feat(data-management): 实现数据集文件版本管理和内部路径保护

- 将数据集文件查询方法替换为只查询可见文件的版本
- 引入文件状态管理（ACTIVE/ARCHIVED）和内部目录结构
- 实现文件重复处理策略，支持版本控制模式而非覆盖
- 添加内部数据目录保护，防止访问.datamate等系统目录
- 重构文件上传流程，引入暂存目录和事务后清理机制
- 实现文件版本归档功能，保留历史版本到专用存储位置
- 优化文件路径规范化和安全验证逻辑
- 修复文件删除逻辑，确保归档文件不会被错误移除
- 更新数据集压缩下载功能以排除内部系统文件

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());
         }
@@ -113,13 +115,31 @@ public class DatasetApplicationService {
         if (Objects.nonNull(updateDatasetRequest.getStatus())) {
             dataset.setStatus(updateDatasetRequest.getStatus());
         }
-        if (updateDatasetRequest.getParentDatasetId() != null) {
+        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;
     }
@@ -144,7 +164,7 @@ public class DatasetApplicationService {
     public Dataset getDataset(String datasetId) {
         Dataset dataset = datasetRepository.getById(datasetId);
         BusinessAssert.notNull(dataset, DataManagementErrorCode.DATASET_NOT_FOUND);
-        List<DatasetFile> datasetFiles = datasetFileRepository.findAllByDatasetId(datasetId);
+        List<DatasetFile> datasetFiles = datasetFileRepository.findAllVisibleByDatasetId(datasetId);
         dataset.setFiles(datasetFiles);
         applyVisibleFileCounts(Collections.singletonList(dataset));
         return dataset;
@@ -419,7 +439,7 @@ public class DatasetApplicationService {
 
         Map<String, Object> statistics = new HashMap<>();
 
-        List<DatasetFile> allFiles = datasetFileRepository.findAllByDatasetId(datasetId);
+        List<DatasetFile> allFiles = datasetFileRepository.findAllVisibleByDatasetId(datasetId);
         List<DatasetFile> visibleFiles = filterVisibleFiles(allFiles);
         long totalFiles = visibleFiles.size();
         long completedFiles = visibleFiles.stream()

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

@@ -58,7 +58,6 @@ import java.time.LocalDateTime;
 import java.time.ZoneId;
 import java.time.format.DateTimeFormatter;
 import java.util.*;
-import java.util.concurrent.CompletableFuture;
 import java.util.function.Function;
 import java.util.stream.Collectors;
 import java.util.stream.Stream;
@@ -83,6 +82,11 @@ public class DatasetFileApplicationService {
         XLSX_FILE_TYPE
     );
     private static final String DERIVED_METADATA_KEY = "derived_from_file_id";
+    private static final String FILE_STATUS_ACTIVE = "ACTIVE";
+    private static final String FILE_STATUS_ARCHIVED = "ARCHIVED";
+    private static final String INTERNAL_DIR_NAME = ".datamate";
+    private static final String INTERNAL_UPLOAD_DIR_NAME = "uploading";
+    private static final String INTERNAL_VERSIONS_DIR_NAME = "versions";
 
     private final DatasetFileRepository datasetFileRepository;
     private final DatasetRepository datasetRepository;
@@ -93,7 +97,7 @@ public class DatasetFileApplicationService {
     @Value("${datamate.data-management.base-path:/dataset}")
     private String datasetBasePath;
 
-    @Value("${datamate.data-management.file.duplicate:COVER}")
+    @Value("${datamate.data-management.file.duplicate:VERSION}")
     private DuplicateMethod duplicateMethod;
 
     @Autowired
@@ -162,9 +166,19 @@ public class DatasetFileApplicationService {
         if (dataset == null) {
             return PagedResponse.of(new Page<>(page, size));
         }
-        String datasetPath = dataset.getPath();
+        Path datasetRoot = Paths.get(dataset.getPath()).toAbsolutePath().normalize();
-        Path queryPath = Path.of(dataset.getPath() + File.separator + prefix);
+        prefix = Optional.ofNullable(prefix).orElse("").trim().replace("\\", "/");
-        Map<String, DatasetFile> datasetFilesMap = datasetFileRepository.findAllByDatasetId(datasetId)
+        while (prefix.startsWith("/")) {
+            prefix = prefix.substring(1);
+        }
+        if (prefix.equals(INTERNAL_DIR_NAME) || prefix.startsWith(INTERNAL_DIR_NAME + "/")) {
+            return new PagedResponse<>(page, size, 0, 0, Collections.emptyList());
+        }
+        Path queryPath = datasetRoot.resolve(prefix.replace("/", File.separator)).normalize();
+        if (!queryPath.startsWith(datasetRoot)) {
+            return new PagedResponse<>(page, size, 0, 0, Collections.emptyList());
+        }
+        Map<String, DatasetFile> datasetFilesMap = datasetFileRepository.findAllVisibleByDatasetId(datasetId)
             .stream()
             .filter(file -> file.getFilePath() != null)
             .collect(Collectors.toMap(
@@ -186,7 +200,8 @@ public class DatasetFileApplicationService {
         }
         try (Stream<Path> pathStream = Files.list(queryPath)) {
             List<Path> allFiles = pathStream
-                .filter(path -> path.toString().startsWith(datasetPath))
+                .filter(path -> path.toAbsolutePath().normalize().startsWith(datasetRoot))
+                .filter(path -> !isInternalDatasetPath(datasetRoot, path))
                 .filter(path -> !excludeDerivedFiles
                     || Files.isDirectory(path)
                     || !derivedFilePaths.contains(normalizeFilePath(path.toString())))
@@ -298,6 +313,86 @@ public class DatasetFileApplicationService {
         }
     }
 
+    private boolean isSameNormalizedPath(String left, String right) {
+        String normalizedLeft = normalizeFilePath(left);
+        String normalizedRight = normalizeFilePath(right);
+        if (normalizedLeft == null || normalizedRight == null) {
+            return false;
+        }
+        return normalizedLeft.equals(normalizedRight);
+    }
+
+    private boolean isInternalDatasetPath(Path datasetRoot, Path path) {
+        if (datasetRoot == null || path == null) {
+            return false;
+        }
+        try {
+            Path normalizedRoot = datasetRoot.toAbsolutePath().normalize();
+            Path normalizedPath = path.toAbsolutePath().normalize();
+            if (!normalizedPath.startsWith(normalizedRoot)) {
+                return false;
+            }
+            Path relative = normalizedRoot.relativize(normalizedPath);
+            if (relative.getNameCount() == 0) {
+                return false;
+            }
+            return INTERNAL_DIR_NAME.equals(relative.getName(0).toString());
+        } catch (Exception e) {
+            return false;
+        }
+    }
+
+    private String normalizeLogicalPrefix(String prefix) {
+        if (prefix == null) {
+            return "";
+        }
+        String normalized = prefix.trim().replace("\\", "/");
+        while (normalized.startsWith("/")) {
+            normalized = normalized.substring(1);
+        }
+        while (normalized.endsWith("/")) {
+            normalized = normalized.substring(0, normalized.length() - 1);
+        }
+        while (normalized.contains("//")) {
+            normalized = normalized.replace("//", "/");
+        }
+        return normalized;
+    }
+
+    private String normalizeLogicalPath(String logicalPath) {
+        return normalizeLogicalPrefix(logicalPath);
+    }
+
+    private String joinLogicalPath(String prefix, String relativePath) {
+        String normalizedPrefix = normalizeLogicalPrefix(prefix);
+        String normalizedRelative = normalizeLogicalPath(relativePath);
+        if (normalizedPrefix.isEmpty()) {
+            return normalizedRelative;
+        }
+        if (normalizedRelative.isEmpty()) {
+            return normalizedPrefix;
+        }
+        return normalizeLogicalPath(normalizedPrefix + "/" + normalizedRelative);
+    }
+
+    private void assertNotInternalPrefix(String prefix) {
+        if (prefix == null || prefix.isBlank()) {
+            return;
+        }
+        String normalized = normalizeLogicalPrefix(prefix);
+        if (normalized.equals(INTERNAL_DIR_NAME) || normalized.startsWith(INTERNAL_DIR_NAME + "/")) {
+            throw BusinessException.of(CommonErrorCode.PARAM_ERROR);
+        }
+    }
+
+    private boolean isArchivedStatus(DatasetFile datasetFile) {
+        if (datasetFile == null) {
+            return false;
+        }
+        String status = datasetFile.getStatus();
+        return status != null && FILE_STATUS_ARCHIVED.equalsIgnoreCase(status);
+    }
+
     private boolean isSourceDocument(DatasetFile datasetFile) {
         if (datasetFile == null) {
             return false;
@@ -327,6 +422,144 @@ public class DatasetFileApplicationService {
         }
     }
 
+    private Path resolveDatasetRootPath(Dataset dataset, String datasetId) {
+        String datasetPath = dataset == null ? null : dataset.getPath();
+        if (datasetPath == null || datasetPath.isBlank()) {
+            datasetPath = datasetBasePath + File.separator + datasetId;
+            if (dataset != null) {
+                dataset.setPath(datasetPath);
+                datasetRepository.updateById(dataset);
+            }
+        }
+        Path datasetRoot = Paths.get(datasetPath).toAbsolutePath().normalize();
+        try {
+            Files.createDirectories(datasetRoot);
+        } catch (IOException e) {
+            log.error("Failed to create dataset root dir: {}", datasetRoot, e);
+            throw BusinessException.of(SystemErrorCode.FILE_SYSTEM_ERROR);
+        }
+        return datasetRoot;
+    }
+
+    private Path resolveStagingRootPath(Path datasetRoot,
+                                        DatasetFileUploadCheckInfo checkInfo,
+                                        List<FileUploadResult> uploadedFiles) {
+        if (datasetRoot == null) {
+            return null;
+        }
+        String stagingPath = checkInfo == null ? null : checkInfo.getStagingPath();
+        if (stagingPath != null && !stagingPath.isBlank()) {
+            try {
+                Path stagingRoot = Paths.get(stagingPath).toAbsolutePath().normalize();
+                if (!stagingRoot.startsWith(datasetRoot)) {
+                    log.warn("Staging root out of dataset root, datasetId={}, stagingRoot={}, datasetRoot={}",
+                        checkInfo == null ? null : checkInfo.getDatasetId(), stagingRoot, datasetRoot);
+                    return null;
+                }
+                Path relative = datasetRoot.relativize(stagingRoot);
+                if (relative.getNameCount() < 3) {
+                    return null;
+                }
+                if (!INTERNAL_DIR_NAME.equals(relative.getName(0).toString())
+                    || !INTERNAL_UPLOAD_DIR_NAME.equals(relative.getName(1).toString())) {
+                    return null;
+                }
+                return stagingRoot;
+            } catch (Exception e) {
+                log.warn("Invalid staging path: {}", stagingPath, e);
+                return null;
+            }
+        }
+        if (uploadedFiles == null || uploadedFiles.isEmpty()) {
+            return null;
+        }
+        FileUploadResult firstResult = uploadedFiles.get(0);
+        File firstFile = firstResult == null ? null : firstResult.getSavedFile();
+        if (firstFile == null) {
+            return null;
+        }
+        try {
+            return Paths.get(firstFile.getParent()).toAbsolutePath().normalize();
+        } catch (Exception e) {
+            return null;
+        }
+    }
+
+    private void scheduleCleanupStagingDirAfterCommit(Path stagingRoot) {
+        if (stagingRoot == null) {
+            return;
+        }
+        Runnable cleanup = () -> deleteDirectoryRecursivelyQuietly(stagingRoot);
+        if (TransactionSynchronizationManager.isSynchronizationActive()) {
+            TransactionSynchronizationManager.registerSynchronization(new TransactionSynchronization() {
+                @Override
+                public void afterCommit() {
+                    cleanup.run();
+                }
+            });
+            return;
+        }
+        cleanup.run();
+    }
+
+    private void deleteDirectoryRecursivelyQuietly(Path directory) {
+        if (directory == null) {
+            return;
+        }
+        if (!Files.exists(directory)) {
+            return;
+        }
+        try (Stream<Path> paths = Files.walk(directory)) {
+            paths.sorted(Comparator.reverseOrder()).forEach(path -> {
+                try {
+                    Files.deleteIfExists(path);
+                } catch (IOException e) {
+                    log.debug("Failed to delete: {}", path, e);
+                }
+            });
+        } catch (IOException e) {
+            log.debug("Failed to cleanup staging dir: {}", directory, e);
+        }
+    }
+
+    private String sanitizeArchiveFileName(String fileName) {
+        String input = fileName == null ? "" : fileName.trim();
+        if (input.isBlank()) {
+            return "file";
+        }
+        StringBuilder builder = new StringBuilder(input.length());
+        for (int i = 0; i < input.length(); i++) {
+            char c = input.charAt(i);
+            if (c <= 31 || c == 127) {
+                builder.append('_');
+                continue;
+            }
+            if (c == '/' || c == '\\' || c == ':' || c == '*' || c == '?' || c == '\"'
+                || c == '<' || c == '>' || c == '|') {
+                builder.append('_');
+                continue;
+            }
+            builder.append(c);
+        }
+        String sanitized = builder.toString().trim();
+        return sanitized.isEmpty() ? "file" : sanitized;
+    }
+
+    private String sha256Hex(String value) {
+        String input = value == null ? "" : value;
+        try {
+            java.security.MessageDigest digest = java.security.MessageDigest.getInstance("SHA-256");
+            byte[] hashed = digest.digest(input.getBytes(java.nio.charset.StandardCharsets.UTF_8));
+            StringBuilder builder = new StringBuilder(hashed.length * 2);
+            for (byte b : hashed) {
+                builder.append(String.format("%02x", b));
+            }
+            return builder.toString();
+        } catch (Exception e) {
+            return Integer.toHexString(input.hashCode());
+        }
+    }
+
     /**
      * 获取文件详情
      */
@@ -349,10 +582,12 @@ public class DatasetFileApplicationService {
     public void deleteDatasetFile(String datasetId, String fileId) {
         DatasetFile file = getDatasetFile(datasetId, fileId);
         Dataset dataset = datasetRepository.getById(datasetId);
-        dataset.setFiles(new ArrayList<>(Collections.singleton(file)));
         datasetFileRepository.removeById(fileId);
+        if (!isArchivedStatus(file)) {
+            dataset.setFiles(new ArrayList<>(Collections.singleton(file)));
             dataset.removeFile(file);
             datasetRepository.updateById(dataset);
+        }
         datasetFilePreviewService.deletePreviewFileQuietly(datasetId, fileId);
         // 删除文件时，上传到数据集中的文件会同时删除数据库中的记录和文件系统中的文件，归集过来的文件仅删除数据库中的记录
         if (file.getFilePath().startsWith(dataset.getPath())) {
@@ -393,18 +628,26 @@ public class DatasetFileApplicationService {
         if (Objects.isNull(dataset)) {
             throw BusinessException.of(DataManagementErrorCode.DATASET_NOT_FOUND);
         }
-        List<DatasetFile> allByDatasetId = datasetFileRepository.findAllByDatasetId(datasetId);
+        Path datasetRoot = Paths.get(dataset.getPath()).toAbsolutePath().normalize();
-        Set<String> filePaths = allByDatasetId.stream().map(DatasetFile::getFilePath).collect(Collectors.toSet());
+        Set<Path> filePaths = datasetFileRepository.findAllVisibleByDatasetId(datasetId).stream()
-        String datasetPath = dataset.getPath();
+            .map(DatasetFile::getFilePath)
-        Path downloadPath = Path.of(datasetPath);
+            .filter(Objects::nonNull)
+            .map(path -> Paths.get(path).toAbsolutePath().normalize())
+            .filter(path -> path.startsWith(datasetRoot))
+            .filter(path -> !isInternalDatasetPath(datasetRoot, path))
+            .collect(Collectors.toSet());
+        Path downloadPath = datasetRoot;
         response.setContentType("application/zip");
         String zipName = String.format("dataset_%s.zip",
                 LocalDateTime.now().format(DateTimeFormatter.ofPattern("yyyyMMddHHmmss")));
         response.setHeader(HttpHeaders.CONTENT_DISPOSITION, "attachment; filename=" + zipName);
         try (ZipArchiveOutputStream zos = new ZipArchiveOutputStream(response.getOutputStream())) {
             try (Stream<Path> pathStream = Files.walk(downloadPath)) {
-                List<Path> allPaths = pathStream.filter(path -> path.toString().startsWith(datasetPath))
+                List<Path> allPaths = pathStream
-                    .filter(path -> filePaths.stream().anyMatch(filePath -> filePath.startsWith(path.toString())))
+                    .map(path -> path.toAbsolutePath().normalize())
+                    .filter(path -> path.startsWith(datasetRoot))
+                    .filter(path -> !isInternalDatasetPath(datasetRoot, path))
+                    .filter(path -> filePaths.stream().anyMatch(filePath -> filePath.startsWith(path)))
                     .toList();
                 for (Path path : allPaths) {
                     addToZipFile(path, downloadPath, zos);
@@ -461,29 +704,33 @@ public class DatasetFileApplicationService {
             throw BusinessException.of(DataManagementErrorCode.DATASET_NOT_FOUND);
         }
 
-        // 构建上传路径，如果有 prefix 则追加到路径中
+        String prefix = normalizeLogicalPrefix(chunkUploadRequest == null ? null : chunkUploadRequest.getPrefix());
-        String prefix = Optional.ofNullable(chunkUploadRequest.getPrefix()).orElse("").trim();
+        assertNotInternalPrefix(prefix);
-        prefix = prefix.replace("\\", "/");
-        while (prefix.startsWith("/")) {
-            prefix = prefix.substring(1);
-        }
 
-        String uploadPath = dataset.getPath();
+        Path datasetRoot = resolveDatasetRootPath(dataset, datasetId);
-        if (uploadPath == null || uploadPath.isBlank()) {
+        Path stagingRoot = datasetRoot
-            uploadPath = datasetBasePath + File.separator + datasetId;
+            .resolve(INTERNAL_DIR_NAME)
-        }
+            .resolve(INTERNAL_UPLOAD_DIR_NAME)
-        if (!prefix.isEmpty()) {
+            .resolve(UUID.randomUUID().toString())
-            uploadPath = uploadPath + File.separator + prefix.replace("/", File.separator);
+            .toAbsolutePath()
+            .normalize();
+        BusinessAssert.isTrue(stagingRoot.startsWith(datasetRoot), CommonErrorCode.PARAM_ERROR);
+        try {
+            Files.createDirectories(stagingRoot);
+        } catch (IOException e) {
+            log.error("Failed to create staging dir: {}", stagingRoot, e);
+            throw BusinessException.of(SystemErrorCode.FILE_SYSTEM_ERROR);
         }
 
         ChunkUploadPreRequest request = ChunkUploadPreRequest.builder().build();
-        request.setUploadPath(uploadPath);
+        request.setUploadPath(stagingRoot.toString());
         request.setTotalFileNum(chunkUploadRequest.getTotalFileNum());
         request.setServiceId(DatasetConstant.SERVICE_ID);
         DatasetFileUploadCheckInfo checkInfo = new DatasetFileUploadCheckInfo();
         checkInfo.setDatasetId(datasetId);
         checkInfo.setHasArchive(chunkUploadRequest.isHasArchive());
         checkInfo.setPrefix(prefix);
+        checkInfo.setStagingPath(stagingRoot.toString());
         try {
             ObjectMapper objectMapper = new ObjectMapper();
             String checkInfoJson = objectMapper.writeValueAsString(checkInfo);
@@ -505,6 +752,14 @@ 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
@@ -527,32 +782,251 @@ public class DatasetFileApplicationService {
         } else {
             files = Collections.singletonList(fileUploadResult);
         }
-        addFileToDataset(datasetId, files);
+        commitUploadedFiles(datasetId, checkInfo, files, fileUploadResult.isAllFilesUploaded());
     }
 
-    private void addFileToDataset(String datasetId, List<FileUploadResult> unpacked) {
+    private void commitUploadedFiles(String datasetId,
+                                     DatasetFileUploadCheckInfo checkInfo,
+                                     List<FileUploadResult> uploadedFiles,
+                                     boolean cleanupStagingAfterCommit) {
         Dataset dataset = datasetRepository.getById(datasetId);
-        dataset.setFiles(datasetFileRepository.findAllByDatasetId(datasetId));
+        BusinessAssert.notNull(dataset, DataManagementErrorCode.DATASET_NOT_FOUND);
-        for (FileUploadResult file : unpacked) {
+
-            File savedFile = file.getSavedFile();
+        Path datasetRoot = resolveDatasetRootPath(dataset, datasetId);
+        String prefix = checkInfo == null ? "" : normalizeLogicalPrefix(checkInfo.getPrefix());
+        assertNotInternalPrefix(prefix);
+
+        Path stagingRoot = resolveStagingRootPath(datasetRoot, checkInfo, uploadedFiles);
+        BusinessAssert.notNull(stagingRoot, CommonErrorCode.PARAM_ERROR);
+
+        dataset.setFiles(new ArrayList<>(datasetFileRepository.findAllVisibleByDatasetId(datasetId)));
+        for (FileUploadResult fileResult : uploadedFiles) {
+            commitSingleUploadedFile(dataset, datasetRoot, stagingRoot, prefix, fileResult);
+        }
+
+        dataset.active();
+        datasetRepository.updateById(dataset);
+
+        if (cleanupStagingAfterCommit) {
+            scheduleCleanupStagingDirAfterCommit(stagingRoot);
+        }
+    }
+
+    private void commitSingleUploadedFile(Dataset dataset,
+                                          Path datasetRoot,
+                                          Path stagingRoot,
+                                          String prefix,
+                                          FileUploadResult fileResult) {
+        if (dataset == null || fileResult == null || fileResult.getSavedFile() == null) {
+            return;
+        }
+        Path incomingPath = Paths.get(fileResult.getSavedFile().getPath()).toAbsolutePath().normalize();
+        BusinessAssert.isTrue(incomingPath.startsWith(stagingRoot), CommonErrorCode.PARAM_ERROR);
+
+        String relativePath = stagingRoot.relativize(incomingPath).toString().replace(File.separator, "/");
+        String logicalPath = joinLogicalPath(prefix, relativePath);
+        assertNotInternalPrefix(logicalPath);
+
+        commitNewFileVersion(dataset, datasetRoot, logicalPath, incomingPath, true);
+    }
+
+    private DatasetFile commitNewFileVersion(Dataset dataset,
+                                             Path datasetRoot,
+                                             String logicalPath,
+                                             Path incomingFilePath,
+                                             boolean moveIncoming) {
+        BusinessAssert.notNull(dataset, CommonErrorCode.PARAM_ERROR);
+        BusinessAssert.isTrue(datasetRoot != null && Files.exists(datasetRoot), CommonErrorCode.PARAM_ERROR);
+
+        String normalizedLogicalPath = normalizeLogicalPath(logicalPath);
+        BusinessAssert.isTrue(!normalizedLogicalPath.isEmpty(), CommonErrorCode.PARAM_ERROR);
+        assertNotInternalPrefix(normalizedLogicalPath);
+
+        Path targetFilePath = datasetRoot.resolve(normalizedLogicalPath.replace("/", File.separator))
+            .toAbsolutePath()
+            .normalize();
+        BusinessAssert.isTrue(targetFilePath.startsWith(datasetRoot), CommonErrorCode.PARAM_ERROR);
+
+        DuplicateMethod effectiveDuplicateMethod = resolveEffectiveDuplicateMethod();
+        DatasetFile latest = datasetFileRepository.findLatestByDatasetIdAndLogicalPath(dataset.getId(), normalizedLogicalPath);
+        if (latest == null && dataset.getFiles() != null) {
+            latest = dataset.getFiles().stream()
+                .filter(existing -> isSameNormalizedPath(existing == null ? null : existing.getFilePath(), targetFilePath.toString()))
+                .findFirst()
+                .orElse(null);
+        }
+        if (latest != null && effectiveDuplicateMethod == DuplicateMethod.ERROR) {
+            throw BusinessException.of(DataManagementErrorCode.DATASET_FILE_ALREADY_EXISTS);
+        }
+
+        long nextVersion = 1L;
+        if (latest != null) {
+            long latestVersion = Optional.ofNullable(latest.getVersion()).orElse(1L);
+            if (latest.getVersion() == null) {
+                latest.setVersion(latestVersion);
+            }
+            if (latest.getLogicalPath() == null || latest.getLogicalPath().isBlank()) {
+                latest.setLogicalPath(normalizedLogicalPath);
+            }
+            nextVersion = latestVersion + 1L;
+        }
+
+        if (latest != null && effectiveDuplicateMethod == DuplicateMethod.VERSION) {
+            Path archivedPath = archiveDatasetFileVersion(datasetRoot, normalizedLogicalPath, latest);
+            if (archivedPath != null) {
+                latest.setFilePath(archivedPath.toString());
+            } else if (Files.exists(targetFilePath)) {
+                log.error("Failed to archive latest file, refuse to overwrite. datasetId={}, fileId={}, logicalPath={}, targetPath={}",
+                    dataset.getId(), latest.getId(), normalizedLogicalPath, targetFilePath);
+                throw BusinessException.of(SystemErrorCode.FILE_SYSTEM_ERROR);
+            }
+            latest.setStatus(FILE_STATUS_ARCHIVED);
+            datasetFileRepository.updateById(latest);
+            dataset.removeFile(latest);
+        } else if (latest == null && Files.exists(targetFilePath)) {
+            archiveOrphanTargetFile(datasetRoot, normalizedLogicalPath, targetFilePath);
+        }
+
+        try {
+            Files.createDirectories(targetFilePath.getParent());
+            if (moveIncoming) {
+                Files.move(incomingFilePath, targetFilePath, java.nio.file.StandardCopyOption.REPLACE_EXISTING);
+            } else {
+                Files.copy(incomingFilePath, targetFilePath, java.nio.file.StandardCopyOption.REPLACE_EXISTING);
+            }
+        } catch (IOException e) {
+            log.error("Failed to write dataset file, datasetId={}, logicalPath={}, targetPath={}",
+                dataset.getId(), normalizedLogicalPath, targetFilePath, e);
+            throw BusinessException.of(SystemErrorCode.FILE_SYSTEM_ERROR);
+        }
+
         LocalDateTime currentTime = LocalDateTime.now();
+        String fileName = targetFilePath.getFileName().toString();
+        long fileSize;
+        try {
+            fileSize = Files.size(targetFilePath);
+        } catch (IOException e) {
+            fileSize = 0L;
+        }
+
         DatasetFile datasetFile = DatasetFile.builder()
             .id(UUID.randomUUID().toString())
-                .datasetId(datasetId)
+            .datasetId(dataset.getId())
-                .fileSize(savedFile.length())
+            .fileName(fileName)
+            .fileType(AnalyzerUtils.getExtension(fileName))
+            .fileSize(fileSize)
+            .filePath(targetFilePath.toString())
+            .logicalPath(normalizedLogicalPath)
+            .version(nextVersion)
+            .status(FILE_STATUS_ACTIVE)
             .uploadTime(currentTime)
             .lastAccessTime(currentTime)
-                .fileName(file.getFileName())
-                .filePath(savedFile.getPath())
-                .fileType(AnalyzerUtils.getExtension(file.getFileName()))
             .build();
-            setDatasetFileId(datasetFile, dataset);
+
         datasetFileRepository.saveOrUpdate(datasetFile);
         dataset.addFile(datasetFile);
         triggerPdfTextExtraction(dataset, datasetFile);
+        return datasetFile;
+    }
+
+    private DuplicateMethod resolveEffectiveDuplicateMethod() {
+        if (duplicateMethod == null) {
+            return DuplicateMethod.VERSION;
+        }
+        if (duplicateMethod == DuplicateMethod.COVER) {
+            log.warn("duplicateMethod=COVER 会导致标注引用的 fileId 对应内容被覆盖，已强制按 VERSION 处理。");
+            return DuplicateMethod.VERSION;
+        }
+        return duplicateMethod;
+    }
+
+    private Path archiveDatasetFileVersion(Path datasetRoot, String logicalPath, DatasetFile latest) {
+        if (latest == null || latest.getId() == null || latest.getId().isBlank()) {
+            return null;
+        }
+        Path currentPath;
+        try {
+            currentPath = Paths.get(latest.getFilePath()).toAbsolutePath().normalize();
+        } catch (Exception e) {
+            log.warn("Invalid latest file path, skip archiving. datasetId={}, fileId={}, filePath={}",
+                latest.getDatasetId(), latest.getId(), latest.getFilePath());
+            return null;
+        }
+
+        if (!Files.exists(currentPath) || !Files.isRegularFile(currentPath)) {
+            log.warn("Latest file not found on disk, skip archiving. datasetId={}, fileId={}, filePath={}",
+                latest.getDatasetId(), latest.getId(), currentPath);
+            return null;
+        }
+        if (!currentPath.startsWith(datasetRoot)) {
+            log.warn("Latest file path out of dataset root, skip archiving. datasetId={}, fileId={}, filePath={}",
+                latest.getDatasetId(), latest.getId(), currentPath);
+            return null;
+        }
+
+        long latestVersion = Optional.ofNullable(latest.getVersion()).orElse(1L);
+        String logicalPathHash = sha256Hex(logicalPath);
+        Path archiveDir = datasetRoot
+            .resolve(INTERNAL_DIR_NAME)
+            .resolve(INTERNAL_VERSIONS_DIR_NAME)
+            .resolve(logicalPathHash)
+            .resolve("v" + latestVersion)
+            .toAbsolutePath()
+            .normalize();
+        BusinessAssert.isTrue(archiveDir.startsWith(datasetRoot), CommonErrorCode.PARAM_ERROR);
+
+        try {
+            Files.createDirectories(archiveDir);
+        } catch (IOException e) {
+            log.error("Failed to create archive dir: {}", archiveDir, e);
+            throw BusinessException.of(SystemErrorCode.FILE_SYSTEM_ERROR);
+        }
+
+        String fileName = sanitizeArchiveFileName(Optional.ofNullable(latest.getFileName()).orElse(currentPath.getFileName().toString()));
+        Path archivedPath = archiveDir.resolve(latest.getId() + "__" + fileName).toAbsolutePath().normalize();
+        BusinessAssert.isTrue(archivedPath.startsWith(archiveDir), CommonErrorCode.PARAM_ERROR);
+
+        try {
+            Files.move(currentPath, archivedPath, java.nio.file.StandardCopyOption.REPLACE_EXISTING);
+            return archivedPath;
+        } catch (IOException e) {
+            log.error("Failed to archive latest file, datasetId={}, fileId={}, from={}, to={}",
+                latest.getDatasetId(), latest.getId(), currentPath, archivedPath, e);
+            throw BusinessException.of(SystemErrorCode.FILE_SYSTEM_ERROR);
+        }
+    }
+
+    private void archiveOrphanTargetFile(Path datasetRoot, String logicalPath, Path targetFilePath) {
+        if (datasetRoot == null || targetFilePath == null) {
+            return;
+        }
+        if (!Files.exists(targetFilePath) || !Files.isRegularFile(targetFilePath)) {
+            return;
+        }
+        String logicalPathHash = sha256Hex(logicalPath);
+        Path orphanDir = datasetRoot
+            .resolve(INTERNAL_DIR_NAME)
+            .resolve(INTERNAL_VERSIONS_DIR_NAME)
+            .resolve(logicalPathHash)
+            .resolve("orphan")
+            .toAbsolutePath()
+            .normalize();
+        if (!orphanDir.startsWith(datasetRoot)) {
+            return;
+        }
+        try {
+            Files.createDirectories(orphanDir);
+            String safeName = sanitizeArchiveFileName(targetFilePath.getFileName().toString());
+            Path orphanPath = orphanDir.resolve("orphan_" + System.currentTimeMillis() + "__" + safeName)
+                .toAbsolutePath()
+                .normalize();
+            if (!orphanPath.startsWith(orphanDir)) {
+                return;
+            }
+            Files.move(targetFilePath, orphanPath, java.nio.file.StandardCopyOption.REPLACE_EXISTING);
+        } catch (Exception e) {
+            log.warn("Failed to archive orphan target file, logicalPath={}, targetPath={}", logicalPath, targetFilePath, e);
         }
-        dataset.active();
-        datasetRepository.updateById(dataset);
     }
 
     /**
@@ -570,11 +1044,16 @@ public class DatasetFileApplicationService {
         while (parentPrefix.startsWith("/")) {
             parentPrefix = parentPrefix.substring(1);
         }
+        parentPrefix = normalizeLogicalPrefix(parentPrefix);
+        assertNotInternalPrefix(parentPrefix);
 
         String directoryName = Optional.ofNullable(req.getDirectoryName()).orElse("").trim();
         if (directoryName.isEmpty()) {
             throw BusinessException.of(CommonErrorCode.PARAM_ERROR);
         }
+        if (INTERNAL_DIR_NAME.equals(directoryName)) {
+            throw BusinessException.of(CommonErrorCode.PARAM_ERROR);
+        }
         if (directoryName.contains("..") || directoryName.contains("/") || directoryName.contains("\\")) {
             throw BusinessException.of(CommonErrorCode.PARAM_ERROR);
         }
@@ -616,6 +1095,9 @@ public class DatasetFileApplicationService {
         while (prefix.endsWith("/")) {
             prefix = prefix.substring(0, prefix.length() - 1);
         }
+        if (prefix.equals(INTERNAL_DIR_NAME) || prefix.startsWith(INTERNAL_DIR_NAME + "/")) {
+            throw BusinessException.of(CommonErrorCode.PARAM_ERROR);
+        }
 
         Path basePath = Paths.get(datasetPath);
         Path targetPath = prefix.isEmpty() ? basePath : basePath.resolve(prefix);
@@ -652,6 +1134,7 @@ public class DatasetFileApplicationService {
     private void zipDirectory(Path sourceDir, Path basePath, ZipArchiveOutputStream zipOut) throws IOException {
         try (Stream<Path> paths = Files.walk(sourceDir)) {
             paths.filter(path -> !Files.isDirectory(path))
+                .filter(path -> !isInternalDatasetPath(basePath.toAbsolutePath().normalize(), path))
                 .forEach(path -> {
                     try {
                         Path relativePath = basePath.relativize(path);
@@ -690,6 +1173,9 @@ public class DatasetFileApplicationService {
         if (prefix.isEmpty()) {
             throw BusinessException.of(CommonErrorCode.PARAM_ERROR);
         }
+        if (prefix.equals(INTERNAL_DIR_NAME) || prefix.startsWith(INTERNAL_DIR_NAME + "/")) {
+            throw BusinessException.of(CommonErrorCode.PARAM_ERROR);
+        }
 
         String datasetPath = dataset.getPath();
         Path basePath = Paths.get(datasetPath);
@@ -761,28 +1247,6 @@ public class DatasetFileApplicationService {
         }
     }
 
-    /**
-     * 为数据集文件设置文件id
-     *
-     * @param datasetFile 要设置id的文件
-     * @param dataset 数据集（包含文件列表）
-     */
-    private void setDatasetFileId(DatasetFile datasetFile, Dataset dataset) {
-        Map<String, DatasetFile> existDatasetFilMap = dataset.getFiles().stream().collect(Collectors.toMap(DatasetFile::getFilePath, Function.identity()));
-        DatasetFile existDatasetFile = existDatasetFilMap.get(datasetFile.getFilePath());
-        if (Objects.isNull(existDatasetFile)) {
-            return;
-        }
-        if (duplicateMethod == DuplicateMethod.ERROR) {
-            log.error("file {} already exists in dataset {}", datasetFile.getFileName(), datasetFile.getDatasetId());
-            throw BusinessException.of(DataManagementErrorCode.DATASET_FILE_ALREADY_EXISTS);
-        }
-        if (duplicateMethod == DuplicateMethod.COVER) {
-            dataset.removeFile(existDatasetFile);
-            datasetFile.setId(existDatasetFile.getId());
-        }
-    }
-
     /**
      * 复制文件到数据集目录
      *
@@ -794,36 +1258,21 @@ public class DatasetFileApplicationService {
     public List<DatasetFile> copyFilesToDatasetDir(String datasetId, CopyFilesRequest req) {
         Dataset dataset = datasetRepository.getById(datasetId);
         BusinessAssert.notNull(dataset, SystemErrorCode.RESOURCE_NOT_FOUND);
+        Path datasetRoot = resolveDatasetRootPath(dataset, datasetId);
+        dataset.setFiles(new ArrayList<>(datasetFileRepository.findAllVisibleByDatasetId(datasetId)));
         List<DatasetFile> copiedFiles = new ArrayList<>();
-        List<DatasetFile> existDatasetFiles = datasetFileRepository.findAllByDatasetId(datasetId);
-        dataset.setFiles(existDatasetFiles);
         for (String sourceFilePath : req.sourcePaths()) {
-            Path sourcePath = Paths.get(sourceFilePath);
+            Path sourcePath = Paths.get(sourceFilePath).toAbsolutePath().normalize();
             if (!Files.exists(sourcePath) || !Files.isRegularFile(sourcePath)) {
                 log.warn("Source file does not exist or is not a regular file: {}", sourceFilePath);
                 continue;
             }
-            String fileName = sourcePath.getFileName().toString();
+            String logicalPath = sourcePath.getFileName().toString();
-            File sourceFile = sourcePath.toFile();
+            DatasetFile datasetFile = commitNewFileVersion(dataset, datasetRoot, logicalPath, sourcePath, false);
-            LocalDateTime currentTime = LocalDateTime.now();
-            DatasetFile datasetFile = DatasetFile.builder()
-                    .id(UUID.randomUUID().toString())
-                    .datasetId(datasetId)
-                    .fileName(fileName)
-                    .fileType(AnalyzerUtils.getExtension(fileName))
-                    .fileSize(sourceFile.length())
-                    .filePath(Paths.get(dataset.getPath(), fileName).toString())
-                    .uploadTime(currentTime)
-                    .lastAccessTime(currentTime)
-                    .build();
-            setDatasetFileId(datasetFile, dataset);
-            dataset.addFile(datasetFile);
             copiedFiles.add(datasetFile);
         }
-        datasetFileRepository.saveOrUpdateBatch(copiedFiles, 100);
         dataset.active();
         datasetRepository.updateById(dataset);
-        CompletableFuture.runAsync(() -> copyFilesToDatasetDir(req.sourcePaths(), dataset));
         return copiedFiles;
     }
 
@@ -839,13 +1288,11 @@ public class DatasetFileApplicationService {
     public List<DatasetFile> copyFilesToDatasetDirWithSourceRoot(String datasetId, Path sourceRoot, List<String> sourcePaths) {
         Dataset dataset = datasetRepository.getById(datasetId);
         BusinessAssert.notNull(dataset, SystemErrorCode.RESOURCE_NOT_FOUND);
-
+        Path datasetRoot = resolveDatasetRootPath(dataset, datasetId);
         Path normalizedRoot = sourceRoot.toAbsolutePath().normalize();
-        List<DatasetFile> copiedFiles = new ArrayList<>();
+        dataset.setFiles(new ArrayList<>(datasetFileRepository.findAllVisibleByDatasetId(datasetId)));
-        List<DatasetFile> existDatasetFiles = datasetFileRepository.findAllByDatasetId(datasetId);
-        dataset.setFiles(existDatasetFiles);
-        Map<String, DatasetFile> copyTargets = new LinkedHashMap<>();
 
+        List<DatasetFile> copiedFiles = new ArrayList<>();
         for (String sourceFilePath : sourcePaths) {
             if (sourceFilePath == null || sourceFilePath.isBlank()) {
                 continue;
@@ -859,86 +1306,16 @@ public class DatasetFileApplicationService {
                 log.warn("Source file does not exist or is not a regular file: {}", sourceFilePath);
                 continue;
             }
-
             Path relativePath = normalizedRoot.relativize(sourcePath);
-            String fileName = sourcePath.getFileName().toString();
+            String logicalPath = relativePath.toString().replace("\\", "/");
-            File sourceFile = sourcePath.toFile();
+            DatasetFile datasetFile = commitNewFileVersion(dataset, datasetRoot, logicalPath, sourcePath, false);
-            LocalDateTime currentTime = LocalDateTime.now();
-            Path targetPath = Paths.get(dataset.getPath(), relativePath.toString());
-
-            DatasetFile datasetFile = DatasetFile.builder()
-                .id(UUID.randomUUID().toString())
-                .datasetId(datasetId)
-                .fileName(fileName)
-                .fileType(AnalyzerUtils.getExtension(fileName))
-                .fileSize(sourceFile.length())
-                .filePath(targetPath.toString())
-                .uploadTime(currentTime)
-                .lastAccessTime(currentTime)
-                .build();
-            setDatasetFileId(datasetFile, dataset);
-            dataset.addFile(datasetFile);
             copiedFiles.add(datasetFile);
-            copyTargets.put(sourceFilePath, datasetFile);
         }
-
-        if (copiedFiles.isEmpty()) {
-            return copiedFiles;
-        }
-        datasetFileRepository.saveOrUpdateBatch(copiedFiles, 100);
         dataset.active();
         datasetRepository.updateById(dataset);
-        CompletableFuture.runAsync(() -> copyFilesToDatasetDirWithRelativePath(copyTargets, dataset, normalizedRoot));
         return copiedFiles;
     }
 
-    private void copyFilesToDatasetDir(List<String> sourcePaths, Dataset dataset) {
-        for (String sourcePath : sourcePaths) {
-            Path sourceFilePath = Paths.get(sourcePath);
-            Path targetFilePath = Paths.get(dataset.getPath(), sourceFilePath.getFileName().toString());
-            try {
-                Files.createDirectories(Path.of(dataset.getPath()));
-                Files.copy(sourceFilePath, targetFilePath);
-                DatasetFile datasetFile = datasetFileRepository.findByDatasetIdAndFileName(
-                    dataset.getId(),
-                    sourceFilePath.getFileName().toString()
-                );
-                triggerPdfTextExtraction(dataset, datasetFile);
-            } catch (IOException e) {
-                log.error("Failed to copy file from {} to {}", sourcePath, targetFilePath, e);
-            }
-        }
-    }
-
-    private void copyFilesToDatasetDirWithRelativePath(
-        Map<String, DatasetFile> copyTargets,
-        Dataset dataset,
-        Path sourceRoot
-    ) {
-        Path datasetRoot = Paths.get(dataset.getPath()).toAbsolutePath().normalize();
-        Path normalizedRoot = sourceRoot.toAbsolutePath().normalize();
-        for (Map.Entry<String, DatasetFile> entry : copyTargets.entrySet()) {
-            Path sourcePath = Paths.get(entry.getKey()).toAbsolutePath().normalize();
-            if (!sourcePath.startsWith(normalizedRoot)) {
-                log.warn("Source file path is out of root: {}", sourcePath);
-                continue;
-            }
-            Path relativePath = normalizedRoot.relativize(sourcePath);
-            Path targetFilePath = datasetRoot.resolve(relativePath).normalize();
-            if (!targetFilePath.startsWith(datasetRoot)) {
-                log.warn("Target file path is out of dataset path: {}", targetFilePath);
-                continue;
-            }
-            try {
-                Files.createDirectories(targetFilePath.getParent());
-                Files.copy(sourcePath, targetFilePath);
-                triggerPdfTextExtraction(dataset, entry.getValue());
-            } catch (IOException e) {
-                log.error("Failed to copy file from {} to {}", sourcePath, targetFilePath, e);
-            }
-        }
-    }
-
     /**
      * 添加文件到数据集（仅创建数据库记录，不执行文件系统操作）
      *
@@ -951,8 +1328,7 @@ public class DatasetFileApplicationService {
         Dataset dataset = datasetRepository.getById(datasetId);
         BusinessAssert.notNull(dataset, SystemErrorCode.RESOURCE_NOT_FOUND);
         List<DatasetFile> addedFiles = new ArrayList<>();
-        List<DatasetFile> existDatasetFiles = datasetFileRepository.findAllByDatasetId(datasetId);
+        dataset.setFiles(new ArrayList<>(datasetFileRepository.findAllVisibleByDatasetId(datasetId)));
-        dataset.setFiles(existDatasetFiles);
 
         boolean softAdd = req.softAdd();
         String metadata;
@@ -969,8 +1345,43 @@ public class DatasetFileApplicationService {
             Path sourcePath = Paths.get(sourceFilePath);
             String fileName = sourcePath.getFileName().toString();
             File sourceFile = sourcePath.toFile();
-            LocalDateTime currentTime = LocalDateTime.now();
+            String logicalPath = normalizeLogicalPath(fileName);
+            assertNotInternalPrefix(logicalPath);
 
+            DatasetFile latest = datasetFileRepository.findLatestByDatasetIdAndLogicalPath(datasetId, logicalPath);
+            if (latest == null && dataset.getFiles() != null) {
+                latest = dataset.getFiles().stream()
+                    .filter(existing -> existing != null
+                        && !isArchivedStatus(existing)
+                        && Objects.equals(existing.getFileName(), fileName))
+                    .findFirst()
+                    .orElse(null);
+            }
+
+            DuplicateMethod effectiveDuplicateMethod = resolveEffectiveDuplicateMethod();
+            if (latest != null && effectiveDuplicateMethod == DuplicateMethod.ERROR) {
+                throw BusinessException.of(DataManagementErrorCode.DATASET_FILE_ALREADY_EXISTS);
+            }
+
+            long nextVersion = 1L;
+            if (latest != null) {
+                long latestVersion = Optional.ofNullable(latest.getVersion()).orElse(1L);
+                if (latest.getVersion() == null) {
+                    latest.setVersion(latestVersion);
+                }
+                if (latest.getLogicalPath() == null || latest.getLogicalPath().isBlank()) {
+                    latest.setLogicalPath(logicalPath);
+                }
+                nextVersion = latestVersion + 1L;
+            }
+
+            if (latest != null && effectiveDuplicateMethod == DuplicateMethod.VERSION) {
+                latest.setStatus(FILE_STATUS_ARCHIVED);
+                datasetFileRepository.updateById(latest);
+                dataset.removeFile(latest);
+            }
+
+            LocalDateTime currentTime = LocalDateTime.now();
             DatasetFile datasetFile = DatasetFile.builder()
                 .id(UUID.randomUUID().toString())
                 .datasetId(datasetId)
@@ -978,16 +1389,19 @@ public class DatasetFileApplicationService {
                 .fileType(AnalyzerUtils.getExtension(fileName))
                 .fileSize(sourceFile.length())
                 .filePath(sourceFilePath)
+                .logicalPath(logicalPath)
+                .version(nextVersion)
+                .status(FILE_STATUS_ACTIVE)
                 .uploadTime(currentTime)
                 .lastAccessTime(currentTime)
                 .metadata(metadata)
                 .build();
-            setDatasetFileId(datasetFile, dataset);
+
+            datasetFileRepository.saveOrUpdate(datasetFile);
             dataset.addFile(datasetFile);
             addedFiles.add(datasetFile);
             triggerPdfTextExtraction(dataset, datasetFile);
         }
-        datasetFileRepository.saveOrUpdateBatch(addedFiles, 100);
         dataset.active();
         datasetRepository.updateById(dataset);
         // Note: addFilesToDataset only creates DB records, no file system operations

backend/services/data-management-service/src/main/java/com/datamate/datamanagement/common/enums/DuplicateMethod.java

@@ -7,5 +7,6 @@ package com.datamate.datamanagement.common.enums;
  */
 public enum DuplicateMethod {
     ERROR,
-    COVER
+    COVER,
+    VERSION
 }

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

@@ -152,12 +152,20 @@ public class Dataset extends BaseEntity<String> {
     }
 
     public void removeFile(DatasetFile file) {
-        if (this.files.remove(file)) {
+        if (file == null) {
+            return;
+        }
+        boolean removed = this.files.remove(file);
+        if (!removed && file.getId() != null) {
+            removed = this.files.removeIf(existing -> Objects.equals(existing.getId(), file.getId()));
+        }
+        if (!removed) {
+            return;
+        }
         this.fileCount = Math.max(0, this.fileCount - 1);
         this.sizeBytes = Math.max(0, this.sizeBytes - (file.getFileSize() != null ? file.getFileSize() : 0L));
         this.updatedAt = LocalDateTime.now();
     }
-    }
 
     public void active() {
         if (this.status == DatasetStatusType.DRAFT) {

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

@@ -28,12 +28,16 @@ public class DatasetFile {
     private String datasetId; // UUID
     private String fileName;
     private String filePath;
+    /** 文件逻辑路径（相对数据集根目录，包含子目录） */
+    private String logicalPath;
+    /** 文件版本号（同一个 logicalPath 下递增） */
+    private Long version;
     private String fileType; // JPG/PNG/DCM/TXT
     private Long fileSize; // bytes
     private String checkSum;
     private String tags;
     private String metadata;
-    private String status; // UPLOADED, PROCESSING, COMPLETED, ERROR
+    private String status; // ACTIVE/ARCHIVED/DELETED/PROCESSING...
     private LocalDateTime uploadTime;
     private LocalDateTime lastAccessTime;
     private LocalDateTime createdAt;

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

@@ -21,4 +21,7 @@ public class DatasetFileUploadCheckInfo {
 
     /** 目标子目录前缀，例如 "images/"，为空表示数据集根目录 */
     private String prefix;
+
+    /** 上传临时落盘目录（仅服务端使用，不对外暴露） */
+    private String stagingPath;
 }

backend/services/data-management-service/src/main/java/com/datamate/datamanagement/infrastructure/persistence/repository/DatasetFileRepository.java

@@ -24,8 +24,19 @@ public interface DatasetFileRepository extends IRepository<DatasetFile> {
 
     List<DatasetFile> findAllByDatasetId(String datasetId);
 
+    /**
+     * 查询数据集内“可见文件”（默认不包含历史归档版本）。
+     * 约定：status 为 NULL 视为可见；status = ARCHIVED 视为历史版本。
+     */
+    List<DatasetFile> findAllVisibleByDatasetId(String datasetId);
+
     DatasetFile findByDatasetIdAndFileName(String datasetId, String fileName);
 
+    /**
+     * 查询指定逻辑路径的最新版本（ACTIVE/NULL）。
+     */
+    DatasetFile findLatestByDatasetIdAndLogicalPath(String datasetId, String logicalPath);
+
     IPage<DatasetFile> findByCriteria(String datasetId, String fileType, String status, String name,
                                       Boolean hasAnnotation, IPage<DatasetFile> page);
 

backend/services/data-management-service/src/main/java/com/datamate/datamanagement/infrastructure/persistence/repository/impl/DatasetFileRepositoryImpl.java

@@ -25,6 +25,8 @@ public class DatasetFileRepositoryImpl extends CrudRepository<DatasetFileMapper,
     private final DatasetFileMapper datasetFileMapper;
     private static final String ANNOTATION_EXISTS_SQL =
             "SELECT 1 FROM t_dm_annotation_results ar WHERE ar.file_id = t_dm_dataset_files.id";
+    private static final String FILE_STATUS_ARCHIVED = "ARCHIVED";
+    private static final String FILE_STATUS_ACTIVE = "ACTIVE";
 
     @Override
     public Long countByDatasetId(String datasetId) {
@@ -51,19 +53,54 @@ public class DatasetFileRepositoryImpl extends CrudRepository<DatasetFileMapper,
         return datasetFileMapper.findAllByDatasetId(datasetId);
     }
 
+    @Override
+    public List<DatasetFile> findAllVisibleByDatasetId(String datasetId) {
+        return datasetFileMapper.selectList(new LambdaQueryWrapper<DatasetFile>()
+            .eq(DatasetFile::getDatasetId, datasetId)
+            .and(wrapper -> wrapper.isNull(DatasetFile::getStatus)
+                .or()
+                .ne(DatasetFile::getStatus, FILE_STATUS_ARCHIVED))
+            .orderByDesc(DatasetFile::getUploadTime));
+    }
+
     @Override
     public DatasetFile findByDatasetIdAndFileName(String datasetId, String fileName) {
         return datasetFileMapper.findByDatasetIdAndFileName(datasetId, fileName);
     }
 
+    @Override
+    public DatasetFile findLatestByDatasetIdAndLogicalPath(String datasetId, String logicalPath) {
+        if (!StringUtils.hasText(datasetId) || !StringUtils.hasText(logicalPath)) {
+            return null;
+        }
+        return datasetFileMapper.selectOne(new LambdaQueryWrapper<DatasetFile>()
+            .eq(DatasetFile::getDatasetId, datasetId)
+            .eq(DatasetFile::getLogicalPath, logicalPath)
+            .and(wrapper -> wrapper.isNull(DatasetFile::getStatus)
+                .or()
+                .eq(DatasetFile::getStatus, FILE_STATUS_ACTIVE))
+            .orderByDesc(DatasetFile::getVersion)
+            .orderByDesc(DatasetFile::getUploadTime)
+            .last("LIMIT 1"));
+    }
+
     public IPage<DatasetFile> findByCriteria(String datasetId, String fileType, String status, String name,
                                              Boolean hasAnnotation, IPage<DatasetFile> page) {
-        return datasetFileMapper.selectPage(page, new LambdaQueryWrapper<DatasetFile>()
+        LambdaQueryWrapper<DatasetFile> wrapper = new LambdaQueryWrapper<DatasetFile>()
             .eq(DatasetFile::getDatasetId, datasetId)
             .eq(StringUtils.hasText(fileType), DatasetFile::getFileType, fileType)
-                .eq(StringUtils.hasText(status), DatasetFile::getStatus, status)
             .like(StringUtils.hasText(name), DatasetFile::getFileName, name)
-                .exists(Boolean.TRUE.equals(hasAnnotation), ANNOTATION_EXISTS_SQL));
+            .exists(Boolean.TRUE.equals(hasAnnotation), ANNOTATION_EXISTS_SQL);
+
+        if (StringUtils.hasText(status)) {
+            wrapper.eq(DatasetFile::getStatus, status);
+        } else {
+            wrapper.and(visibility -> visibility.isNull(DatasetFile::getStatus)
+                .or()
+                .ne(DatasetFile::getStatus, FILE_STATUS_ARCHIVED));
+        }
+
+        return datasetFileMapper.selectPage(page, wrapper);
     }
 
     @Override

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/services/data-management-service/src/main/resources/mappers/DatasetFileMapper.xml

@@ -3,7 +3,7 @@
     "http://mybatis.org/dtd/mybatis-3-mapper.dtd">
 <mapper namespace="com.datamate.datamanagement.infrastructure.persistence.mapper.DatasetFileMapper">
     <sql id="Base_Column_List">
-      id, dataset_id, file_name, file_path, file_type, file_size, check_sum, tags, metadata, status,
+      id, dataset_id, file_name, file_path, logical_path, version, file_type, file_size, check_sum, tags, metadata, status,
           upload_time, last_access_time, created_at, updated_at
     </sql>
 
@@ -39,13 +39,17 @@
     </select>
 
     <select id="countByDatasetId" parameterType="string" resultType="long">
-        SELECT COUNT(*) FROM t_dm_dataset_files WHERE dataset_id = #{datasetId}
+        SELECT COUNT(*)
+        FROM t_dm_dataset_files
+        WHERE dataset_id = #{datasetId}
+          AND (status IS NULL OR status <> 'ARCHIVED')
     </select>
 
     <select id="countNonDerivedByDatasetId" parameterType="string" resultType="long">
         SELECT COUNT(*)
         FROM t_dm_dataset_files
         WHERE dataset_id = #{datasetId}
+          AND (status IS NULL OR status <> 'ARCHIVED')
           AND (metadata IS NULL OR JSON_EXTRACT(metadata, '$.derived_from_file_id') IS NULL)
     </select>
 
@@ -54,13 +58,19 @@
     </select>
 
     <select id="sumSizeByDatasetId" parameterType="string" resultType="long">
-        SELECT COALESCE(SUM(file_size), 0) FROM t_dm_dataset_files WHERE dataset_id = #{datasetId}
+        SELECT COALESCE(SUM(file_size), 0)
+        FROM t_dm_dataset_files
+        WHERE dataset_id = #{datasetId}
+          AND (status IS NULL OR status <> 'ARCHIVED')
     </select>
 
   <select id="findByDatasetIdAndFileName" resultType="com.datamate.datamanagement.domain.model.dataset.DatasetFile">
         SELECT <include refid="Base_Column_List"/>
         FROM t_dm_dataset_files
-        WHERE dataset_id = #{datasetId} AND file_name = #{fileName}
+        WHERE dataset_id = #{datasetId}
+          AND file_name = #{fileName}
+          AND (status IS NULL OR status <> 'ARCHIVED')
+        ORDER BY version DESC, upload_time DESC
         LIMIT 1
     </select>
 
@@ -91,6 +101,8 @@
         UPDATE t_dm_dataset_files
         SET file_name = #{fileName},
             file_path = #{filePath},
+            logical_path = #{logicalPath},
+            version = #{version},
             file_type = #{fileType},
             file_size = #{fileSize},
             upload_time = #{uploadTime},
@@ -126,6 +138,7 @@
         <foreach collection="datasetIds" item="datasetId" open="(" separator="," close=")">
             #{datasetId}
         </foreach>
+          AND (status IS NULL OR status <> 'ARCHIVED')
           AND (metadata IS NULL OR JSON_EXTRACT(metadata, '$.derived_from_file_id') IS NULL)
         GROUP BY dataset_id
     </select>

backend/services/data-management-service/src/test/java/com/datamate/datamanagement/application/DatasetFileApplicationServiceVersioningTest.java

@@ -0,0 +1,147 @@
+package com.datamate.datamanagement.application;
+
+import com.datamate.common.domain.service.FileService;
+import com.datamate.datamanagement.domain.model.dataset.Dataset;
+import com.datamate.datamanagement.domain.model.dataset.DatasetFile;
+import com.datamate.datamanagement.infrastructure.persistence.repository.DatasetFileRepository;
+import com.datamate.datamanagement.infrastructure.persistence.repository.DatasetRepository;
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.extension.ExtendWith;
+import org.junit.jupiter.api.io.TempDir;
+import org.mockito.ArgumentCaptor;
+import org.mockito.Mock;
+import org.mockito.junit.jupiter.MockitoExtension;
+
+import java.nio.charset.StandardCharsets;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.nio.file.Paths;
+import java.security.MessageDigest;
+import java.util.List;
+
+import static org.assertj.core.api.Assertions.assertThat;
+import static org.mockito.ArgumentMatchers.anyString;
+import static org.mockito.Mockito.verify;
+import static org.mockito.Mockito.when;
+
+@ExtendWith(MockitoExtension.class)
+class DatasetFileApplicationServiceVersioningTest {
+
+    @TempDir
+    Path tempDir;
+
+    @Mock
+    DatasetFileRepository datasetFileRepository;
+
+    @Mock
+    DatasetRepository datasetRepository;
+
+    @Mock
+    FileService fileService;
+
+    @Mock
+    PdfTextExtractAsyncService pdfTextExtractAsyncService;
+
+    @Mock
+    DatasetFilePreviewService datasetFilePreviewService;
+
+    @Test
+    void copyFilesToDatasetDirWithSourceRoot_shouldArchiveOldFileAndCreateNewVersionWhenDuplicateLogicalPath()
+        throws Exception {
+        String datasetId = "dataset-1";
+
+        Path datasetRoot = tempDir.resolve("dataset-root");
+        Files.createDirectories(datasetRoot);
+
+        Path sourceRoot = tempDir.resolve("source-root");
+        Files.createDirectories(sourceRoot);
+
+        Path existingPath = datasetRoot.resolve("a.txt");
+        Files.writeString(existingPath, "old-content", StandardCharsets.UTF_8);
+
+        Path incomingPath = sourceRoot.resolve("a.txt");
+        Files.writeString(incomingPath, "new-content", StandardCharsets.UTF_8);
+
+        Dataset dataset = new Dataset();
+        dataset.setId(datasetId);
+        dataset.setPath(datasetRoot.toString());
+
+        DatasetFile oldRecord = DatasetFile.builder()
+            .id("old-file-id")
+            .datasetId(datasetId)
+            .fileName("a.txt")
+            .filePath(existingPath.toString())
+            .logicalPath(null)
+            .version(null)
+            .status(null)
+            .fileSize(Files.size(existingPath))
+            .build();
+
+        when(datasetRepository.getById(datasetId)).thenReturn(dataset);
+        when(datasetFileRepository.findAllVisibleByDatasetId(datasetId)).thenReturn(List.of(oldRecord));
+        when(datasetFileRepository.findLatestByDatasetIdAndLogicalPath(anyString(), anyString())).thenReturn(null);
+
+        DatasetFileApplicationService service = new DatasetFileApplicationService(
+            datasetFileRepository,
+            datasetRepository,
+            fileService,
+            pdfTextExtractAsyncService,
+            datasetFilePreviewService
+        );
+
+        List<DatasetFile> copied = service.copyFilesToDatasetDirWithSourceRoot(
+            datasetId,
+            sourceRoot,
+            List.of(incomingPath.toString())
+        );
+
+        assertThat(copied).hasSize(1);
+        assertThat(Files.readString(existingPath, StandardCharsets.UTF_8)).isEqualTo("new-content");
+
+        String logicalPathHash = sha256Hex("a.txt");
+        Path archivedPath = datasetRoot
+            .resolve(".datamate")
+            .resolve("versions")
+            .resolve(logicalPathHash)
+            .resolve("v1")
+            .resolve("old-file-id__a.txt")
+            .toAbsolutePath()
+            .normalize();
+
+        assertThat(Files.exists(archivedPath)).isTrue();
+        assertThat(Files.readString(archivedPath, StandardCharsets.UTF_8)).isEqualTo("old-content");
+
+        ArgumentCaptor<DatasetFile> archivedCaptor = ArgumentCaptor.forClass(DatasetFile.class);
+        verify(datasetFileRepository).updateById(archivedCaptor.capture());
+        DatasetFile archivedRecord = archivedCaptor.getValue();
+        assertThat(archivedRecord.getId()).isEqualTo("old-file-id");
+        assertThat(archivedRecord.getStatus()).isEqualTo("ARCHIVED");
+        assertThat(archivedRecord.getLogicalPath()).isEqualTo("a.txt");
+        assertThat(archivedRecord.getVersion()).isEqualTo(1L);
+        assertThat(Paths.get(archivedRecord.getFilePath()).toAbsolutePath().normalize()).isEqualTo(archivedPath);
+
+        ArgumentCaptor<DatasetFile> createdCaptor = ArgumentCaptor.forClass(DatasetFile.class);
+        verify(datasetFileRepository).saveOrUpdate(createdCaptor.capture());
+        DatasetFile newRecord = createdCaptor.getValue();
+        assertThat(newRecord.getId()).isNotEqualTo("old-file-id");
+        assertThat(newRecord.getStatus()).isEqualTo("ACTIVE");
+        assertThat(newRecord.getLogicalPath()).isEqualTo("a.txt");
+        assertThat(newRecord.getVersion()).isEqualTo(2L);
+        assertThat(Paths.get(newRecord.getFilePath()).toAbsolutePath().normalize()).isEqualTo(existingPath.toAbsolutePath().normalize());
+    }
+
+    private static String sha256Hex(String value) {
+        try {
+            MessageDigest digest = MessageDigest.getInstance("SHA-256");
+            byte[] hashed = digest.digest((value == null ? "" : value).getBytes(StandardCharsets.UTF_8));
+            StringBuilder builder = new StringBuilder(hashed.length * 2);
+            for (byte b : hashed) {
+                builder.append(String.format("%02x", b));
+            }
+            return builder.toString();
+        } catch (Exception e) {
+            return Integer.toHexString((value == null ? "" : value).hashCode());
+        }
+    }
+}
+

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));

backend/shared/domain-common/src/main/java/com/datamate/common/domain/utils/ArchiveAnalyzer.java

@@ -143,7 +143,20 @@ public class ArchiveAnalyzer {
     private static Optional<FileUploadResult> extractEntity(ArchiveInputStream<?> archiveInputStream, ArchiveEntry archiveEntry, Path archivePath)
         throws IOException {
         byte[] buffer = new byte[DEFAULT_BUFFER_SIZE];
-        Path path = Paths.get(archivePath.getParent().toString(), archiveEntry.getName());
+        Path archiveRoot = archivePath.getParent().toAbsolutePath().normalize();
+        String entryName = archiveEntry.getName();
+        if (entryName == null || entryName.isBlank()) {
+            return Optional.empty();
+        }
+        entryName = entryName.replace("\\", "/");
+        while (entryName.startsWith("/")) {
+            entryName = entryName.substring(1);
+        }
+        Path path = archiveRoot.resolve(entryName).normalize();
+        if (!path.startsWith(archiveRoot)) {
+            log.warn("Skip unsafe archive entry path traversal: {}", archiveEntry.getName());
+            return Optional.empty();
+        }
         File file = path.toFile();
         long fileSize = 0L;
         FileUtils.createParentDirectories(file);

backend/shared/domain-common/src/main/java/com/datamate/common/domain/utils/CommonUtils.java

@@ -13,7 +13,10 @@ public class CommonUtils {
      * @return 文件名（带后缀）
      */
     public static String trimFilePath(String filePath) {
-        int lastSlashIndex = filePath.lastIndexOf(File.separator);
+        if (filePath == null || filePath.isBlank()) {
+            return "";
+        }
+        int lastSlashIndex = Math.max(filePath.lastIndexOf('/'), filePath.lastIndexOf('\\'));
 
         String filename = filePath;
         if (lastSlashIndex != -1) {

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();
@@ -37,6 +38,14 @@ export function useFileSliceUpload(
     taskListRef.current = [task, ...taskListRef.current];
 
     setTaskList(taskListRef.current);
+    
+    // 立即显示任务中心，让用户感知上传已开始
+    if (showTaskCenter) {
+      window.dispatchEvent(
+        new CustomEvent("show:task-popover", { detail: { show: true } })
+      );
+    }
+    
     return task;
   };
 
@@ -60,7 +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 },
         })
       );
     }
@@ -71,7 +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]);
@@ -86,12 +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,
@@ -101,6 +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);
@@ -116,7 +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,
@@ -132,24 +148,29 @@ 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);
-    if (showTaskCenter) {
+    // 注意：show:task-popover 事件已在 createTask 中触发，此处不再重复触发
-      window.dispatchEvent(
-        new CustomEvent("show:task-popover", { detail: { show: true } })
-      );
-    }
     // // 更新数据状态
     if (task.updateEvent) window.dispatchEvent(new Event(task.updateEvent));
 
     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,
@@ -163,7 +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");
@@ -189,10 +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

@@ -1,6 +1,6 @@
 import { useCallback, useEffect, useMemo, useRef, useState } from "react";
-import { App, Button, Card, List, Spin, Typography, Tag, Switch, Tree, Empty } from "antd";
+import { App, Button, Card, List, Spin, Typography, Tag, Empty } from "antd";
-import { LeftOutlined, ReloadOutlined, SaveOutlined, MenuFoldOutlined, MenuUnfoldOutlined, CheckOutlined } from "@ant-design/icons";
+import { LeftOutlined, ReloadOutlined, SaveOutlined, MenuFoldOutlined, MenuUnfoldOutlined } from "@ant-design/icons";
 import { useNavigate, useParams } from "react-router";
 
 import {
@@ -28,7 +28,6 @@ type EditorTaskListItem = {
   hasAnnotation: boolean;
   annotationUpdatedAt?: string | null;
   annotationStatus?: AnnotationResultStatus | null;
-  segmentStats?: SegmentStats;
 };
 
 type LsfMessage = {
@@ -36,21 +35,6 @@ type LsfMessage = {
   payload?: unknown;
 };
 
-type SegmentInfo = {
-  idx: number;
-  text: string;
-  start: number;
-  end: number;
-  hasAnnotation: boolean;
-  lineIndex: number;
-  chunkIndex: number;
-};
-
-type SegmentStats = {
-  done: number;
-  total: number;
-};
-
 type ApiResponse<T> = {
   code?: number;
   message?: string;
@@ -66,10 +50,11 @@ type EditorTaskPayload = {
 type EditorTaskResponse = {
   task?: EditorTaskPayload;
   segmented?: boolean;
-  segments?: SegmentInfo[];
+  totalSegments?: number;
   currentSegmentIndex?: number;
 };
 
+
 type EditorTaskListResponse = {
   content?: EditorTaskListItem[];
   totalElements?: number;
@@ -91,8 +76,6 @@ type ExportPayload = {
   requestId?: string | null;
 };
 
-type SwitchDecision = "save" | "discard" | "cancel";
-
 const LSF_IFRAME_SRC = "/lsf/lsf.html";
 const TASK_PAGE_START = 0;
 const TASK_PAGE_SIZE = 200;
@@ -154,16 +137,6 @@ const isAnnotationResultEmpty = (annotation?: Record<string, unknown>) => {
 };
 
 const resolveTaskStatusMeta = (item: EditorTaskListItem) => {
-  const segmentSummary = resolveSegmentSummary(item);
-  if (segmentSummary) {
-    if (segmentSummary.done >= segmentSummary.total) {
-      return { text: "已标注", type: "success" as const };
-    }
-    if (segmentSummary.done > 0) {
-      return { text: "标注中", type: "warning" as const };
-    }
-    return { text: "未标注", type: "secondary" as const };
-  }
   if (!item.hasAnnotation) {
     return { text: "未标注", type: "secondary" as const };
   }
@@ -216,25 +189,6 @@ const buildAnnotationSnapshot = (annotation?: Record<string, unknown>) => {
 const buildSnapshotKey = (fileId: string, segmentIndex?: number) =>
   `${fileId}::${segmentIndex ?? "full"}`;
 
-const buildSegmentStats = (segmentList?: SegmentInfo[] | null): SegmentStats | null => {
-  if (!Array.isArray(segmentList) || segmentList.length === 0) return null;
-  const total = segmentList.length;
-  const done = segmentList.reduce((count, seg) => count + (seg.hasAnnotation ? 1 : 0), 0);
-  return { done, total };
-};
-
-const normalizeSegmentStats = (stats?: SegmentStats | null): SegmentStats | null => {
-  if (!stats) return null;
-  const total = Number(stats.total);
-  const done = Number(stats.done);
-  if (!Number.isFinite(total) || total <= 0) return null;
-  const safeDone = Math.min(Math.max(done, 0), total);
-  return { done: safeDone, total };
-};
-
-const resolveSegmentSummary = (item: EditorTaskListItem) =>
-  normalizeSegmentStats(item.segmentStats);
-
 const mergeTaskItems = (base: EditorTaskListItem[], next: EditorTaskListItem[]) => {
   if (next.length === 0) return base;
   const seen = new Set(base.map((item) => item.fileId));
@@ -282,18 +236,13 @@ export default function LabelStudioTextEditor() {
     resolve: (payload?: ExportPayload) => void;
     timer?: number;
   } | null>(null);
-  const exportCheckSeqRef = useRef(0);
   const savedSnapshotsRef = useRef<Record<string, string>>({});
   const pendingAutoAdvanceRef = useRef(false);
-  const segmentStatsCacheRef = useRef<Record<string, SegmentStats>>({});
-  const segmentStatsSeqRef = useRef(0);
-  const segmentStatsLoadingRef = useRef<Set<string>>(new Set());
 
   const [loadingProject, setLoadingProject] = useState(true);
   const [loadingTasks, setLoadingTasks] = useState(false);
   const [loadingTaskDetail, setLoadingTaskDetail] = useState(false);
   const [saving, setSaving] = useState(false);
-  const [segmentSwitching, setSegmentSwitching] = useState(false);
 
   const [iframeReady, setIframeReady] = useState(false);
   const [lsReady, setLsReady] = useState(false);
@@ -306,16 +255,19 @@ export default function LabelStudioTextEditor() {
   const [prefetching, setPrefetching] = useState(false);
   const [selectedFileId, setSelectedFileId] = useState<string>("");
   const [sidebarCollapsed, setSidebarCollapsed] = useState(false);
-  const [autoSaveOnSwitch, setAutoSaveOnSwitch] = useState(false);
 
   // 分段相关状态
   const [segmented, setSegmented] = useState(false);
-  const [segments, setSegments] = useState<SegmentInfo[]>([]);
   const [currentSegmentIndex, setCurrentSegmentIndex] = useState(0);
+  const [segmentTotal, setSegmentTotal] = useState(0);
   const isTextProject = useMemo(
     () => (project?.datasetType || "").toUpperCase() === "TEXT",
     [project?.datasetType],
   );
+  const segmentIndices = useMemo(() => {
+    if (segmentTotal <= 0) return [] as number[];
+    return Array.from({ length: segmentTotal }, (_, index) => index);
+  }, [segmentTotal]);
 
   const focusIframe = useCallback(() => {
     const iframe = iframeRef.current;
@@ -330,70 +282,6 @@ export default function LabelStudioTextEditor() {
     win.postMessage({ type, payload }, origin);
   }, [origin]);
 
-  const applySegmentStats = useCallback((fileId: string, stats: SegmentStats | null) => {
-    if (!fileId) return;
-    const normalized = normalizeSegmentStats(stats);
-    setTasks((prev) =>
-      prev.map((item) =>
-        item.fileId === fileId
-          ? { ...item, segmentStats: normalized || undefined }
-          : item
-      )
-    );
-  }, []);
-
-  const updateSegmentStatsCache = useCallback((fileId: string, stats: SegmentStats | null) => {
-    if (!fileId) return;
-    const normalized = normalizeSegmentStats(stats);
-    if (normalized) {
-      segmentStatsCacheRef.current[fileId] = normalized;
-    } else {
-      delete segmentStatsCacheRef.current[fileId];
-    }
-    applySegmentStats(fileId, normalized);
-  }, [applySegmentStats]);
-
-  const fetchSegmentStatsForFile = useCallback(async (fileId: string, seq: number) => {
-    if (!projectId || !fileId) return;
-    if (segmentStatsCacheRef.current[fileId] || segmentStatsLoadingRef.current.has(fileId)) return;
-    segmentStatsLoadingRef.current.add(fileId);
-    try {
-      const resp = (await getEditorTaskUsingGet(projectId, fileId, {
-        segmentIndex: 0,
-      })) as ApiResponse<EditorTaskResponse>;
-      if (segmentStatsSeqRef.current !== seq) return;
-      const data = resp?.data;
-      if (!data?.segmented) return;
-      const stats = buildSegmentStats(data.segments);
-      if (!stats) return;
-      segmentStatsCacheRef.current[fileId] = stats;
-      applySegmentStats(fileId, stats);
-    } catch (e) {
-      console.error(e);
-    } finally {
-      segmentStatsLoadingRef.current.delete(fileId);
-    }
-  }, [applySegmentStats, projectId]);
-
-  const prefetchSegmentStats = useCallback((items: EditorTaskListItem[]) => {
-    if (!projectId) return;
-    const fileIds = items
-      .map((item) => item.fileId)
-      .filter((fileId) => fileId && !segmentStatsCacheRef.current[fileId]);
-    if (fileIds.length === 0) return;
-    const seq = segmentStatsSeqRef.current;
-    let cursor = 0;
-    const workerCount = Math.min(3, fileIds.length);
-    const runWorker = async () => {
-      while (cursor < fileIds.length && segmentStatsSeqRef.current === seq) {
-        const fileId = fileIds[cursor];
-        cursor += 1;
-        await fetchSegmentStatsForFile(fileId, seq);
-      }
-    };
-    void Promise.all(Array.from({ length: workerCount }, () => runWorker()));
-  }, [fetchSegmentStatsForFile, projectId]);
-
   const confirmEmptyAnnotationStatus = useCallback(() => {
     return new Promise<AnnotationResultStatus | null>((resolve) => {
       let resolved = false;
@@ -446,8 +334,6 @@ export default function LabelStudioTextEditor() {
 
   const updateTaskSelection = useCallback((items: EditorTaskListItem[]) => {
     const isCompleted = (item: EditorTaskListItem) => {
-      const summary = resolveSegmentSummary(item);
-      if (summary) return summary.done >= summary.total;
       return item.hasAnnotation;
     };
     const defaultFileId =
@@ -508,9 +394,6 @@ export default function LabelStudioTextEditor() {
     if (mode === "reset") {
       prefetchSeqRef.current += 1;
       setPrefetching(false);
-      segmentStatsSeqRef.current += 1;
-      segmentStatsCacheRef.current = {};
-      segmentStatsLoadingRef.current = new Set();
     }
     if (mode === "append") {
       setLoadingMore(true);
@@ -591,20 +474,19 @@ 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);
         setSegmented(true);
-        setSegments(data.segments || []);
         setCurrentSegmentIndex(segmentIndex ?? 0);
-        updateSegmentStatsCache(fileId, stats);
+        const totalSegments = Number(data?.totalSegments ?? 0);
+        setSegmentTotal(Number.isFinite(totalSegments) && totalSegments > 0 ? totalSegments : 0);
       } else {
         setSegmented(false);
-        setSegments([]);
         setCurrentSegmentIndex(0);
-        updateSegmentStatsCache(fileId, null);
+        setSegmentTotal(0);
       }
 
       const taskData = {
@@ -664,19 +546,14 @@ export default function LabelStudioTextEditor() {
     } finally {
       if (seq === initSeqRef.current) setLoadingTaskDetail(false);
     }
-  }, [iframeReady, message, postToIframe, project, projectId, updateSegmentStatsCache]);
+  }, [iframeReady, message, postToIframe, project, projectId]);
 
   const advanceAfterSave = useCallback(async (fileId: string, segmentIndex?: number) => {
     if (!fileId) return;
-    if (segmented && segments.length > 0) {
+    if (segmented && segmentTotal > 0) {
-      const sortedSegmentIndices = segments
+      const baseIndex = Math.max(segmentIndex ?? currentSegmentIndex, 0);
-        .map((seg) => seg.idx)
+      const nextSegmentIndex = baseIndex + 1;
-        .sort((a, b) => a - b);
+      if (nextSegmentIndex < segmentTotal) {
-      const baseIndex = segmentIndex ?? currentSegmentIndex;
-      const currentPos = sortedSegmentIndices.indexOf(baseIndex);
-      const nextSegmentIndex =
-        currentPos >= 0 ? sortedSegmentIndices[currentPos + 1] : sortedSegmentIndices[0];
-      if (nextSegmentIndex !== undefined) {
         await initEditorForFile(fileId, nextSegmentIndex);
         return;
       }
@@ -698,7 +575,7 @@ export default function LabelStudioTextEditor() {
     initEditorForFile,
     message,
     segmented,
-    segments,
+    segmentTotal,
     tasks,
   ]);
 
@@ -772,16 +649,6 @@ export default function LabelStudioTextEditor() {
       const snapshot = buildAnnotationSnapshot(isRecord(annotation) ? annotation : undefined);
       savedSnapshotsRef.current[snapshotKey] = snapshot;
 
-      // 分段模式下更新当前段落的标注状态
-      if (segmented && segmentIndex !== undefined) {
-        const nextSegments = segments.map((seg) =>
-          seg.idx === segmentIndex
-            ? { ...seg, hasAnnotation: true }
-            : seg
-        );
-        setSegments(nextSegments);
-        updateSegmentStatsCache(String(fileId), buildSegmentStats(nextSegments));
-      }
       if (options?.autoAdvance) {
         await advanceAfterSave(String(fileId), segmentIndex);
       }
@@ -800,69 +667,10 @@ export default function LabelStudioTextEditor() {
     message,
     projectId,
     segmented,
-    segments,
     selectedFileId,
     tasks,
-    updateSegmentStatsCache,
   ]);
 
-  const requestExportForCheck = useCallback(() => {
-    if (!iframeReady || !lsReady) return Promise.resolve(undefined);
-    if (exportCheckRef.current) {
-      if (exportCheckRef.current.timer) {
-        window.clearTimeout(exportCheckRef.current.timer);
-      }
-      exportCheckRef.current.resolve(undefined);
-      exportCheckRef.current = null;
-    }
-    const requestId = `check_${Date.now()}_${++exportCheckSeqRef.current}`;
-    return new Promise<ExportPayload | undefined>((resolve) => {
-      const timer = window.setTimeout(() => {
-        if (exportCheckRef.current?.requestId === requestId) {
-          exportCheckRef.current = null;
-        }
-        resolve(undefined);
-      }, 3000);
-      exportCheckRef.current = {
-        requestId,
-        resolve,
-        timer,
-      };
-      postToIframe("LS_EXPORT_CHECK", { requestId });
-    });
-  }, [iframeReady, lsReady, postToIframe]);
-
-  const confirmSaveBeforeSwitch = useCallback(() => {
-    return new Promise<SwitchDecision>((resolve) => {
-      let resolved = false;
-      let modalInstance: { destroy: () => void } | null = null;
-      const settle = (decision: SwitchDecision) => {
-        if (resolved) return;
-        resolved = true;
-        resolve(decision);
-      };
-      const handleDiscard = () => {
-        if (modalInstance) modalInstance.destroy();
-        settle("discard");
-      };
-      modalInstance = modal.confirm({
-        title: "当前段落有未保存标注",
-        content: (
-          <div className="flex flex-col gap-2">
-            <Typography.Text>切换段落前请先保存当前标注。</Typography.Text>
-            <Button type="link" danger style={{ padding: 0, height: "auto" }} onClick={handleDiscard}>
-              放弃未保存并切换
-            </Button>
-          </div>
-        ),
-        okText: "保存并切换",
-        cancelText: "取消",
-        onOk: () => settle("save"),
-        onCancel: () => settle("cancel"),
-      });
-    });
-  }, [modal]);
-
   const requestExport = useCallback((autoAdvance: boolean) => {
     if (!selectedFileId) {
       message.warning("请先选择文件");
@@ -875,7 +683,7 @@ export default function LabelStudioTextEditor() {
   useEffect(() => {
     const handleSaveShortcut = (event: KeyboardEvent) => {
       if (!isSaveShortcut(event) || event.repeat) return;
-      if (saving || loadingTaskDetail || segmentSwitching) return;
+      if (saving || loadingTaskDetail) return;
       if (!iframeReady || !lsReady) return;
       event.preventDefault();
       event.stopPropagation();
@@ -883,83 +691,7 @@ export default function LabelStudioTextEditor() {
     };
     window.addEventListener("keydown", handleSaveShortcut);
     return () => window.removeEventListener("keydown", handleSaveShortcut);
-  }, [iframeReady, loadingTaskDetail, lsReady, requestExport, saving, segmentSwitching]);
+  }, [iframeReady, loadingTaskDetail, lsReady, requestExport, saving]);
-
-  // 段落切换处理
-  const handleSegmentChange = useCallback(async (newIndex: number) => {
-    if (newIndex === currentSegmentIndex) return;
-    if (segmentSwitching || saving || loadingTaskDetail) return;
-    if (!iframeReady || !lsReady) {
-      message.warning("编辑器未就绪，无法切换段落");
-      return;
-    }
-
-    setSegmentSwitching(true);
-    try {
-      const payload = await requestExportForCheck();
-      if (!payload) {
-        message.warning("无法读取当前标注，已取消切换");
-        return;
-      }
-
-      const payloadTaskId = payload.taskId;
-      if (expectedTaskIdRef.current && payloadTaskId) {
-        if (Number(payloadTaskId) !== expectedTaskIdRef.current) {
-          message.warning("已忽略过期的标注数据");
-          return;
-        }
-      }
-
-      const payloadFileId = payload.fileId || selectedFileId;
-      const payloadSegmentIndex = resolveSegmentIndex(payload.segmentIndex);
-      const resolvedSegmentIndex =
-        payloadSegmentIndex !== undefined
-          ? payloadSegmentIndex
-          : segmented
-            ? currentSegmentIndex
-            : undefined;
-      const annotation = isRecord(payload.annotation) ? payload.annotation : undefined;
-      const snapshotKey = payloadFileId
-        ? buildSnapshotKey(String(payloadFileId), resolvedSegmentIndex)
-        : undefined;
-      const latestSnapshot = buildAnnotationSnapshot(annotation);
-      const lastSnapshot = snapshotKey ? savedSnapshotsRef.current[snapshotKey] : undefined;
-      const hasUnsavedChange = snapshotKey !== undefined && lastSnapshot !== undefined && latestSnapshot !== lastSnapshot;
-
-      if (hasUnsavedChange) {
-        if (autoSaveOnSwitch) {
-          const saved = await saveFromExport(payload);
-          if (!saved) return;
-        } else {
-          const decision = await confirmSaveBeforeSwitch();
-          if (decision === "cancel") return;
-          if (decision === "save") {
-            const saved = await saveFromExport(payload);
-            if (!saved) return;
-          }
-        }
-      }
-
-      await initEditorForFile(selectedFileId, newIndex);
-    } finally {
-      setSegmentSwitching(false);
-    }
-  }, [
-    autoSaveOnSwitch,
-    confirmSaveBeforeSwitch,
-    currentSegmentIndex,
-    iframeReady,
-    initEditorForFile,
-    loadingTaskDetail,
-    lsReady,
-    message,
-    requestExportForCheck,
-    saveFromExport,
-    segmented,
-    selectedFileId,
-    segmentSwitching,
-    saving,
-  ]);
 
   useEffect(() => {
     setIframeReady(false);
@@ -977,12 +709,9 @@ export default function LabelStudioTextEditor() {
     expectedTaskIdRef.current = null;
     // 重置分段状态
     setSegmented(false);
-    setSegments([]);
     setCurrentSegmentIndex(0);
+    setSegmentTotal(0);
     savedSnapshotsRef.current = {};
-    segmentStatsSeqRef.current += 1;
-    segmentStatsCacheRef.current = {};
-    segmentStatsLoadingRef.current = new Set();
     if (exportCheckRef.current?.timer) {
       window.clearTimeout(exportCheckRef.current.timer);
     }
@@ -996,12 +725,6 @@ export default function LabelStudioTextEditor() {
     loadTasks({ mode: "reset" });
   }, [project?.supported, loadTasks]);
 
-  useEffect(() => {
-    if (!segmented) return;
-    if (tasks.length === 0) return;
-    prefetchSegmentStats(tasks);
-  }, [prefetchSegmentStats, segmented, tasks]);
-
   useEffect(() => {
     if (!selectedFileId) return;
     initEditorForFile(selectedFileId);
@@ -1026,60 +749,6 @@ export default function LabelStudioTextEditor() {
     return () => window.removeEventListener("focus", handleWindowFocus);
   }, [focusIframe, lsReady]);
 
-  const segmentTreeData = useMemo(() => {
-    if (!segmented || segments.length === 0) return [];
-    const lineMap = new Map<number, SegmentInfo[]>();
-    segments.forEach((seg) => {
-      const list = lineMap.get(seg.lineIndex) || [];
-      list.push(seg);
-      lineMap.set(seg.lineIndex, list);
-    });
-    return Array.from(lineMap.entries())
-      .sort((a, b) => a[0] - b[0])
-      .map(([lineIndex, lineSegments]) => ({
-        key: `line-${lineIndex}`,
-        title: `第${lineIndex + 1}行`,
-        selectable: false,
-        children: lineSegments
-          .sort((a, b) => a.chunkIndex - b.chunkIndex)
-          .map((seg) => ({
-            key: `seg-${seg.idx}`,
-            title: (
-              <span className="flex items-center gap-1">
-                <span>{`片${seg.chunkIndex + 1}`}</span>
-                {seg.hasAnnotation && (
-                  <CheckOutlined style={{ fontSize: 10, color: "#52c41a" }} />
-                )}
-              </span>
-            ),
-          })),
-      }));
-  }, [segmented, segments]);
-
-  const segmentLineKeys = useMemo(
-    () => segmentTreeData.map((item) => String(item.key)),
-    [segmentTreeData]
-  );
-
-  const inProgressSegmentedCount = useMemo(() => {
-    if (tasks.length === 0) return 0;
-    return tasks.reduce((count, item) => {
-      const summary = resolveSegmentSummary(item);
-      if (!summary) return count;
-      return summary.done < summary.total ? count + 1 : count;
-    }, 0);
-  }, [tasks]);
-
-  const handleSegmentSelect = useCallback((keys: Array<string | number>) => {
-    const [first] = keys;
-    if (first === undefined || first === null) return;
-    const key = String(first);
-    if (!key.startsWith("seg-")) return;
-    const nextIndex = Number(key.replace("seg-", ""));
-    if (!Number.isFinite(nextIndex)) return;
-    handleSegmentChange(nextIndex);
-  }, [handleSegmentChange]);
-
   useEffect(() => {
     const handler = (event: MessageEvent<LsfMessage>) => {
       if (event.origin !== origin) return;
@@ -1148,7 +817,7 @@ export default function LabelStudioTextEditor() {
 
   const canLoadMore = taskTotalPages > 0 && taskPage + 1 < taskTotalPages;
   const saveDisabled =
-    !iframeReady || !selectedFileId || saving || segmentSwitching || loadingTaskDetail;
+    !iframeReady || !selectedFileId || saving || loadingTaskDetail;
   const loadMoreNode = canLoadMore ? (
     <div className="p-2 text-center">
       <Button
@@ -1265,11 +934,6 @@ export default function LabelStudioTextEditor() {
         >
           <div className="px-3 py-2 border-b border-gray-200 bg-white font-medium text-sm flex items-center justify-between gap-2">
             <span>文件列表</span>
-            {segmented && (
-              <Tag color="orange" style={{ margin: 0 }}>
-                标注中 {inProgressSegmentedCount}
-              </Tag>
-            )}
           </div>
           <div className="flex-1 min-h-0 overflow-auto">
             <List
@@ -1278,7 +942,6 @@ export default function LabelStudioTextEditor() {
               dataSource={tasks}
               loadMore={loadMoreNode}
               renderItem={(item) => {
-                const segmentSummary = resolveSegmentSummary(item);
                 const statusMeta = resolveTaskStatusMeta(item);
                 return (
                   <List.Item
@@ -1300,11 +963,6 @@ export default function LabelStudioTextEditor() {
                             <Typography.Text type={statusMeta.type} style={{ fontSize: 11 }}>
                               {statusMeta.text}
                             </Typography.Text>
-                            {segmentSummary && (
-                              <Typography.Text type="secondary" style={{ fontSize: 10 }}>
-                                已标注 {segmentSummary.done}/{segmentSummary.total}
-                              </Typography.Text>
-                            )}
                           </div>
                           {item.annotationUpdatedAt && (
                             <Typography.Text type="secondary" style={{ fontSize: 10 }}>
@@ -1323,21 +981,28 @@ export default function LabelStudioTextEditor() {
               <div className="px-3 py-2 border-b border-gray-200 bg-gray-50 font-medium text-sm flex items-center justify-between">
                 <span>段落/分段</span>
                 <Tag color="blue" style={{ margin: 0 }}>
-                  {currentSegmentIndex + 1} / {segments.length}
+                  {segmentTotal > 0 ? currentSegmentIndex + 1 : 0} / {segmentTotal}
                 </Tag>
               </div>
               <div className="flex-1 min-h-0 overflow-auto px-2 py-2">
-                {segments.length > 0 ? (
+                {segmentTotal > 0 ? (
-                  <Tree
+                  <div className="grid grid-cols-[repeat(auto-fill,minmax(44px,1fr))] gap-1">
-                    showLine
+                    {segmentIndices.map((segmentIndex) => {
-                    blockNode
+                      const isCurrent = segmentIndex === currentSegmentIndex;
-                    selectedKeys={
+                      return (
-                      segmented ? [`seg-${currentSegmentIndex}`] : []
+                        <div
+                          key={segmentIndex}
+                          className={
+                            isCurrent
+                              ? "h-7 leading-7 rounded bg-blue-500 text-white text-center text-xs font-medium"
+                              : "h-7 leading-7 rounded bg-gray-100 text-gray-700 text-center text-xs"
                           }
-                    expandedKeys={segmentLineKeys}
+                        >
-                    onSelect={handleSegmentSelect}
+                          {segmentIndex + 1}
-                    treeData={segmentTreeData}
+                        </div>
-                  />
+                      );
+                    })}
+                  </div>
                 ) : (
                   <div className="py-6">
                     <Empty
@@ -1347,17 +1012,6 @@ export default function LabelStudioTextEditor() {
                   </div>
                 )}
               </div>
-              <div className="px-3 py-2 border-t border-gray-200 flex items-center justify-between">
-                <Typography.Text style={{ fontSize: 12 }}>
-                  切段自动保存
-                </Typography.Text>
-                <Switch
-                  size="small"
-                  checked={autoSaveOnSwitch}
-                  onChange={(checked) => setAutoSaveOnSwitch(checked)}
-                  disabled={segmentSwitching || saving || loadingTaskDetail || !lsReady}
-                />
-              </div>
             </div>
           )}
         </div>

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 +80,19 @@ export function getEditorTaskUsingGet(
   return get(`/api/annotation/editor/projects/${projectId}/tasks/${fileId}`, params);
 }
 
+export function getEditorTaskSegmentUsingGet(
+  projectId: string,
+  fileId: string,
+  params: { segmentIndex: number }
+) {
+  return get(`/api/annotation/editor/projects/${projectId}/tasks/${fileId}/segments`, params);
+}
+
 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;
@@ -234,6 +294,10 @@ export default function ImportConfiguration({
     if (!data) return;
     console.log('[ImportConfiguration] handleImportData called, currentPrefix:', currentPrefix);
     if (importConfig.source === DataSource.UPLOAD) {
+      // 立即显示任务中心，让用户感知上传已开始（在文件分割等耗时操作之前）
+      window.dispatchEvent(
+        new CustomEvent("show:task-popover", { detail: { show: true } })
+      );
       await handleUpload(data);
     } else if (importConfig.source === DataSource.COLLECTION) {
       await updateDatasetByIdUsingPut(data.id, {

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/db/models/dataset_management.py

@@ -61,13 +61,15 @@ class DatasetFiles(Base):
     dataset_id = Column(String(36), nullable=False, comment="所属数据集ID（UUID）")
     file_name = Column(String(255), nullable=False, comment="文件名")
     file_path = Column(String(1000), nullable=False, comment="文件路径")
+    logical_path = Column(String(1000), nullable=False, comment="文件逻辑路径（相对数据集根目录）")
+    version = Column(BigInteger, nullable=False, default=1, comment="文件版本号（同 logical_path 递增）")
     file_type = Column(String(50), nullable=True, comment="文件格式：JPG/PNG/DCM/TXT等")
     file_size = Column(BigInteger, default=0, comment="文件大小(字节)")
     check_sum = Column(String(64), nullable=True, comment="文件校验和")
     tags = Column(JSON, nullable=True, comment="文件标签信息")
     tags_updated_at = Column(TIMESTAMP, nullable=True, comment="标签最后更新时间")
     dataset_filemetadata = Column("metadata", JSON, nullable=True, comment="文件元数据")
-    status = Column(String(50), default='ACTIVE', comment="文件状态：ACTIVE/DELETED/PROCESSING")
+    status = Column(String(50), default='ACTIVE', comment="文件状态：ACTIVE/ARCHIVED/DELETED/PROCESSING")
     upload_time = Column(TIMESTAMP, server_default=func.current_timestamp(), comment="上传时间")
     last_access_time = Column(TIMESTAMP, nullable=True, comment="最后访问时间")
     created_at = Column(TIMESTAMP, server_default=func.current_timestamp(), comment="创建时间")

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

@@ -19,6 +19,7 @@ from app.db.session import get_db
 from app.module.annotation.schema.editor import (
     EditorProjectInfo,
     EditorTaskListResponse,
+    EditorTaskSegmentResponse,
     EditorTaskResponse,
     UpsertAnnotationRequest,
     UpsertAnnotationResponse,
@@ -87,6 +88,21 @@ 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[EditorTaskSegmentResponse],
+)
+async def get_editor_task_segment(
+    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）"),
+    segment_index: int = Query(..., ge=0, alias="segmentIndex", description="段落索引（从0开始）"),
+    db: AsyncSession = Depends(get_db),
+):
+    service = AnnotationEditorService(db)
+    result = await service.get_task_segment(project_id, file_id, segment_index)
+    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）")
@@ -100,7 +97,29 @@ 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 SegmentDetail(BaseModel):
+    """段落内容"""
+
+    idx: int = Field(..., description="段落索引")
+    text: str = Field(..., description="段落文本")
+    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）")
+
+    model_config = ConfigDict(populate_by_name=True)
+
+
+class EditorTaskSegmentResponse(BaseModel):
+    """编辑器单段内容响应"""
+
+    segmented: bool = Field(False, description="是否启用分段模式")
+    segment: Optional[SegmentDetail] = Field(None, description="段落内容")
     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")
 

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

@@ -36,7 +36,9 @@ from app.module.annotation.schema.editor import (
     EditorProjectInfo,
     EditorTaskListItem,
     EditorTaskListResponse,
+    EditorTaskSegmentResponse,
     EditorTaskResponse,
+    SegmentDetail,
     SegmentInfo,
     UpsertAnnotationRequest,
     UpsertAnnotationResponse,
@@ -538,6 +540,50 @@ 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)
+        segments: List[SegmentInfo] = []
+        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 +714,124 @@ class AnnotationEditorService:
 
         return await self._build_text_task(project, file_record, file_id, segment_index)
 
+    async def get_task_segment(
+        self,
+        project_id: str,
+        file_id: str,
+        segment_index: int,
+    ) -> EditorTaskSegmentResponse:
+        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 项目的段落内容",
+            )
+
+        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 EditorTaskSegmentResponse(
+                segmented=False,
+                segment=None,
+                totalSegments=0,
+                currentSegmentIndex=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 EditorTaskSegmentResponse(
+                segmented=False,
+                segment=None,
+                totalSegments=0,
+                currentSegmentIndex=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, segment_contexts = self._build_segment_contexts(
+            records,
+            record_texts,
+            segment_annotation_keys,
+        )
+
+        total_segments = len(segment_contexts)
+        if total_segments == 0:
+            return EditorTaskSegmentResponse(
+                segmented=False,
+                segment=None,
+                totalSegments=0,
+                currentSegmentIndex=0,
+            )
+
+        if segment_index < 0 or segment_index >= total_segments:
+            raise HTTPException(
+                status_code=400,
+                detail=f"segmentIndex 超出范围: {segment_index}",
+            )
+
+        segment_info = segments[segment_index]
+        _, _, segment_text, line_index, chunk_index = segment_contexts[segment_index]
+        segment_detail = SegmentDetail(
+            idx=segment_info.idx,
+            text=segment_text,
+            hasAnnotation=segment_info.has_annotation,
+            lineIndex=line_index,
+            chunkIndex=chunk_index,
+        )
+
+        return EditorTaskSegmentResponse(
+            segmented=True,
+            segment=segment_detail,
+            totalSegments=total_segments,
+            currentSegmentIndex=segment_index,
+        )
+
     async def _build_text_task(
         self,
         project: LabelingProject,
@@ -723,7 +887,8 @@ 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 +897,13 @@ 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 +981,7 @@ 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 +1316,195 @@ 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,
+        }
+

runtime/python-executor/datamate/auto_annotation_worker.py

@@ -375,9 +375,9 @@ def _register_output_dataset(
     insert_file_sql = text(
         """
         INSERT INTO t_dm_dataset_files (
-            id, dataset_id, file_name, file_path, file_type, file_size, status
+            id, dataset_id, file_name, file_path, logical_path, version, file_type, file_size, status
         ) VALUES (
-            :id, :dataset_id, :file_name, :file_path, :file_type, :file_size, :status
+            :id, :dataset_id, :file_name, :file_path, :logical_path, :version, :file_type, :file_size, :status
         )
         """
     )
@@ -395,6 +395,7 @@ def _register_output_dataset(
 
         for file_name, file_path, file_size in image_files:
             ext = os.path.splitext(file_name)[1].lstrip(".").upper() or None
+            logical_path = os.path.relpath(file_path, output_dir).replace("\\", "/")
             conn.execute(
                 insert_file_sql,
                 {
@@ -402,6 +403,8 @@ def _register_output_dataset(
                     "dataset_id": output_dataset_id,
                     "file_name": file_name,
                     "file_path": file_path,
+                    "logical_path": logical_path,
+                    "version": 1,
                     "file_type": ext,
                     "file_size": int(file_size),
                     "status": "ACTIVE",
@@ -411,6 +414,7 @@ def _register_output_dataset(
 
         for file_name, file_path, file_size in annotation_files:
             ext = os.path.splitext(file_name)[1].lstrip(".").upper() or None
+            logical_path = os.path.relpath(file_path, output_dir).replace("\\", "/")
             conn.execute(
                 insert_file_sql,
                 {
@@ -418,6 +422,8 @@ def _register_output_dataset(
                     "dataset_id": output_dataset_id,
                     "file_name": file_name,
                     "file_path": file_path,
+                    "logical_path": logical_path,
+                    "version": 1,
                     "file_type": ext,
                     "file_size": int(file_size),
                     "status": "ACTIVE",

runtime/python-executor/datamate/sql_manager/sql/sql_config.json

@@ -1,9 +1,9 @@
 {
   "query_sql": "SELECT * FROM t_task_instance_info WHERE instance_id IN (:instance_id)",
   "insert_sql": "INSERT INTO t_task_instance_info (instance_id, meta_file_name, meta_file_type, meta_file_id, meta_file_size, file_id, file_size, file_type, file_name, file_path, status, operator_id, error_code, incremental, child_id, slice_num) VALUES (:instance_id, :meta_file_name, :meta_file_type, :meta_file_id, :meta_file_size, :file_id, :file_size, :file_type, :file_name, :file_path, :status, :operator_id, :error_code, :incremental, :child_id, :slice_num)",
-  "insert_dataset_file_sql": "INSERT INTO t_dm_dataset_files (id, dataset_id, file_name, file_path, file_type, file_size, status, upload_time, last_access_time, created_at, updated_at) VALUES (:id, :dataset_id, :file_name, :file_path, :file_type, :file_size, :status, :upload_time, :last_access_time, :created_at, :updated_at)",
+  "insert_dataset_file_sql": "INSERT INTO t_dm_dataset_files (id, dataset_id, file_name, file_path, logical_path, version, file_type, file_size, status, upload_time, last_access_time, created_at, updated_at) VALUES (:id, :dataset_id, :file_name, :file_path, :logical_path, :version, :file_type, :file_size, :status, :upload_time, :last_access_time, :created_at, :updated_at)",
   "insert_clean_result_sql": "INSERT INTO t_clean_result (instance_id, src_file_id, dest_file_id, src_name, dest_name, src_type, dest_type, src_size, dest_size, status, result) VALUES (:instance_id, :src_file_id, :dest_file_id, :src_name, :dest_name, :src_type, :dest_type, :src_size, :dest_size, :status, :result)",
-  "query_dataset_sql": "SELECT file_size FROM t_dm_dataset_files WHERE dataset_id = :dataset_id",
+  "query_dataset_sql": "SELECT file_size FROM t_dm_dataset_files WHERE dataset_id = :dataset_id AND (status IS NULL OR status <> 'ARCHIVED')",
   "update_dataset_sql": "UPDATE t_dm_datasets SET size_bytes = :total_size, file_count = :file_count WHERE id = :dataset_id;",
   "update_task_sql": "UPDATE t_clean_task SET status = :status, after_size = :total_size, finished_at = :finished_time WHERE id = :task_id",
   "create_tables_sql": "CREATE TABLE IF NOT EXISTS t_task_instance_info (instance_id VARCHAR(255), meta_file_name TEXT, meta_file_type VARCHAR(100), meta_file_id BIGINT, meta_file_size VARCHAR(100), file_id BIGINT, file_size VARCHAR(100), file_type VARCHAR(100), file_name TEXT, file_path TEXT, status INT, operator_id VARCHAR(255), error_code VARCHAR(100), incremental VARCHAR(50), child_id BIGINT, slice_num INT DEFAULT 0);",

scripts/db/data-management-init.sql

@@ -54,19 +54,22 @@ CREATE TABLE IF NOT EXISTS t_dm_dataset_files (
     dataset_id VARCHAR(36) NOT NULL COMMENT '所属数据集ID（UUID）',
     file_name VARCHAR(255) NOT NULL COMMENT '文件名',
     file_path VARCHAR(1000) NOT NULL COMMENT '文件路径',
+    logical_path VARCHAR(1000) NOT NULL COMMENT '文件逻辑路径（相对数据集根目录）',
+    version BIGINT NOT NULL DEFAULT 1 COMMENT '文件版本号（同 logical_path 递增）',
     file_type VARCHAR(50) COMMENT '文件格式：JPG/PNG/DCM/TXT等',
     file_size BIGINT DEFAULT 0 COMMENT '文件大小(字节)',
     check_sum VARCHAR(64) COMMENT '文件校验和',
     tags      JSON COMMENT '文件标签信息',
     tags_updated_at TIMESTAMP NULL COMMENT '标签最后更新时间',
     metadata JSON COMMENT '文件元数据',
-    status VARCHAR(50) DEFAULT 'ACTIVE' COMMENT '文件状态：ACTIVE/DELETED/PROCESSING',
+    status VARCHAR(50) DEFAULT 'ACTIVE' COMMENT '文件状态：ACTIVE/ARCHIVED/DELETED/PROCESSING',
     upload_time TIMESTAMP DEFAULT CURRENT_TIMESTAMP COMMENT '上传时间',
     last_access_time TIMESTAMP NULL COMMENT '最后访问时间',
     created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP COMMENT '创建时间',
     updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT '更新时间',
     FOREIGN KEY (dataset_id) REFERENCES t_dm_datasets(id) ON DELETE CASCADE,
     INDEX idx_dm_dataset (dataset_id),
+    INDEX idx_dm_dataset_logical_path (dataset_id, logical_path, version),
     INDEX idx_dm_file_type (file_type),
     INDEX idx_dm_file_status (status),
     INDEX idx_dm_upload_time (upload_time)