBytes,
+              percent: curPercent >= 100 ? 99.99 : curPercent,
+              streamUploadInfo: {
+                currentFile: file.name,
+                fileIndex: i + 1,
+                totalFiles: files.length,
+                uploadedLines: totalUploadedLines + uploadedLines,
+              },
+            };
+            updateTaskList(updatedTask);
+          }, 
+          1024 * 1024, // 1MB chunk size
+          {
+            resolveReqId: async ({ totalFileNum, totalSize }) => {
+              const { data: reqId } = await preUpload(task.key, {
+                totalFileNum,
+                totalSize,
+                datasetId: task.key,
+                hasArchive: task.hasArchive,
+                prefix: task.prefix,
+              });
+              console.log(`[useSliceUpload] File ${file.name} preUpload response reqId:`, reqId);
+              reqIds.push(reqId);
+              return reqId;
+            },
+            hasArchive: newTask.hasArchive,
+            prefix: newTask.prefix,
+            signal: newTask.controller.signal,
+            maxConcurrency: 3,
+          }
+        );
+        
+        results.push(result);
+        totalUploadedLines += result.uploadedCount;
+        totalProcessedBytes += file.size;
+        
+        console.log(`[useSliceUpload] File ${file.name} processed, uploaded ${result.uploadedCount} lines`);
+      }
+
+      console.log('[useSliceUpload] Stream upload completed, total lines:', totalUploadedLines);
+      removeTask(newTask);
+      
+      message.success(`成功上传 ${totalUploadedLines} 个文件（按行分割）`);
+    } catch (err) {
+      console.error('[useSliceUpload] Stream upload error:', err);
+      if (err.message === "Upload cancelled") {
+        message.info("上传已取消");
+      } else {
+        message.error("文件上传失败，请稍后重试");
+      }
+      removeTask({
+        ...task,
+        isCancel: true,
+        ...taskListRef.current.find((item) => item.key === task.key),
+      });
+    }
+  };
+
+  /**
+   * 注册流式上传事件监听
+   * 返回注销函数
+   */
+  const registerStreamUploadListener = () => {
+    if (!enableStreamUpload) return () => {};
+
+    const streamUploadHandler = async (e: Event) => {
+      const customEvent = e as CustomEvent;
+      const { dataset, files, updateEvent, hasArchive, prefix } = customEvent.detail;
+      
+      const controller = new AbortController();
+      const task: TaskItem = {
+        key: dataset.id,
+        title: `上传数据集: ${dataset.name} (按行分割)`,
+        percent: 0,
+        reqId: -1,
+        controller,
+        size: 0,
+        updateEvent,
+        hasArchive,
+        prefix,
+      };
+      
+      taskListRef.current = [task, ...taskListRef.current];
+      setTaskList(taskListRef.current);
+      
+      // 显示任务中心
+      if (showTaskCenter) {
+        window.dispatchEvent(
+          new CustomEvent("show:task-popover", { detail: { show: true } })
+        );
+      }
+      
+      await handleStreamUpload({ task, files });
+    };
+
+    window.addEventListener("upload:dataset-stream", streamUploadHandler);
+    
+    return () => {
+      window.removeEventListener("upload:dataset-stream", streamUploadHandler);
+    };
+  };
+
   return {
     taskList,
     createTask,
     removeTask,
     handleUpload,
+    handleStreamUpload,
+    registerStreamUploadListener,
   };
 }

frontend/src/pages/DataAnnotation/Annotate/LabelStudioTextEditor.tsx

@@ -6,6 +6,7 @@ import { useNavigate, useParams } from "react-router";
 import {
   getEditorProjectInfoUsingGet,
   getEditorTaskUsingGet,
+  getEditorTaskSegmentsUsingGet,
   listEditorTasksUsingGet,
   upsertEditorAnnotationUsingPut,
 } from "../annotation.api";
@@ -38,9 +39,6 @@ type LsfMessage = {
 
 type SegmentInfo = {
   idx: number;
-  text: string;
-  start: number;
-  end: number;
   hasAnnotation: boolean;
   lineIndex: number;
   chunkIndex: number;
@@ -66,10 +64,16 @@ type EditorTaskPayload = {
 type EditorTaskResponse = {
   task?: EditorTaskPayload;
   segmented?: boolean;
-  segments?: SegmentInfo[];
+  totalSegments?: number;
   currentSegmentIndex?: number;
 };
 
+type EditorTaskSegmentsResponse = {
+  segmented?: boolean;
+  segments?: SegmentInfo[];
+  totalSegments?: number;
+};
+
 type EditorTaskListResponse = {
   content?: EditorTaskListItem[];
   totalElements?: number;
@@ -288,6 +292,7 @@ export default function LabelStudioTextEditor() {
   const segmentStatsCacheRef = useRef<Record<string, SegmentStats>>({});
   const segmentStatsSeqRef = useRef(0);
   const segmentStatsLoadingRef = useRef<Set<string>>(new Set());
+  const segmentSummaryFileRef = useRef<string>("");
 
   const [loadingProject, setLoadingProject] = useState(true);
   const [loadingTasks, setLoadingTasks] = useState(false);
@@ -358,9 +363,7 @@ export default function LabelStudioTextEditor() {
     if (segmentStatsCacheRef.current[fileId] || segmentStatsLoadingRef.current.has(fileId)) return;
     segmentStatsLoadingRef.current.add(fileId);
     try {
-      const resp = (await getEditorTaskUsingGet(projectId, fileId, {
+      const resp = (await getEditorTaskSegmentsUsingGet(projectId, fileId)) as ApiResponse<EditorTaskSegmentsResponse>;
-        segmentIndex: 0,
-      })) as ApiResponse<EditorTaskResponse>;
       if (segmentStatsSeqRef.current !== seq) return;
       const data = resp?.data;
       if (!data?.segmented) return;
@@ -591,20 +594,38 @@ export default function LabelStudioTextEditor() {
       if (seq !== initSeqRef.current) return;
 
       // 更新分段状态
-      const segmentIndex = data?.segmented
+      const isSegmented = !!data?.segmented;
+      const segmentIndex = isSegmented
         ? resolveSegmentIndex(data.currentSegmentIndex) ?? 0
         : undefined;
-      if (data?.segmented) {
+      if (isSegmented) {
-        const stats = buildSegmentStats(data.segments);
+        let nextSegments: SegmentInfo[] = [];
+        if (segmentSummaryFileRef.current === fileId && segments.length > 0) {
+          nextSegments = segments;
+        } else {
+          try {
+            const segmentResp = (await getEditorTaskSegmentsUsingGet(projectId, fileId)) as ApiResponse<EditorTaskSegmentsResponse>;
+            if (seq !== initSeqRef.current) return;
+            const segmentData = segmentResp?.data;
+            if (segmentData?.segmented) {
+              nextSegments = Array.isArray(segmentData.segments) ? segmentData.segments : [];
+            }
+          } catch (e) {
+            console.error(e);
+          }
+        }
+        const stats = buildSegmentStats(nextSegments);
         setSegmented(true);
-        setSegments(data.segments || []);
+        setSegments(nextSegments);
         setCurrentSegmentIndex(segmentIndex ?? 0);
         updateSegmentStatsCache(fileId, stats);
+        segmentSummaryFileRef.current = fileId;
       } else {
         setSegmented(false);
         setSegments([]);
         setCurrentSegmentIndex(0);
         updateSegmentStatsCache(fileId, null);
+        segmentSummaryFileRef.current = fileId;
       }
 
       const taskData = {
@@ -664,7 +685,7 @@ export default function LabelStudioTextEditor() {
     } finally {
       if (seq === initSeqRef.current) setLoadingTaskDetail(false);
     }
-  }, [iframeReady, message, postToIframe, project, projectId, updateSegmentStatsCache]);
+  }, [iframeReady, message, postToIframe, project, projectId, segments, updateSegmentStatsCache]);
 
   const advanceAfterSave = useCallback(async (fileId: string, segmentIndex?: number) => {
     if (!fileId) return;
@@ -979,6 +1000,7 @@ export default function LabelStudioTextEditor() {
     setSegmented(false);
     setSegments([]);
     setCurrentSegmentIndex(0);
+    segmentSummaryFileRef.current = "";
     savedSnapshotsRef.current = {};
     segmentStatsSeqRef.current += 1;
     segmentStatsCacheRef.current = {};

frontend/src/pages/DataAnnotation/annotation.api.ts

@@ -3,16 +3,19 @@ import { get, post, put, del, download } from "@/utils/request";
 // 导出格式类型
 export type ExportFormat = "json" | "jsonl" | "csv" | "coco" | "yolo";
 
+type RequestParams = Record<string, unknown>;
+type RequestPayload = Record<string, unknown>;
+
 // 标注任务管理相关接口
-export function queryAnnotationTasksUsingGet(params?: any) {
+export function queryAnnotationTasksUsingGet(params?: RequestParams) {
   return get("/api/annotation/project", params);
 }
 
-export function createAnnotationTaskUsingPost(data: any) {
+export function createAnnotationTaskUsingPost(data: RequestPayload) {
   return post("/api/annotation/project", data);
 }
 
-export function syncAnnotationTaskUsingPost(data: any) {
+export function syncAnnotationTaskUsingPost(data: RequestPayload) {
   return post(`/api/annotation/task/sync`, data);
 }
 
@@ -25,7 +28,7 @@ export function getAnnotationTaskByIdUsingGet(taskId: string) {
   return get(`/api/annotation/project/${taskId}`);
 }
 
-export function updateAnnotationTaskByIdUsingPut(taskId: string, data: any) {
+export function updateAnnotationTaskByIdUsingPut(taskId: string, data: RequestPayload) {
   return put(`/api/annotation/project/${taskId}`, data);
 }
 
@@ -35,17 +38,17 @@ export function getTagConfigUsingGet() {
 }
 
 // 标注模板管理
-export function queryAnnotationTemplatesUsingGet(params?: any) {
+export function queryAnnotationTemplatesUsingGet(params?: RequestParams) {
   return get("/api/annotation/template", params);
 }
 
-export function createAnnotationTemplateUsingPost(data: any) {
+export function createAnnotationTemplateUsingPost(data: RequestPayload) {
   return post("/api/annotation/template", data);
 }
 
 export function updateAnnotationTemplateByIdUsingPut(
   templateId: string | number,
-  data: any
+  data: RequestPayload
 ) {
   return put(`/api/annotation/template/${templateId}`, data);
 }
@@ -65,7 +68,7 @@ export function getEditorProjectInfoUsingGet(projectId: string) {
   return get(`/api/annotation/editor/projects/${projectId}`);
 }
 
-export function listEditorTasksUsingGet(projectId: string, params?: any) {
+export function listEditorTasksUsingGet(projectId: string, params?: RequestParams) {
   return get(`/api/annotation/editor/projects/${projectId}/tasks`, params);
 }
 
@@ -77,11 +80,15 @@ export function getEditorTaskUsingGet(
   return get(`/api/annotation/editor/projects/${projectId}/tasks/${fileId}`, params);
 }
 
+export function getEditorTaskSegmentsUsingGet(projectId: string, fileId: string) {
+  return get(`/api/annotation/editor/projects/${projectId}/tasks/${fileId}/segments`);
+}
+
 export function upsertEditorAnnotationUsingPut(
   projectId: string,
   fileId: string,
   data: {
-    annotation: any;
+    annotation: Record<string, unknown>;
     expectedUpdatedAt?: string;
     segmentIndex?: number;
   }

frontend/src/pages/DataManagement/Detail/components/ImportConfiguration.tsx

@@ -5,7 +5,7 @@ import { Dataset, DatasetType, DataSource } from "../../dataset.model";
 import { useCallback, useEffect, useMemo, useState } from "react";
 import { queryTasksUsingGet } from "@/pages/DataCollection/collection.apis";
 import { updateDatasetByIdUsingPut } from "../../dataset.api";
-import { sliceFile } from "@/utils/file.util";
+import { sliceFile, shouldStreamUpload } from "@/utils/file.util";
 import Dragger from "antd/es/upload/Dragger";
 
 const TEXT_FILE_MIME_PREFIX = "text/";
@@ -90,14 +90,16 @@ async function splitFileByLines(file: UploadFile): Promise<UploadFile[]> {
   const lines = text.split(/\r?\n/).filter((line: string) => line.trim() !== "");
   if (lines.length === 0) return [];
 
-  // 生成文件名：原文件名_序号.扩展名
+  // 生成文件名：原文件名_序号（不保留后缀）
   const nameParts = file.name.split(".");
-  const ext = nameParts.length > 1 ? "." + nameParts.pop() : "";
+  if (nameParts.length > 1) {
+    nameParts.pop();
+  }
   const baseName = nameParts.join(".");
   const padLength = String(lines.length).length;
 
   return lines.map((line: string, index: number) => {
-    const newFileName = `${baseName}_${String(index + 1).padStart(padLength, "0")}${ext}`;
+    const newFileName = `${baseName}_${String(index + 1).padStart(padLength, "0")}`;
     const blob = new Blob([line], { type: "text/plain" });
     const newFile = new File([blob], newFileName, { type: "text/plain" });
     return {
@@ -164,17 +166,75 @@ export default function ImportConfiguration({
   // 本地上传文件相关逻辑
 
   const handleUpload = async (dataset: Dataset) => {
-    let filesToUpload =
+    const filesToUpload =
       (form.getFieldValue("files") as UploadFile[] | undefined) || [];
 
-    // 如果启用分行分割，处理文件
+    // 如果启用分行分割，对大文件使用流式处理
     if (importConfig.splitByLine && !hasNonTextFile) {
-      const splitResults = await Promise.all(
+      // 检查是否有大文件需要流式分割上传
-        filesToUpload.map((file) => splitFileByLines(file))
+      const filesForStreamUpload: File[] = [];
-      );
+      const filesForNormalUpload: UploadFile[] = [];
-      filesToUpload = splitResults.flat();
+
+      for (const file of filesToUpload) {
+        const originFile = file.originFileObj ?? file;
+        if (originFile instanceof File && shouldStreamUpload(originFile)) {
+          filesForStreamUpload.push(originFile);
+        } else {
+          filesForNormalUpload.push(file);
+        }
       }
 
+      // 大文件使用流式分割上传
+      if (filesForStreamUpload.length > 0) {
+        window.dispatchEvent(
+          new CustomEvent("upload:dataset-stream", {
+            detail: {
+              dataset,
+              files: filesForStreamUpload,
+              updateEvent,
+              hasArchive: importConfig.hasArchive,
+              prefix: currentPrefix,
+            },
+          })
+        );
+      }
+
+      // 小文件使用传统分割方式
+      if (filesForNormalUpload.length > 0) {
+        const splitResults = await Promise.all(
+          filesForNormalUpload.map((file) => splitFileByLines(file))
+        );
+        const smallFilesToUpload = splitResults.flat();
+        
+        // 计算分片列表
+        const sliceList = smallFilesToUpload.map((file) => {
+          const originFile = (file.originFileObj ?? file) as Blob;
+          const slices = sliceFile(originFile);
+          return {
+            originFile: originFile,
+            slices,
+            name: file.name,
+            size: originFile.size || 0,
+          };
+        });
+
+        console.log("[ImportConfiguration] Uploading small files with currentPrefix:", currentPrefix);
+        window.dispatchEvent(
+          new CustomEvent("upload:dataset", {
+            detail: {
+              dataset,
+              files: sliceList,
+              updateEvent,
+              hasArchive: importConfig.hasArchive,
+              prefix: currentPrefix,
+            },
+          })
+        );
+      }
+      return;
+    }
+
+    // 未启用分行分割，使用普通上传
     // 计算分片列表
     const sliceList = filesToUpload.map((file) => {
       const originFile = (file.originFileObj ?? file) as Blob;

frontend/src/pages/DataManagement/dataset.model.ts

@@ -102,6 +102,13 @@ export interface DatasetTask {
   executionHistory?: { time: string; status: string }[];
 }
 
+export interface StreamUploadInfo {
+  currentFile: string;
+  fileIndex: number;
+  totalFiles: number;
+  uploadedLines: number;
+}
+
 export interface TaskItem {
   key: string;
   title: string;
@@ -113,4 +120,6 @@ export interface TaskItem {
   updateEvent?: string;
   size?: number;
   hasArchive?: boolean;
+  prefix?: string;
+  streamUploadInfo?: StreamUploadInfo;
 }

frontend/src/pages/Layout/TaskUpload.tsx

@@ -3,25 +3,28 @@ import {
   preUploadUsingPost,
   uploadFileChunkUsingPost,
 } from "@/pages/DataManagement/dataset.api";
-import { Button, Empty, Progress } from "antd";
+import { Button, Empty, Progress, Tag } from "antd";
-import { DeleteOutlined } from "@ant-design/icons";
+import { DeleteOutlined, FileTextOutlined } from "@ant-design/icons";
 import { useEffect } from "react";
 import { useFileSliceUpload } from "@/hooks/useSliceUpload";
 
 export default function TaskUpload() {
-  const { createTask, taskList, removeTask, handleUpload } = useFileSliceUpload(
+  const { createTask, taskList, removeTask, handleUpload, registerStreamUploadListener } = useFileSliceUpload(
     {
       preUpload: preUploadUsingPost,
       uploadChunk: uploadFileChunkUsingPost,
       cancelUpload: cancelUploadUsingPut,
-    }
+    },
+    true, // showTaskCenter
+    true  // enableStreamUpload
   );
 
   useEffect(() => {
-    const uploadHandler = (e: any) => {
+    const uploadHandler = (e: Event) => {
-      console.log('[TaskUpload] Received upload event detail:', e.detail);
+      const customEvent = e as CustomEvent;
-      const { files } = e.detail;
+      console.log('[TaskUpload] Received upload event detail:', customEvent.detail);
-      const task = createTask(e.detail);
+      const { files } = customEvent.detail;
+      const task = createTask(customEvent.detail);
       console.log('[TaskUpload] Created task with prefix:', task.prefix);
       handleUpload({ task, files });
     };
@@ -29,7 +32,13 @@ export default function TaskUpload() {
     return () => {
       window.removeEventListener("upload:dataset", uploadHandler);
     };
-  }, []);
+  }, [createTask, handleUpload]);
+
+  // 注册流式上传监听器
+  useEffect(() => {
+    const unregister = registerStreamUploadListener();
+    return unregister;
+  }, [registerStreamUploadListener]);
 
   return (
     <div
@@ -55,7 +64,22 @@ export default function TaskUpload() {
               ></Button>
             </div>
 
-            <Progress size="small" percent={task.percent} />
+            <Progress size="small" percent={Number(task.percent)} />
+            {task.streamUploadInfo && (
+              <div className="flex items-center gap-2 text-xs text-gray-500 mt-1">
+                <Tag icon={<FileTextOutlined />} size="small">
+                  按行分割
+                </Tag>
+                <span>
+                  已上传: {task.streamUploadInfo.uploadedLines} 行
+                </span>
+                {task.streamUploadInfo.totalFiles > 1 && (
+                  <span>
+                    ({task.streamUploadInfo.fileIndex}/{task.streamUploadInfo.totalFiles} 文件)
+                  </span>
+                )}
+              </div>
+            )}
           </div>
         ))}
       {taskList.length === 0 && (

frontend/src/utils/file.util.ts

@@ -1,79 +1,657 @@
 import { UploadFile } from "antd";
 import jsSHA from "jssha";
 
-const CHUNK_SIZE = 1024 * 1024 * 60;
+// 默认分片大小：5MB（适合大多数网络环境）
+export const DEFAULT_CHUNK_SIZE = 1024 * 1024 * 5;
+// 大文件阈值：10MB
+export const LARGE_FILE_THRESHOLD = 1024 * 1024 * 10;
+// 最大并发上传数
+export const MAX_CONCURRENT_UPLOADS = 3;
+// 文本文件读取块大小：20MB（用于计算 SHA256）
+const BUFFER_CHUNK_SIZE = 1024 * 1024 * 20;
 
-export function sliceFile(file, chunkSize = CHUNK_SIZE): Blob[] {
+/**
+ * 将文件分割为多个分片
+ * @param file 文件对象
+ * @param chunkSize 分片大小（字节），默认 5MB
+ * @returns 分片数组（Blob 列表）
+ */
+export function sliceFile(file: Blob, chunkSize = DEFAULT_CHUNK_SIZE): Blob[] {
   const totalSize = file.size;
+  const chunks: Blob[] = [];
+  
+  // 小文件不需要分片
+  if (totalSize <= chunkSize) {
+    return [file];
+  }
+  
   let start = 0;
-  let end = start + chunkSize;
-  const chunks = [];
   while (start < totalSize) {
+    const end = Math.min(start + chunkSize, totalSize);
     const blob = file.slice(start, end);
     chunks.push(blob);
-
     start = end;
-    end = start + chunkSize;
   }
+  
   return chunks;
 }
 
-export function calculateSHA256(file: Blob): Promise<string> {
+/**
-  let count = 0;
+ * 计算文件的 SHA256 哈希值
-  const hash = new jsSHA("SHA-256", "ARRAYBUFFER", { encoding: "UTF8" });
+ * @param file 文件 Blob
+ * @param onProgress 进度回调（可选）
+ * @returns SHA256 哈希字符串
+ */
+export function calculateSHA256(
+  file: Blob, 
+  onProgress?: (percent: number) => void
+): Promise<string> {
   return new Promise((resolve, reject) => {
+    const hash = new jsSHA("SHA-256", "ARRAYBUFFER", { encoding: "UTF8" });
     const reader = new FileReader();
+    let processedSize = 0;
 
     function readChunk(start: number, end: number) {
       const slice = file.slice(start, end);
       reader.readAsArrayBuffer(slice);
     }
 
-    const bufferChunkSize = 1024 * 1024 * 20;
-
     function processChunk(offset: number) {
       const start = offset;
-      const end = Math.min(start + bufferChunkSize, file.size);
+      const end = Math.min(start + BUFFER_CHUNK_SIZE, file.size);
-      count = end;
-
       readChunk(start, end);
     }
 
-    reader.onloadend = function () {
+    reader.onloadend = function (e) {
-      const arraybuffer = reader.result;
+      const arraybuffer = reader.result as ArrayBuffer;
+      if (!arraybuffer) {
+        reject(new Error("Failed to read file"));
+        return;
+      }
 
       hash.update(arraybuffer);
-      if (count < file.size) {
+      processedSize += (e.target as FileReader).result?.byteLength || 0;
-        processChunk(count);
+      
+      if (onProgress) {
+        const percent = Math.min(100, Math.round((processedSize / file.size) * 100));
+        onProgress(percent);
+      }
+      
+      if (processedSize < file.size) {
+        processChunk(processedSize);
       } else {
         resolve(hash.getHash("HEX", { outputLen: 256 }));
       }
     };
 
+    reader.onerror = () => reject(new Error("File reading failed"));
     processChunk(0);
   });
 }
 
+/**
+ * 批量计算多个文件的 SHA256
+ * @param files 文件列表
+ * @param onFileProgress 单个文件进度回调（可选）
+ * @returns 哈希值数组
+ */
+export async function calculateSHA256Batch(
+  files: Blob[],
+  onFileProgress?: (index: number, percent: number) => void
+): Promise<string[]> {
+  const results: string[] = [];
+  
+  for (let i = 0; i < files.length; i++) {
+    const hash = await calculateSHA256(files[i], (percent) => {
+      onFileProgress?.(i, percent);
+    });
+    results.push(hash);
+  }
+  
+  return results;
+}
+
+/**
+ * 检查文件是否存在（未被修改或删除）
+ * @param fileList 文件列表
+ * @returns 返回第一个不存在的文件，或 null（如果都存在）
+ */
 export function checkIsFilesExist(
-  fileList: UploadFile[]
+  fileList: Array<{ originFile?: Blob }>
-): Promise<UploadFile | null> {
+): Promise<{ originFile?: Blob } | null> {
   return new Promise((resolve) => {
-    const loadEndFn = (file: UploadFile, reachEnd: boolean, e) => {
+    if (!fileList.length) {
-      const fileNotExist = !e.target.result;
+      resolve(null);
+      return;
+    }
+
+    let checkedCount = 0;
+    const totalCount = fileList.length;
+
+    const loadEndFn = (file: { originFile?: Blob }, e: ProgressEvent<FileReader>) => {
+      checkedCount++;
+      const fileNotExist = !e.target?.result;
       if (fileNotExist) {
         resolve(file);
+        return;
       }
-      if (reachEnd) {
+      if (checkedCount >= totalCount) {
         resolve(null);
       }
     };
 
-    for (let i = 0; i < fileList.length; i++) {
+    for (const file of fileList) {
-      const { originFile: file } = fileList[i];
       const fileReader = new FileReader();
-      fileReader.readAsArrayBuffer(file);
+      const actualFile = file.originFile;
-      fileReader.onloadend = (e) =>
+      
-        loadEndFn(fileList[i], i === fileList.length - 1, e);
+      if (!actualFile) {
+        checkedCount++;
+        if (checkedCount >= totalCount) {
+          resolve(null);
+        }
+        continue;
+      }
+      
+      fileReader.readAsArrayBuffer(actualFile.slice(0, 1));
+      fileReader.onloadend = (e) => loadEndFn(file, e);
+      fileReader.onerror = () => {
+        checkedCount++;
+        resolve(file);
+      };
     }
   });
 }
+
+/**
+ * 判断文件是否为大文件
+ * @param size 文件大小（字节）
+ * @param threshold 阈值（字节），默认 10MB
+ */
+export function isLargeFile(size: number, threshold = LARGE_FILE_THRESHOLD): boolean {
+  return size > threshold;
+}
+
+/**
+ * 格式化文件大小为人类可读格式
+ * @param bytes 字节数
+ * @param decimals 小数位数
+ */
+export function formatFileSize(bytes: number, decimals = 2): string {
+  if (bytes === 0) return "0 B";
+  
+  const k = 1024;
+  const sizes = ["B", "KB", "MB", "GB", "TB", "PB"];
+  const i = Math.floor(Math.log(bytes) / Math.log(k));
+  
+  return `${parseFloat((bytes / Math.pow(k, i)).toFixed(decimals))} ${sizes[i]}`;
+}
+
+/**
+ * 并发执行异步任务
+ * @param tasks 任务函数数组
+ * @param maxConcurrency 最大并发数
+ * @param onTaskComplete 单个任务完成回调（可选）
+ */
+export async function runConcurrentTasks<T>(
+  tasks: (() => Promise<T>)[],
+  maxConcurrency: number,
+  onTaskComplete?: (index: number, result: T) => void
+): Promise<T[]> {
+  const results: T[] = new Array(tasks.length);
+  let index = 0;
+
+  async function runNext(): Promise<void> {
+    const currentIndex = index++;
+    if (currentIndex >= tasks.length) return;
+
+    const result = await tasks[currentIndex]();
+    results[currentIndex] = result;
+    onTaskComplete?.(currentIndex, result);
+    
+    await runNext();
+  }
+
+  const workers = Array(Math.min(maxConcurrency, tasks.length))
+    .fill(null)
+    .map(() => runNext());
+
+  await Promise.all(workers);
+  return results;
+}
+
+/**
+ * 按行分割文本文件内容
+ * @param text 文本内容
+ * @param skipEmptyLines 是否跳过空行，默认 true
+ * @returns 行数组
+ */
+export function splitTextByLines(text: string, skipEmptyLines = true): string[] {
+  const lines = text.split(/\r?\n/);
+  if (skipEmptyLines) {
+    return lines.filter((line) => line.trim() !== "");
+  }
+  return lines;
+}
+
+/**
+ * 创建分片信息对象
+ * @param file 原始文件
+ * @param chunkSize 分片大小
+ */
+export function createFileSliceInfo(
+  file: File | Blob,
+  chunkSize = DEFAULT_CHUNK_SIZE
+): {
+  originFile: Blob;
+  slices: Blob[];
+  name: string;
+  size: number;
+  totalChunks: number;
+} {
+  const slices = sliceFile(file, chunkSize);
+  return {
+    originFile: file,
+    slices,
+    name: (file as File).name || "unnamed",
+    size: file.size,
+    totalChunks: slices.length,
+  };
+}
+
+/**
+ * 支持的文本文件 MIME 类型前缀
+ */
+export const TEXT_FILE_MIME_PREFIX = "text/";
+
+/**
+ * 支持的文本文件 MIME 类型集合
+ */
+export const TEXT_FILE_MIME_TYPES = new Set([
+  "application/json",
+  "application/xml",
+  "application/csv",
+  "application/ndjson",
+  "application/x-ndjson",
+  "application/x-yaml",
+  "application/yaml",
+  "application/javascript",
+  "application/x-javascript",
+  "application/sql",
+  "application/rtf",
+  "application/xhtml+xml",
+  "application/svg+xml",
+]);
+
+/**
+ * 支持的文本文件扩展名集合
+ */
+export const TEXT_FILE_EXTENSIONS = new Set([
+  ".txt",
+  ".md",
+  ".markdown",
+  ".csv",
+  ".tsv",
+  ".json",
+  ".jsonl",
+  ".ndjson",
+  ".log",
+  ".xml",
+  ".yaml",
+  ".yml",
+  ".sql",
+  ".js",
+  ".ts",
+  ".jsx",
+  ".tsx",
+  ".html",
+  ".htm",
+  ".css",
+  ".scss",
+  ".less",
+  ".py",
+  ".java",
+  ".c",
+  ".cpp",
+  ".h",
+  ".hpp",
+  ".go",
+  ".rs",
+  ".rb",
+  ".php",
+  ".sh",
+  ".bash",
+  ".zsh",
+  ".ps1",
+  ".bat",
+  ".cmd",
+  ".svg",
+  ".rtf",
+]);
+
+/**
+ * 判断文件是否为文本文件（支持 UploadFile 类型）
+ * @param file UploadFile 对象
+ */
+export function isTextUploadFile(file: UploadFile): boolean {
+  const mimeType = (file.type || "").toLowerCase();
+  if (mimeType) {
+    if (mimeType.startsWith(TEXT_FILE_MIME_PREFIX)) return true;
+    if (TEXT_FILE_MIME_TYPES.has(mimeType)) return true;
+  }
+
+  const fileName = file.name || "";
+  const dotIndex = fileName.lastIndexOf(".");
+  if (dotIndex < 0) return false;
+  const ext = fileName.slice(dotIndex).toLowerCase();
+  return TEXT_FILE_EXTENSIONS.has(ext);
+}
+
+/**
+ * 判断文件名是否为文本文件
+ * @param fileName 文件名
+ */
+export function isTextFileByName(fileName: string): boolean {
+  const lowerName = fileName.toLowerCase();
+  
+  // 先检查 MIME 类型（如果有）
+  // 这里简化处理，主要通过扩展名判断
+  
+  const dotIndex = lowerName.lastIndexOf(".");
+  if (dotIndex < 0) return false;
+  const ext = lowerName.slice(dotIndex);
+  return TEXT_FILE_EXTENSIONS.has(ext);
+}
+
+/**
+ * 获取文件扩展名
+ * @param fileName 文件名
+ */
+export function getFileExtension(fileName: string): string {
+  const dotIndex = fileName.lastIndexOf(".");
+  if (dotIndex < 0) return "";
+  return fileName.slice(dotIndex).toLowerCase();
+}
+
+/**
+ * 安全地读取文件为文本
+ * @param file 文件对象
+ * @param encoding 编码，默认 UTF-8
+ */
+export function readFileAsText(
+  file: File | Blob,
+  encoding = "UTF-8"
+): Promise<string> {
+  return new Promise((resolve, reject) => {
+    const reader = new FileReader();
+    reader.onload = (e) => resolve(e.target?.result as string);
+    reader.onerror = () => reject(new Error("Failed to read file"));
+    reader.readAsText(file, encoding);
+  });
+}
+
+/**
+ * 流式分割文件并逐行上传
+ * 使用 Blob.slice 逐块读取，避免一次性加载大文件到内存
+ * @param file 文件对象
+ * @param datasetId 数据集ID
+ * @param uploadFn 上传函数，接收 FormData 和配置，返回 Promise
+ * @param onProgress 进度回调 (currentBytes, totalBytes, uploadedLines)
+ * @param chunkSize 每次读取的块大小，默认 1MB
+ * @param options 其他选项
+ * @returns 上传结果统计
+ */
+export interface StreamUploadOptions {
+  reqId?: number;
+  resolveReqId?: (params: { totalFileNum: number; totalSize: number }) => Promise<number>;
+  onReqIdResolved?: (reqId: number) => void;
+  fileNamePrefix?: string;
+  hasArchive?: boolean;
+  prefix?: string;
+  signal?: AbortSignal;
+  maxConcurrency?: number;
+}
+
+export interface StreamUploadResult {
+  uploadedCount: number;
+  totalBytes: number;
+  skippedEmptyCount: number;
+}
+
+async function processFileLines(
+  file: File,
+  chunkSize: number,
+  signal: AbortSignal | undefined,
+  onLine?: (line: string, index: number) => Promise<void> | void,
+  onProgress?: (currentBytes: number, totalBytes: number, processedLines: number) => void
+): Promise<{ lineCount: number; skippedEmptyCount: number }> {
+  const fileSize = file.size;
+  let offset = 0;
+  let buffer = "";
+  let skippedEmptyCount = 0;
+  let lineIndex = 0;
+
+  while (offset < fileSize) {
+    if (signal?.aborted) {
+      throw new Error("Upload cancelled");
+    }
+
+    const end = Math.min(offset + chunkSize, fileSize);
+    const chunk = file.slice(offset, end);
+    const text = await readFileAsText(chunk);
+    const combined = buffer + text;
+    const lines = combined.split(/\r?\n/);
+    buffer = lines.pop() || "";
+
+    for (const line of lines) {
+      if (signal?.aborted) {
+        throw new Error("Upload cancelled");
+      }
+      if (!line.trim()) {
+        skippedEmptyCount++;
+        continue;
+      }
+      const currentIndex = lineIndex;
+      lineIndex += 1;
+      if (onLine) {
+        await onLine(line, currentIndex);
+      }
+    }
+
+    offset = end;
+    onProgress?.(offset, fileSize, lineIndex);
+  }
+
+  if (buffer.trim()) {
+    const currentIndex = lineIndex;
+    lineIndex += 1;
+    if (onLine) {
+      await onLine(buffer, currentIndex);
+    }
+  } else if (buffer.length > 0) {
+    skippedEmptyCount++;
+  }
+
+  return { lineCount: lineIndex, skippedEmptyCount };
+}
+
+export async function streamSplitAndUpload(
+  file: File,
+  uploadFn: (formData: FormData, config?: { onUploadProgress?: (e: { loaded: number; total: number }) => void }) => Promise<unknown>,
+  onProgress?: (currentBytes: number, totalBytes: number, uploadedLines: number) => void,
+  chunkSize: number = 1024 * 1024, // 1MB
+  options: StreamUploadOptions
+): Promise<StreamUploadResult> {
+  const {
+    reqId: initialReqId,
+    resolveReqId,
+    onReqIdResolved,
+    fileNamePrefix,
+    prefix,
+    signal,
+    maxConcurrency = 3,
+  } = options;
+
+  const fileSize = file.size;
+  let uploadedCount = 0;
+  let skippedEmptyCount = 0;
+  
+  // 获取文件名基础部分和扩展名
+  const originalFileName = fileNamePrefix || file.name;
+  const lastDotIndex = originalFileName.lastIndexOf(".");
+  const baseName = lastDotIndex > 0 ? originalFileName.slice(0, lastDotIndex) : originalFileName;
+  const fileExtension = lastDotIndex > 0 ? originalFileName.slice(lastDotIndex) : "";
+
+  let resolvedReqId = initialReqId;
+  if (!resolvedReqId) {
+    const scanResult = await processFileLines(file, chunkSize, signal);
+    const totalFileNum = scanResult.lineCount;
+    skippedEmptyCount = scanResult.skippedEmptyCount;
+    if (totalFileNum === 0) {
+      return {
+        uploadedCount: 0,
+        totalBytes: fileSize,
+        skippedEmptyCount,
+      };
+    }
+    if (signal?.aborted) {
+      throw new Error("Upload cancelled");
+    }
+    if (!resolveReqId) {
+      throw new Error("Missing pre-upload request id");
+    }
+    resolvedReqId = await resolveReqId({ totalFileNum, totalSize: fileSize });
+    if (!resolvedReqId) {
+      throw new Error("Failed to resolve pre-upload request id");
+    }
+    onReqIdResolved?.(resolvedReqId);
+  }
+  if (!resolvedReqId) {
+    throw new Error("Missing pre-upload request id");
+  }
+
+  /**
+   * 上传单行内容
+   * 每行作为独立文件上传，fileNo 对应行序号，chunkNo 固定为 1
+   */
+  async function uploadLine(line: string, index: number): Promise<void> {
+    // 检查是否已取消
+    if (signal?.aborted) {
+      throw new Error("Upload cancelled");
+    }
+    
+    if (!line.trim()) {
+      skippedEmptyCount++;
+      return;
+    }
+
+    // 保留原始文件扩展名
+    const fileIndex = index + 1;
+    const newFileName = `${baseName}_${String(fileIndex).padStart(6, "0")}${fileExtension}`;
+    const blob = new Blob([line], { type: "text/plain" });
+    const lineFile = new File([blob], newFileName, { type: "text/plain" });
+    
+    // 计算分片（小文件通常只需要一个分片）
+    const slices = sliceFile(lineFile, DEFAULT_CHUNK_SIZE);
+    const checkSum = await calculateSHA256(slices[0]);
+    
+    // 检查是否已取消（计算哈希后）
+    if (signal?.aborted) {
+      throw new Error("Upload cancelled");
+    }
+    
+    const formData = new FormData();
+    formData.append("file", slices[0]);
+    formData.append("reqId", resolvedReqId.toString());
+    // 每行作为独立文件上传
+    formData.append("fileNo", fileIndex.toString());
+    formData.append("chunkNo", "1");
+    formData.append("fileName", newFileName);
+    formData.append("fileSize", lineFile.size.toString());
+    formData.append("totalChunkNum", "1");
+    formData.append("checkSumHex", checkSum);
+    if (prefix !== undefined) {
+      formData.append("prefix", prefix);
+    }
+
+    await uploadFn(formData, {
+      onUploadProgress: () => {
+        // 单行文件很小，进度主要用于追踪上传状态
+      },
+    });
+  }
+
+  const inFlight = new Set<Promise<void>>();
+  let uploadError: unknown = null;
+  const enqueueUpload = async (line: string, index: number) => {
+    if (signal?.aborted) {
+      throw new Error("Upload cancelled");
+    }
+    if (uploadError) {
+      throw uploadError;
+    }
+    const uploadPromise = uploadLine(line, index)
+      .then(() => {
+        uploadedCount++;
+      })
+      .catch((err) => {
+        uploadError = err;
+      });
+    inFlight.add(uploadPromise);
+    uploadPromise.finally(() => inFlight.delete(uploadPromise));
+    if (inFlight.size >= maxConcurrency) {
+      await Promise.race(inFlight);
+      if (uploadError) {
+        throw uploadError;
+      }
+    }
+  };
+
+  let uploadResult: { lineCount: number; skippedEmptyCount: number } | null = null;
+  try {
+    uploadResult = await processFileLines(
+      file,
+      chunkSize,
+      signal,
+      enqueueUpload,
+      (currentBytes, totalBytes) => {
+        onProgress?.(currentBytes, totalBytes, uploadedCount);
+      }
+    );
+    if (uploadError) {
+      throw uploadError;
+    }
+  } finally {
+    if (inFlight.size > 0) {
+      await Promise.allSettled(inFlight);
+    }
+  }
+
+  if (!uploadResult || (initialReqId && uploadResult.lineCount === 0)) {
+    return {
+      uploadedCount: 0,
+      totalBytes: fileSize,
+      skippedEmptyCount: uploadResult?.skippedEmptyCount ?? 0,
+    };
+  }
+
+  if (!initialReqId) {
+    skippedEmptyCount = skippedEmptyCount || uploadResult.skippedEmptyCount;
+  } else {
+    skippedEmptyCount = uploadResult.skippedEmptyCount;
+  }
+
+  return {
+    uploadedCount,
+    totalBytes: fileSize,
+    skippedEmptyCount,
+  };
+}
+
+/**
+ * 判断文件是否需要流式分割上传
+ * @param file 文件对象
+ * @param threshold 阈值，默认 5MB
+ */
+export function shouldStreamUpload(file: File, threshold: number = 5 * 1024 * 1024): boolean {
+  return file.size > threshold;
+}

frontend/src/utils/request.ts

@@ -92,6 +92,14 @@ class Request {
         });
       }
 
+      // 监听 AbortSignal 来中止请求
+      if (config.signal) {
+        config.signal.addEventListener("abort", () => {
+          xhr.abort();
+          reject(new Error("上传已取消"));
+        });
+      }
+
       // 监听上传进度
       xhr.upload.addEventListener("progress", function (event) {
         if (event.lengthComputable) {

runtime/datamate-python/app/module/annotation/interface/editor.py

@@ -20,6 +20,7 @@ from app.module.annotation.schema.editor import (
     EditorProjectInfo,
     EditorTaskListResponse,
     EditorTaskResponse,
+    EditorTaskSegmentsResponse,
     UpsertAnnotationRequest,
     UpsertAnnotationResponse,
 )
@@ -87,6 +88,20 @@ async def get_editor_task(
     return StandardResponse(code=200, message="success", data=task)
 
 
+@router.get(
+    "/projects/{project_id}/tasks/{file_id}/segments",
+    response_model=StandardResponse[EditorTaskSegmentsResponse],
+)
+async def list_editor_task_segments(
+    project_id: str = Path(..., description="标注项目ID（t_dm_labeling_projects.id）"),
+    file_id: str = Path(..., description="文件ID（t_dm_dataset_files.id）"),
+    db: AsyncSession = Depends(get_db),
+):
+    service = AnnotationEditorService(db)
+    result = await service.get_task_segments(project_id, file_id)
+    return StandardResponse(code=200, message="success", data=result)
+
+
 @router.put(
     "/projects/{project_id}/tasks/{file_id}/annotation",
     response_model=StandardResponse[UpsertAnnotationResponse],

runtime/datamate-python/app/module/annotation/interface/project.py

@@ -150,6 +150,18 @@ async def create_mapping(
             labeling_project, snapshot_file_ids
         )
 
+        # 如果启用了分段且为文本数据集，预生成切片结构
+        if dataset_type == TEXT_DATASET_TYPE and request.segmentation_enabled:
+            try:
+                from ..service.editor import AnnotationEditorService
+                editor_service = AnnotationEditorService(db)
+                # 异步预计算切片（不阻塞创建响应）
+                segmentation_result = await editor_service.precompute_segmentation_for_project(labeling_project.id)
+                logger.info(f"Precomputed segmentation for project {labeling_project.id}: {segmentation_result}")
+            except Exception as e:
+                logger.warning(f"Failed to precompute segmentation for project {labeling_project.id}: {e}")
+                # 不影响项目创建，只记录警告
+
         response_data = DatasetMappingCreateResponse(
             id=mapping.id,
             labeling_project_id=str(mapping.labeling_project_id),

runtime/datamate-python/app/module/annotation/schema/editor.py

@@ -79,12 +79,9 @@ class EditorTaskListResponse(BaseModel):
 
 
 class SegmentInfo(BaseModel):
-    """段落信息（用于文本分段标注）"""
+    """段落摘要（用于文本分段标注）"""
 
     idx: int = Field(..., description="段落索引")
-    text: str = Field(..., description="段落文本")
-    start: int = Field(..., description="在原文中的起始位置")
-    end: int = Field(..., description="在原文中的结束位置")
     has_annotation: bool = Field(False, alias="hasAnnotation", description="该段落是否已有标注")
     line_index: int = Field(0, alias="lineIndex", description="JSONL 行索引（从0开始）")
     chunk_index: int = Field(0, alias="chunkIndex", description="行内分片索引（从0开始）")
@@ -100,13 +97,22 @@ class EditorTaskResponse(BaseModel):
 
     # 分段相关字段
     segmented: bool = Field(False, description="是否启用分段模式")
-    segments: Optional[List[SegmentInfo]] = Field(None, description="段落列表")
     total_segments: int = Field(0, alias="totalSegments", description="总段落数")
     current_segment_index: int = Field(0, alias="currentSegmentIndex", description="当前段落索引")
 
     model_config = ConfigDict(populate_by_name=True)
 
 
+class EditorTaskSegmentsResponse(BaseModel):
+    """编辑器段落摘要响应"""
+
+    segmented: bool = Field(False, description="是否启用分段模式")
+    segments: List[SegmentInfo] = Field(default_factory=list, description="段落摘要列表")
+    total_segments: int = Field(0, alias="totalSegments", description="总段落数")
+
+    model_config = ConfigDict(populate_by_name=True)
+
+
 class UpsertAnnotationRequest(BaseModel):
     """保存/覆盖最终标注（Label Studio annotation 原始对象）"""
 

runtime/datamate-python/app/module/annotation/service/editor.py

@@ -37,6 +37,7 @@ from app.module.annotation.schema.editor import (
     EditorTaskListItem,
     EditorTaskListResponse,
     EditorTaskResponse,
+    EditorTaskSegmentsResponse,
     SegmentInfo,
     UpsertAnnotationRequest,
     UpsertAnnotationResponse,
@@ -538,6 +539,49 @@ class AnnotationEditorService:
                 return value
         return raw_text
 
+    def _build_segment_contexts(
+        self,
+        records: List[Tuple[Optional[Dict[str, Any]], str]],
+        record_texts: List[str],
+        segment_annotation_keys: set[str],
+    ) -> Tuple[List[SegmentInfo], List[Tuple[Optional[Dict[str, Any]], str, str, int, int]]]:
+        splitter = AnnotationTextSplitter(max_chars=self.SEGMENT_THRESHOLD)
+        segment_contexts: List[Tuple[Optional[Dict[str, Any]], str, str, int, int]] = []
+        segment_cursor = 0
+
+        for record_index, ((payload, raw_text), record_text) in enumerate(zip(records, record_texts)):
+            normalized_text = record_text or ""
+            if len(normalized_text) > self.SEGMENT_THRESHOLD:
+                raw_segments = splitter.split(normalized_text)
+                for chunk_index, seg in enumerate(raw_segments):
+                    segments.append(
+                        SegmentInfo(
+                            idx=segment_cursor,
+                            hasAnnotation=str(segment_cursor) in segment_annotation_keys,
+                            lineIndex=record_index,
+                            chunkIndex=chunk_index,
+                        )
+                    )
+                    segment_contexts.append((payload, raw_text, seg["text"], record_index, chunk_index))
+                    segment_cursor += 1
+            else:
+                segments.append(
+                    SegmentInfo(
+                        idx=segment_cursor,
+                        hasAnnotation=str(segment_cursor) in segment_annotation_keys,
+                        lineIndex=record_index,
+                        chunkIndex=0,
+                    )
+                )
+                segment_contexts.append((payload, raw_text, normalized_text, record_index, 0))
+                segment_cursor += 1
+
+        if not segments:
+            segments = [SegmentInfo(idx=0, hasAnnotation=False, lineIndex=0, chunkIndex=0)]
+            segment_contexts = [(None, "", "", 0, 0)]
+
+        return segments, segment_contexts
+
     async def get_project_info(self, project_id: str) -> EditorProjectInfo:
         project = await self._get_project_or_404(project_id)
 
@@ -668,6 +712,87 @@ class AnnotationEditorService:
 
         return await self._build_text_task(project, file_record, file_id, segment_index)
 
+    async def get_task_segments(
+        self,
+        project_id: str,
+        file_id: str,
+    ) -> EditorTaskSegmentsResponse:
+        project = await self._get_project_or_404(project_id)
+
+        dataset_type = self._normalize_dataset_type(await self._get_dataset_type(project.dataset_id))
+        if dataset_type != DATASET_TYPE_TEXT:
+            raise HTTPException(
+                status_code=400,
+                detail="当前仅支持 TEXT 项目的段落摘要",
+            )
+
+        file_result = await self.db.execute(
+            select(DatasetFiles).where(
+                DatasetFiles.id == file_id,
+                DatasetFiles.dataset_id == project.dataset_id,
+            )
+        )
+        file_record = file_result.scalar_one_or_none()
+        if not file_record:
+            raise HTTPException(status_code=404, detail=f"文件不存在或不属于该项目: {file_id}")
+
+        if not self._resolve_segmentation_enabled(project):
+            return EditorTaskSegmentsResponse(segmented=False, segments=[], totalSegments=0)
+
+        text_content = await self._fetch_text_content_via_download_api(project.dataset_id, file_id)
+        assert isinstance(text_content, str)
+        label_config = await self._resolve_project_label_config(project)
+        primary_text_key = self._resolve_primary_text_key(label_config)
+        file_name = str(getattr(file_record, "file_name", "")).lower()
+
+        records: List[Tuple[Optional[Dict[str, Any]], str]] = []
+        if file_name.endswith(JSONL_EXTENSION):
+            records = self._parse_jsonl_records(text_content)
+        else:
+            parsed_payload = self._try_parse_json_payload(text_content)
+            if parsed_payload:
+                records = [(parsed_payload, text_content)]
+
+        if not records:
+            records = [(None, text_content)]
+
+        record_texts = [
+            self._resolve_primary_text_value(payload, raw_text, primary_text_key)
+            for payload, raw_text in records
+        ]
+        if not record_texts:
+            record_texts = [text_content]
+
+        needs_segmentation = len(records) > 1 or any(
+            len(text or "") > self.SEGMENT_THRESHOLD for text in record_texts
+        )
+        if not needs_segmentation:
+            return EditorTaskSegmentsResponse(segmented=False, segments=[], totalSegments=0)
+
+        ann_result = await self.db.execute(
+            select(AnnotationResult).where(
+                AnnotationResult.project_id == project.id,
+                AnnotationResult.file_id == file_id,
+            )
+        )
+        ann = ann_result.scalar_one_or_none()
+        segment_annotations: Dict[str, Dict[str, Any]] = {}
+        if ann and isinstance(ann.annotation, dict):
+            segment_annotations = self._extract_segment_annotations(ann.annotation)
+        segment_annotation_keys = set(segment_annotations.keys())
+
+        segments, _ = self._build_segment_contexts(
+            records,
+            record_texts,
+            segment_annotation_keys,
+        )
+
+        return EditorTaskSegmentsResponse(
+            segmented=True,
+            segments=segments,
+            totalSegments=len(segments),
+        )
+
     async def _build_text_task(
         self,
         project: LabelingProject,
@@ -723,7 +848,8 @@ class AnnotationEditorService:
         needs_segmentation = segmentation_enabled and (
             len(records) > 1 or any(len(text or "") > self.SEGMENT_THRESHOLD for text in record_texts)
         )
-        segments: Optional[List[SegmentInfo]] = None
+        segments: List[SegmentInfo] = []
+        segment_contexts: List[Tuple[Optional[Dict[str, Any]], str, str, int, int]] = []
         current_segment_index = 0
         display_text = record_texts[0] if record_texts else text_content
         selected_payload = records[0][0] if records else None
@@ -732,46 +858,13 @@ class AnnotationEditorService:
             display_text = "\n".join(record_texts) if record_texts else text_content
 
         if needs_segmentation:
-            splitter = AnnotationTextSplitter(max_chars=self.SEGMENT_THRESHOLD)
+            _, segment_contexts = self._build_segment_contexts(
-            segment_contexts: List[Tuple[Optional[Dict[str, Any]], str, str, int, int]] = []
+                records,
-            segments = []
+                record_texts,
-            segment_cursor = 0
+                segment_annotation_keys,
-
+            )
-            for record_index, ((payload, raw_text), record_text) in enumerate(zip(records, record_texts)):
-                normalized_text = record_text or ""
-                if len(normalized_text) > self.SEGMENT_THRESHOLD:
-                    raw_segments = splitter.split(normalized_text)
-                    for chunk_index, seg in enumerate(raw_segments):
-                        segments.append(SegmentInfo(
-                            idx=segment_cursor,
-                            text=seg["text"],
-                            start=seg["start"],
-                            end=seg["end"],
-                            hasAnnotation=str(segment_cursor) in segment_annotation_keys,
-                            lineIndex=record_index,
-                            chunkIndex=chunk_index,
-                        ))
-                        segment_contexts.append((payload, raw_text, seg["text"], record_index, chunk_index))
-                        segment_cursor += 1
-                else:
-                    segments.append(SegmentInfo(
-                        idx=segment_cursor,
-                        text=normalized_text,
-                        start=0,
-                        end=len(normalized_text),
-                        hasAnnotation=str(segment_cursor) in segment_annotation_keys,
-                        lineIndex=record_index,
-                        chunkIndex=0,
-                    ))
-                    segment_contexts.append((payload, raw_text, normalized_text, record_index, 0))
-                    segment_cursor += 1
-
-            if not segments:
-                segments = [SegmentInfo(idx=0, text="", start=0, end=0, hasAnnotation=False, lineIndex=0, chunkIndex=0)]
-                segment_contexts = [(None, "", "", 0, 0)]
-
             current_segment_index = segment_index if segment_index is not None else 0
-            if current_segment_index < 0 or current_segment_index >= len(segments):
+            if current_segment_index < 0 or current_segment_index >= len(segment_contexts):
                 current_segment_index = 0
 
             selected_payload, _, display_text, _, _ = segment_contexts[current_segment_index]
@@ -849,8 +942,7 @@ class AnnotationEditorService:
             task=task,
             annotationUpdatedAt=annotation_updated_at,
             segmented=needs_segmentation,
-            segments=segments,
+            totalSegments=len(segment_contexts) if needs_segmentation else 1,
-            totalSegments=len(segments) if segments else 1,
             currentSegmentIndex=current_segment_index,
         )
 
@@ -1185,3 +1277,195 @@ class AnnotationEditorService:
         except Exception as exc:
             logger.warning("标注同步知识管理失败：%s", exc)
 
+    async def precompute_segmentation_for_project(
+        self,
+        project_id: str,
+        max_retries: int = 3
+    ) -> Dict[str, Any]:
+        """
+        为指定项目的所有文本文件预计算切片结构并持久化到数据库
+
+        Args:
+            project_id: 标注项目ID
+            max_retries: 失败重试次数
+
+        Returns:
+            统计信息：{total_files, succeeded, failed}
+        """
+        project = await self._get_project_or_404(project_id)
+        dataset_type = self._normalize_dataset_type(await self._get_dataset_type(project.dataset_id))
+
+        # 只处理文本数据集
+        if dataset_type != DATASET_TYPE_TEXT:
+            logger.info(f"项目 {project_id} 不是文本数据集，跳过切片预生成")
+            return {"total_files": 0, "succeeded": 0, "failed": 0}
+
+        # 检查是否启用分段
+        if not self._resolve_segmentation_enabled(project):
+            logger.info(f"项目 {project_id} 未启用分段，跳过切片预生成")
+            return {"total_files": 0, "succeeded": 0, "failed": 0}
+
+        # 获取项目的所有文本文件（排除源文档）
+        files_result = await self.db.execute(
+            select(DatasetFiles)
+            .join(LabelingProjectFile, LabelingProjectFile.file_id == DatasetFiles.id)
+            .where(
+                LabelingProjectFile.project_id == project_id,
+                DatasetFiles.dataset_id == project.dataset_id,
+            )
+        )
+        file_records = files_result.scalars().all()
+
+        if not file_records:
+            logger.info(f"项目 {project_id} 没有文件，跳过切片预生成")
+            return {"total_files": 0, "succeeded": 0, "failed": 0}
+
+        # 过滤源文档文件
+        valid_files = []
+        for file_record in file_records:
+            file_type = str(getattr(file_record, "file_type", "") or "").lower()
+            file_name = str(getattr(file_record, "file_name", "")).lower()
+            is_source_document = (
+                file_type in SOURCE_DOCUMENT_TYPES or
+                any(file_name.endswith(ext) for ext in SOURCE_DOCUMENT_EXTENSIONS)
+            )
+            if not is_source_document:
+                valid_files.append(file_record)
+
+        total_files = len(valid_files)
+        succeeded = 0
+        failed = 0
+
+        label_config = await self._resolve_project_label_config(project)
+        primary_text_key = self._resolve_primary_text_key(label_config)
+
+        for file_record in valid_files:
+            file_id = str(file_record.id)  # type: ignore
+            file_name = str(getattr(file_record, "file_name", ""))
+
+            for retry in range(max_retries):
+                try:
+                    # 读取文本内容
+                    text_content = await self._fetch_text_content_via_download_api(project.dataset_id, file_id)
+                    if not isinstance(text_content, str):
+                        logger.warning(f"文件 {file_id} 内容不是字符串，跳过切片")
+                        failed += 1
+                        break
+
+                    # 解析文本记录
+                    records: List[Tuple[Optional[Dict[str, Any]], str]] = []
+                    if file_name.lower().endswith(JSONL_EXTENSION):
+                        records = self._parse_jsonl_records(text_content)
+                    else:
+                        parsed_payload = self._try_parse_json_payload(text_content)
+                        if parsed_payload:
+                            records = [(parsed_payload, text_content)]
+
+                    if not records:
+                        records = [(None, text_content)]
+
+                    record_texts = [
+                        self._resolve_primary_text_value(payload, raw_text, primary_text_key)
+                        for payload, raw_text in records
+                    ]
+                    if not record_texts:
+                        record_texts = [text_content]
+
+                    # 判断是否需要分段
+                    needs_segmentation = len(records) > 1 or any(
+                        len(text or "") > self.SEGMENT_THRESHOLD for text in record_texts
+                    )
+
+                    if not needs_segmentation:
+                        # 不需要分段的文件，跳过
+                        succeeded += 1
+                        break
+
+                    # 执行切片
+                    splitter = AnnotationTextSplitter(max_chars=self.SEGMENT_THRESHOLD)
+                    segment_cursor = 0
+                    segments = {}
+
+                    for record_index, ((payload, raw_text), record_text) in enumerate(zip(records, record_texts)):
+                        normalized_text = record_text or ""
+
+                        if len(normalized_text) > self.SEGMENT_THRESHOLD:
+                            raw_segments = splitter.split(normalized_text)
+                            for chunk_index, seg in enumerate(raw_segments):
+                                segments[str(segment_cursor)] = {
+                                    SEGMENT_RESULT_KEY: [],
+                                    SEGMENT_CREATED_AT_KEY: datetime.utcnow().isoformat() + "Z",
+                                    SEGMENT_UPDATED_AT_KEY: datetime.utcnow().isoformat() + "Z",
+                                }
+                                segment_cursor += 1
+                        else:
+                            segments[str(segment_cursor)] = {
+                                SEGMENT_RESULT_KEY: [],
+                                SEGMENT_CREATED_AT_KEY: datetime.utcnow().isoformat() + "Z",
+                                SEGMENT_UPDATED_AT_KEY: datetime.utcnow().isoformat() + "Z",
+                            }
+                            segment_cursor += 1
+
+                    if not segments:
+                        succeeded += 1
+                        break
+
+                    # 构造分段标注结构
+                    final_payload = {
+                        SEGMENTED_KEY: True,
+                        "version": 1,
+                        SEGMENTS_KEY: segments,
+                        SEGMENT_TOTAL_KEY: segment_cursor,
+                    }
+
+                    # 检查是否已存在标注
+                    existing_result = await self.db.execute(
+                        select(AnnotationResult).where(
+                            AnnotationResult.project_id == project_id,
+                            AnnotationResult.file_id == file_id,
+                        )
+                    )
+                    existing = existing_result.scalar_one_or_none()
+
+                    now = datetime.utcnow()
+
+                    if existing:
+                        # 更新现有标注
+                        existing.annotation = final_payload  # type: ignore[assignment]
+                        existing.annotation_status = ANNOTATION_STATUS_IN_PROGRESS  # type: ignore[assignment]
+                        existing.updated_at = now  # type: ignore[assignment]
+                    else:
+                        # 创建新标注记录
+                        record = AnnotationResult(
+                            id=str(uuid.uuid4()),
+                            project_id=project_id,
+                            file_id=file_id,
+                            annotation=final_payload,
+                            annotation_status=ANNOTATION_STATUS_IN_PROGRESS,
+                            created_at=now,
+                            updated_at=now,
+                        )
+                        self.db.add(record)
+
+                    await self.db.commit()
+                    succeeded += 1
+                    logger.info(f"成功为文件 {file_id} 预生成 {segment_cursor} 个切片")
+                    break
+
+                except Exception as e:
+                    logger.warning(
+                        f"为文件 {file_id} 预生成切片失败 (重试 {retry + 1}/{max_retries}): {e}"
+                    )
+                    if retry == max_retries - 1:
+                        failed += 1
+                        await self.db.rollback()
+
+        logger.info(
+            f"项目 {project_id} 切片预生成完成: 总计 {total_files}, 成功 {succeeded}, 失败 {failed}"
+        )
+        return {
+            "total_files": total_files,
+            "succeeded": succeeded,
+            "failed": failed,
+        }
+