feat: 支持知识库导入 PPTX 与 XLSX 文档

- 打通 Office 文档桥接解析、解析进度承接与图片引用改写

- 落地 PPTX 按页分块、XLSX 行窗口分块以及预览与检索渲染闭环

easyflow-api/easyflow-api-admin/src/main/java/tech/easyflow/admin/controller/ai/DocumentCollectionController.java

@@ -407,6 +407,8 @@ public class DocumentCollectionController extends BaseCurdController<DocumentCol
             KnowledgeSearchResultItem item = new KnowledgeSearchResultItem();
             item.setSorting(index + 1);
             item.setContent(document.getContent());
+            item.setRenderMarkdown(readMetadataAsString(document, "renderMarkdown"));
+            item.setSourceFileName(readMetadataAsString(document, "sourceFileName"));
             item.setScore(roundScore(document.getScore()));
             item.setHitSource(readMetadataAsString(document, RagRetrievalMetadataKeys.HIT_SOURCE));
             item.setVectorScore(roundScore(readMetadataAsDouble(document, RagRetrievalMetadataKeys.VECTOR_SCORE)));

easyflow-api/easyflow-api-admin/src/main/java/tech/easyflow/admin/controller/ai/ShareKnowledgeController.java

@@ -932,6 +932,10 @@ public class ShareKnowledgeController {
             KnowledgeSearchResultItem item = new KnowledgeSearchResultItem();
             item.setSorting(index + 1);
             item.setContent(document.getContent());
+            Object renderMarkdown = document.getMetadata("renderMarkdown");
+            item.setRenderMarkdown(renderMarkdown == null ? null : String.valueOf(renderMarkdown));
+            Object sourceFileName = document.getMetadata("sourceFileName");
+            item.setSourceFileName(sourceFileName == null ? null : String.valueOf(sourceFileName));
             item.setScore(document.getScore() == null ? null : document.getScore().doubleValue());
             Object hitSource = document.getMetadata("hitSource");
             item.setHitSource(hitSource == null ? null : String.valueOf(hitSource));

easyflow-api/easyflow-api-public/src/main/java/tech/easyflow/publicapi/controller/PublicKnowledgeShareController.java

@@ -638,6 +638,10 @@ public class PublicKnowledgeShareController {
         for (com.easyagents.core.document.Document document : documents) {
             KnowledgeSearchResultItem item = new KnowledgeSearchResultItem();
             item.setContent(document.getContent());
+            Object renderMarkdown = document.getMetadata("renderMarkdown");
+            item.setRenderMarkdown(renderMarkdown == null ? null : String.valueOf(renderMarkdown));
+            Object sourceFileName = document.getMetadata("sourceFileName");
+            item.setSourceFileName(sourceFileName == null ? null : String.valueOf(sourceFileName));
             item.setScore(document.getScore());
             Object hitSource = document.getMetadata("hitSource");
             item.setHitSource(hitSource == null ? null : String.valueOf(hitSource));

easyflow-modules/easyflow-module-ai/pom.xml

@@ -112,7 +112,6 @@
             <groupId>com.easyagents</groupId>
             <artifactId>easy-agents-mcp</artifactId>
         </dependency>
-
         <dependency>
             <groupId>junit</groupId>
             <artifactId>junit</artifactId>

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/document/exception/DocumentParseBridgeException.java

@@ -35,10 +35,14 @@ public class DocumentParseBridgeException extends RuntimeException {
     public static DocumentParseBridgeException serviceNotEnabled() {
         return new DocumentParseBridgeException(
             "service_not_enabled",
-            "统一文档解析服务未启用，请先配置 easy-agents.document.pdf.provider"
+            "统一文档解析服务未启用，请先配置 easy-agents.document.ocr.provider=mineru"
         );
     }
 
+    public static DocumentParseBridgeException serviceNotEnabled(String message) {
+        return new DocumentParseBridgeException("service_not_enabled", message);
+    }
+
     public static DocumentParseBridgeException unsupportedSource(String message) {
         return new DocumentParseBridgeException("unsupported_source", message);
     }

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/document/model/DocumentParseTaskStatus.java

@@ -22,6 +22,11 @@ public class DocumentParseTaskStatus {
     private String statusUrl;
     private String resultUrl;
     private Integer queuedAhead;
+    private Integer progressPercent;
+    private String currentStage;
+    private Integer processedItems;
+    private Integer totalItems;
+    private String statusMessage;
 
     public String getTaskId() {
         return taskId;
@@ -110,4 +115,44 @@ public class DocumentParseTaskStatus {
     public void setQueuedAhead(Integer queuedAhead) {
         this.queuedAhead = queuedAhead;
     }
+
+    public Integer getProgressPercent() {
+        return progressPercent;
+    }
+
+    public void setProgressPercent(Integer progressPercent) {
+        this.progressPercent = progressPercent;
+    }
+
+    public String getCurrentStage() {
+        return currentStage;
+    }
+
+    public void setCurrentStage(String currentStage) {
+        this.currentStage = currentStage;
+    }
+
+    public Integer getProcessedItems() {
+        return processedItems;
+    }
+
+    public void setProcessedItems(Integer processedItems) {
+        this.processedItems = processedItems;
+    }
+
+    public Integer getTotalItems() {
+        return totalItems;
+    }
+
+    public void setTotalItems(Integer totalItems) {
+        this.totalItems = totalItems;
+    }
+
+    public String getStatusMessage() {
+        return statusMessage;
+    }
+
+    public void setStatusMessage(String statusMessage) {
+        this.statusMessage = statusMessage;
+    }
 }

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/document/service/impl/DocumentParseBridgeServiceImpl.java

@@ -5,6 +5,10 @@ import com.easyagents.document.core.entity.ParseResponse;
 import com.easyagents.document.core.entity.ParseResult;
 import com.easyagents.document.core.entity.ParseTaskInfo;
 import com.easyagents.document.core.entity.ParseTaskStatus;
+import com.easyagents.document.pdf.PdfDocumentParseService;
+import com.easyagents.document.pptx.PptxDocumentParseService;
+import com.easyagents.document.xlsx.XlsxDocumentParseService;
+import org.springframework.beans.factory.annotation.Qualifier;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 import org.springframework.lang.Nullable;
@@ -20,8 +24,13 @@ import tech.easyflow.ai.document.service.DocumentParseBridgeService;
 import tech.easyflow.ai.document.support.DocumentSourceLoader;
 import tech.easyflow.ai.document.support.DocumentParseRequestFactory;
 import tech.easyflow.ai.document.support.DocumentParseResultMapper;
+import tech.easyflow.ai.document.support.DocumentParseSourceType;
 import tech.easyflow.ai.document.support.LoadedDocumentSource;
-import tech.easyflow.ai.utils.DocUtil;
+
+import java.util.ArrayList;
+import java.util.LinkedHashSet;
+import java.util.List;
+import java.util.function.Function;
 
 /**
  * 统一文档解析桥接门面默认实现。
@@ -33,18 +42,33 @@ import tech.easyflow.ai.utils.DocUtil;
 public class DocumentParseBridgeServiceImpl implements DocumentParseBridgeService {
 
     private static final Logger LOG = LoggerFactory.getLogger(DocumentParseBridgeServiceImpl.class);
+    private static final String DEFAULT_DOCUMENT_PARSE_SERVICE_BEAN_NAME = "documentParseService";
 
     @Nullable
-    private final DocumentParseService documentParseService;
+    private final DocumentParseService defaultDocumentParseService;
+    @Nullable
+    private final PdfDocumentParseService pdfDocumentParseService;
+    @Nullable
+    private final PptxDocumentParseService pptxDocumentParseService;
+    @Nullable
+    private final XlsxDocumentParseService xlsxDocumentParseService;
     private final DocumentSourceLoader documentSourceLoader;
     private final DocumentParseRequestFactory parseRequestFactory;
     private final DocumentParseResultMapper parseResultMapper;
 
-    public DocumentParseBridgeServiceImpl(@Nullable DocumentParseService documentParseService,
+    public DocumentParseBridgeServiceImpl(@Nullable
+                                          @Qualifier(DEFAULT_DOCUMENT_PARSE_SERVICE_BEAN_NAME)
+                                          DocumentParseService defaultDocumentParseService,
+                                          @Nullable PdfDocumentParseService pdfDocumentParseService,
+                                          @Nullable PptxDocumentParseService pptxDocumentParseService,
+                                          @Nullable XlsxDocumentParseService xlsxDocumentParseService,
                                           DocumentSourceLoader documentSourceLoader,
                                           DocumentParseRequestFactory parseRequestFactory,
                                           DocumentParseResultMapper parseResultMapper) {
-        this.documentParseService = documentParseService;
+        this.defaultDocumentParseService = defaultDocumentParseService;
+        this.pdfDocumentParseService = pdfDocumentParseService;
+        this.pptxDocumentParseService = pptxDocumentParseService;
+        this.xlsxDocumentParseService = xlsxDocumentParseService;
         this.documentSourceLoader = documentSourceLoader;
         this.parseRequestFactory = parseRequestFactory;
         this.parseResultMapper = parseResultMapper;
@@ -59,7 +83,8 @@ public class DocumentParseBridgeServiceImpl implements DocumentParseBridgeServic
             LoadedDocumentSource loadedSource = prepareSupportedSource(source);
             LOG.info("桥接服务开始同步解析文档: fileName={}, contentType={}, scenario={}",
                 loadedSource.getFileName(), loadedSource.getContentType(), scenario);
-            ParseResponse response = requireService().parse(parseRequestFactory.build(loadedSource, scenario));
+            DocumentParseService parseService = resolveService(loadedSource);
+            ParseResponse response = parseService.parse(parseRequestFactory.build(loadedSource, scenario));
             DocumentParsedResult result = parseResultMapper.map(extractSingleResult(response, false));
             LOG.info("桥接服务同步解析完成: fileName={}, scenario={}, preferredTextLength={}",
                 loadedSource.getFileName(), scenario, resolveTextLength(result));
@@ -84,7 +109,8 @@ public class DocumentParseBridgeServiceImpl implements DocumentParseBridgeServic
             LoadedDocumentSource loadedSource = prepareSupportedSource(source);
             LOG.info("桥接服务开始提交异步解析任务: fileName={}, contentType={}, scenario={}",
                 loadedSource.getFileName(), loadedSource.getContentType(), scenario);
-            ParseTaskStatus taskStatus = requireService().submit(parseRequestFactory.build(loadedSource, scenario));
+            DocumentParseService parseService = resolveService(loadedSource);
+            ParseTaskStatus taskStatus = parseService.submit(parseRequestFactory.build(loadedSource, scenario));
             DocumentParseTaskStatus mappedStatus = parseResultMapper.map(taskStatus);
             LOG.info("桥接服务异步解析任务提交完成: fileName={}, scenario={}, providerTaskId={}, status={}",
                 loadedSource.getFileName(), scenario, mappedStatus.getTaskId(), mappedStatus.getStatus());
@@ -109,7 +135,8 @@ public class DocumentParseBridgeServiceImpl implements DocumentParseBridgeServic
             throw DocumentParseBridgeException.taskFailed("taskId 不能为空");
         }
         try {
-            return parseResultMapper.map(requireService().queryTask(taskId));
+            ParseTaskStatus taskStatus = executeAgainstTaskService(taskId, service -> service.queryTask(taskId));
+            return parseResultMapper.map(taskStatus);
         } catch (DocumentParseBridgeException e) {
             throw e;
         } catch (Exception e) {
@@ -127,7 +154,7 @@ public class DocumentParseBridgeServiceImpl implements DocumentParseBridgeServic
         }
         try {
             LOG.info("桥接服务开始获取异步解析结果: providerTaskId={}", taskId);
-            ParseResponse response = requireService().queryResult(taskId);
+            ParseResponse response = executeAgainstTaskService(taskId, service -> service.queryResult(taskId));
             DocumentParsedResult result = parseResultMapper.map(extractSingleResult(response, true));
             LOG.info("桥接服务获取异步解析结果完成: providerTaskId={}, preferredTextLength={}",
                 taskId, resolveTextLength(result));
@@ -150,7 +177,7 @@ public class DocumentParseBridgeServiceImpl implements DocumentParseBridgeServic
             throw DocumentParseBridgeException.taskFailed("taskId 不能为空");
         }
         try {
-            ParseTaskInfo taskInfo = requireService().queryTaskInfo(taskId);
+            ParseTaskInfo taskInfo = executeAgainstTaskService(taskId, service -> service.queryTaskInfo(taskId));
             DocumentParseTaskInfo mappedTaskInfo = parseResultMapper.map(taskInfo);
             LOG.info("桥接服务查询异步解析任务状态: providerTaskId={}, status={}, hasResult={}",
                 taskId,
@@ -177,39 +204,84 @@ public class DocumentParseBridgeServiceImpl implements DocumentParseBridgeServic
         return text == null ? 0 : text.length();
     }
 
-    private DocumentParseService requireService() {
-        if (documentParseService == null) {
-            throw DocumentParseBridgeException.serviceNotEnabled();
-        }
-        return documentParseService;
-    }
-
     private LoadedDocumentSource prepareSupportedSource(DocumentSourceRef source) {
         LoadedDocumentSource loadedSource = documentSourceLoader.load(source);
         if (!isSupportedByBridge(loadedSource)) {
-            throw DocumentParseBridgeException.unsupportedSource("统一文档解析桥接当前仅支持 PDF、DOCX 文件");
+            throw DocumentParseBridgeException.unsupportedSource("统一文档解析桥接当前仅支持 PDF、DOCX、PPTX、XLSX 文件");
         }
         return loadedSource;
     }
 
     private boolean isSupportedByBridge(LoadedDocumentSource loadedSource) {
-        String contentType = loadedSource.getContentType();
+        return DocumentParseSourceType.resolve(loadedSource.getFileName(), loadedSource.getContentType()) != DocumentParseSourceType.UNSUPPORTED;
-        if (StringUtils.hasText(contentType)) {
+    }
-            String normalizedContentType = contentType.toLowerCase();
+
-            if (normalizedContentType.contains("pdf")
+    private DocumentParseService resolveService(LoadedDocumentSource loadedSource) {
-                || normalizedContentType.contains("wordprocessingml.document")) {
+        DocumentParseSourceType sourceType = DocumentParseSourceType.resolve(loadedSource.getFileName(), loadedSource.getContentType());
-                return true;
+        switch (sourceType) {
+            case PDF:
+                return requireSpecificService(pdfDocumentParseService, defaultDocumentParseService, "PDF");
+            case DOCX:
+                return requireSpecificService(defaultDocumentParseService, pdfDocumentParseService, "DOCX");
+            case PPTX:
+                return requireSpecificService(pptxDocumentParseService, null, "PPTX");
+            case XLSX:
+                return requireSpecificService(xlsxDocumentParseService, null, "XLSX");
+            default:
+                throw DocumentParseBridgeException.unsupportedSource("当前文件类型暂不支持桥接解析");
+        }
+    }
+
+    private DocumentParseService requireSpecificService(@Nullable DocumentParseService primaryService,
+                                                        @Nullable DocumentParseService fallbackService,
+                                                        String sourceType) {
+        if (primaryService != null) {
+            return primaryService;
+        }
+        if (fallbackService != null) {
+            return fallbackService;
+        }
+        throw DocumentParseBridgeException.serviceNotEnabled("未启用 " + sourceType + " 文档解析服务");
+    }
+
+    private <T> T executeAgainstTaskService(String taskId, Function<DocumentParseService, T> action) {
+        List<DocumentParseService> services = availableServices();
+        if (services.isEmpty()) {
+            throw DocumentParseBridgeException.serviceNotEnabled();
+        }
+        Exception lastException = null;
+        for (DocumentParseService service : services) {
+            try {
+                return action.apply(service);
+            } catch (Exception exception) {
+                lastException = exception;
+                LOG.debug("桥接服务任务查询尝试失败，准备切换下一个解析服务: taskId={}, service={}",
+                    taskId,
+                    service.getClass().getSimpleName(),
+                    exception);
             }
         }
-        String fileName = loadedSource.getFileName();
+        if (lastException instanceof RuntimeException) {
-        if (!StringUtils.hasText(fileName) || !fileName.contains(".")) {
+            throw (RuntimeException) lastException;
-            return false;
         }
-        String suffix = DocUtil.normalizeSuffix(DocUtil.getSuffix(fileName));
+        throw DocumentParseBridgeException.taskFailed("未找到可处理当前任务ID的文档解析服务", lastException);
-        if ("pdf".equals(suffix) || "docx".equals(suffix)) {
+    }
-            return true;
+
+    private List<DocumentParseService> availableServices() {
+        LinkedHashSet<DocumentParseService> services = new LinkedHashSet<DocumentParseService>();
+        if (pptxDocumentParseService != null) {
+            services.add(pptxDocumentParseService);
         }
-        return false;
+        if (xlsxDocumentParseService != null) {
+            services.add(xlsxDocumentParseService);
+        }
+        if (pdfDocumentParseService != null) {
+            services.add(pdfDocumentParseService);
+        }
+        if (defaultDocumentParseService != null) {
+            services.add(defaultDocumentParseService);
+        }
+        return new ArrayList<DocumentParseService>(services);
     }
 
     private ParseResult extractSingleResult(ParseResponse response, boolean resultFetchPhase) {

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/document/support/DocumentParseRequestFactory.java

@@ -2,6 +2,9 @@ package tech.easyflow.ai.document.support;
 
 import com.easyagents.document.core.entity.ParseFile;
 import com.easyagents.document.core.entity.ParseRequest;
+import com.easyagents.document.core.entity.PdfParseRequest;
+import com.easyagents.document.core.entity.PptxParseRequest;
+import com.easyagents.document.core.entity.XlsxParseRequest;
 import org.springframework.stereotype.Component;
 import tech.easyflow.ai.document.exception.DocumentParseBridgeException;
 import tech.easyflow.ai.document.model.DocumentParseScenario;
@@ -31,12 +34,28 @@ public class DocumentParseRequestFactory {
         if (scenario == null) {
             throw DocumentParseBridgeException.requestBuildFailed("解析场景不能为空");
         }
-        ParseRequest request = new ParseRequest();
+        ParseRequest request = createTypedRequest(source);
         request.addFile(ParseFile.of(source.getFileName(), source.getContentBytes(), source.getContentType()));
         applyScenario(request, scenario);
         return request;
     }
 
+    private ParseRequest createTypedRequest(LoadedDocumentSource source) {
+        DocumentParseSourceType sourceType = DocumentParseSourceType.resolve(source.getFileName(), source.getContentType());
+        switch (sourceType) {
+            case PDF:
+                return new PdfParseRequest();
+            case PPTX:
+                return new PptxParseRequest();
+            case XLSX:
+                return new XlsxParseRequest();
+            case DOCX:
+                return new ParseRequest();
+            default:
+                throw DocumentParseBridgeException.requestBuildFailed("当前文件类型暂不支持桥接解析");
+        }
+    }
+
     private void applyScenario(ParseRequest request, DocumentParseScenario scenario) {
         switch (scenario) {
             case WORKFLOW_TEXT:

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/document/support/DocumentParseResultMapper.java

@@ -69,6 +69,11 @@ public class DocumentParseResultMapper {
         status.setStatusUrl(taskStatus.getStatusUrl());
         status.setResultUrl(taskStatus.getResultUrl());
         status.setQueuedAhead(taskStatus.getQueuedAhead());
+        status.setProgressPercent(taskStatus.getProgressPercent());
+        status.setCurrentStage(taskStatus.getCurrentStage());
+        status.setProcessedItems(taskStatus.getProcessedItems());
+        status.setTotalItems(taskStatus.getTotalItems());
+        status.setStatusMessage(taskStatus.getStatusMessage());
         return status;
     }
 
@@ -104,6 +109,11 @@ public class DocumentParseResultMapper {
         status.setStatusUrl(taskStatus.getStatusUrl());
         status.setResultUrl(taskStatus.getResultUrl());
         status.setQueuedAhead(taskStatus.getQueuedAhead());
+        status.setProgressPercent(taskStatus.getProgressPercent());
+        status.setCurrentStage(taskStatus.getCurrentStage());
+        status.setProcessedItems(taskStatus.getProcessedItems());
+        status.setTotalItems(taskStatus.getTotalItems());
+        status.setStatusMessage(taskStatus.getStatusMessage());
     }
 
     private String resolvePreferredText(ParseResult parseResult) {

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/document/support/DocumentParseSourceType.java

@@ -0,0 +1,70 @@
+package tech.easyflow.ai.document.support;
+
+import org.springframework.util.StringUtils;
+import tech.easyflow.ai.utils.DocUtil;
+
+/**
+ * 统一文档解析桥接支持的源文件类型。
+ *
+ * @author Codex
+ * @since 2026-04-17
+ */
+public enum DocumentParseSourceType {
+
+    PDF,
+    DOCX,
+    PPTX,
+    XLSX,
+    UNSUPPORTED;
+
+    /**
+     * 根据文件名与内容类型推断文档类型。
+     *
+     * @param fileName 文件名
+     * @param contentType MIME 类型
+     * @return 文档类型
+     */
+    public static DocumentParseSourceType resolve(String fileName, String contentType) {
+        if (StringUtils.hasText(contentType)) {
+            String normalizedContentType = contentType.toLowerCase();
+            if (normalizedContentType.contains("pdf")) {
+                return PDF;
+            }
+            if (normalizedContentType.contains("wordprocessingml.document")) {
+                return DOCX;
+            }
+            if (normalizedContentType.contains("presentationml.presentation")) {
+                return PPTX;
+            }
+            if (normalizedContentType.contains("spreadsheetml.sheet")) {
+                return XLSX;
+            }
+        }
+        if (!StringUtils.hasText(fileName) || !fileName.contains(".")) {
+            return UNSUPPORTED;
+        }
+        String suffix = DocUtil.normalizeSuffix(DocUtil.getSuffix(fileName));
+        if ("pdf".equals(suffix)) {
+            return PDF;
+        }
+        if ("docx".equals(suffix)) {
+            return DOCX;
+        }
+        if ("pptx".equals(suffix)) {
+            return PPTX;
+        }
+        if ("xlsx".equals(suffix)) {
+            return XLSX;
+        }
+        return UNSUPPORTED;
+    }
+
+    /**
+     * 判断是否属于 Office 首版接入类型。
+     *
+     * @return 是否是本次 Office 类型
+     */
+    public boolean isOffice() {
+        return this == PPTX || this == XLSX;
+    }
+}

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/documentimport/DocumentImportDtos.java

@@ -286,6 +286,7 @@ public final class DocumentImportDtos {
         private String chunkId;
         private String chunkType;
         private String content;
+        private String renderMarkdown;
         private List<String> headingPath = new ArrayList<String>();
         private Integer partNo;
         private Integer partTotal;
@@ -335,6 +336,14 @@ public final class DocumentImportDtos {
             this.content = content;
         }
 
+        public String getRenderMarkdown() {
+            return renderMarkdown;
+        }
+
+        public void setRenderMarkdown(String renderMarkdown) {
+            this.renderMarkdown = renderMarkdown;
+        }
+
         public List<String> getHeadingPath() {
             return headingPath;
         }

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/documentimport/DocumentImportKeys.java

@@ -22,4 +22,19 @@ public final class DocumentImportKeys {
     public static final String KEY_DOCUMENT_PARSE_METADATA = "parse.metadata";
     public static final String KEY_DOCUMENT_PARSE_WARNINGS = "parse.warnings";
     public static final String KEY_DOCUMENT_PROVIDER_TASK_ID = "parse.providerTaskId";
+    public static final String KEY_DOCUMENT_PARSE_IMAGE_URLS = "parse.imageUrls";
+    public static final String KEY_DOCUMENT_PARSE_IMAGE_COUNT = "parse.imageCount";
+    public static final String KEY_DOCUMENT_PARSE_IMAGE_STORAGE_PREFIX = "parse.imageStoragePrefix";
+    public static final String KEY_DOCUMENT_PARSE_PROGRESS_PERCENT = "parse.progressPercent";
+    public static final String KEY_DOCUMENT_PARSE_CURRENT_STAGE = "parse.currentStage";
+    public static final String KEY_DOCUMENT_PARSE_PROCESSED_ITEMS = "parse.processedItems";
+    public static final String KEY_DOCUMENT_PARSE_TOTAL_ITEMS = "parse.totalItems";
+    public static final String KEY_DOCUMENT_PARSE_STATUS_MESSAGE = "parse.statusMessage";
+    public static final String KEY_DOCUMENT_RENDER_MARKDOWN = "renderMarkdown";
+    public static final String KEY_DOCUMENT_PAGE_INDEX = "pageIndex";
+    public static final String KEY_DOCUMENT_SHEET_NAME = "sheetName";
+    public static final String KEY_DOCUMENT_ROW_START = "rowStart";
+    public static final String KEY_DOCUMENT_ROW_END = "rowEnd";
+    public static final String KEY_DOCUMENT_IMAGE_REFS = "imageRefs";
+    public static final String KEY_DOCUMENT_PARSE_ARTIFACT_SUMMARY = "parseArtifactSummary";
 }

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/documentimport/task/DocumentImportTaskStatusStreamService.java

@@ -6,6 +6,7 @@ import org.springframework.stereotype.Service;
 import org.springframework.transaction.support.TransactionSynchronization;
 import org.springframework.transaction.support.TransactionSynchronizationManager;
 import org.springframework.web.servlet.mvc.method.annotation.SseEmitter;
+import tech.easyflow.ai.documentimport.DocumentImportKeys;
 import tech.easyflow.ai.entity.Document;
 import tech.easyflow.ai.mapper.DocumentMapper;
 import tech.easyflow.common.web.exceptions.BusinessException;
@@ -116,11 +117,21 @@ public class DocumentImportTaskStatusStreamService {
         payload.put("totalChunks", document.getTotalChunks());
         payload.put("completedChunks", document.getCompletedChunks());
         payload.put("failedChunks", document.getFailedChunks());
+        payload.put("parseCurrentStage", readOptionAsString(document, DocumentImportKeys.KEY_DOCUMENT_PARSE_CURRENT_STAGE));
+        payload.put("parseStatusMessage", readOptionAsString(document, DocumentImportKeys.KEY_DOCUMENT_PARSE_STATUS_MESSAGE));
         payload.put("lastTaskError", document.getLastTaskError());
         payload.put("taskModifiedAt", document.getTaskModifiedAt());
         return payload;
     }
 
+    private String readOptionAsString(Document document, String key) {
+        if (document == null || document.getOptions() == null || key == null) {
+            return null;
+        }
+        Object value = document.getOptions().get(key);
+        return value == null ? null : String.valueOf(value);
+    }
+
     private void sendAsync(String topicKey, SseEmitter emitter, String eventName, Map<String, Object> payload) {
         sseThreadPool.execute(() -> {
             try {

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/dto/KnowledgeSearchResultItem.java

@@ -4,6 +4,8 @@ public class KnowledgeSearchResultItem {
 
     private Integer sorting;
     private String content;
+    private String renderMarkdown;
+    private String sourceFileName;
     private Double score;
     private String hitSource;
     private Double vectorScore;
@@ -25,6 +27,22 @@ public class KnowledgeSearchResultItem {
         this.content = content;
     }
 
+    public String getRenderMarkdown() {
+        return renderMarkdown;
+    }
+
+    public void setRenderMarkdown(String renderMarkdown) {
+        this.renderMarkdown = renderMarkdown;
+    }
+
+    public String getSourceFileName() {
+        return sourceFileName;
+    }
+
+    public void setSourceFileName(String sourceFileName) {
+        this.sourceFileName = sourceFileName;
+    }
+
     public Double getScore() {
         return score;
     }

easyflow-modules/easyflow-module-ai/src/main/java/tech/easyflow/ai/service/impl/DocumentCollectionServiceImpl.java

@@ -32,6 +32,7 @@ import tech.easyflow.ai.entity.FaqItem;
 import tech.easyflow.ai.entity.Model;
 import tech.easyflow.ai.enums.DocumentProcessStatus;
 import tech.easyflow.ai.enums.PublishStatus;
+import tech.easyflow.ai.documentimport.DocumentImportKeys;
 import tech.easyflow.ai.mapper.DocumentChunkMapper;
 import tech.easyflow.ai.mapper.DocumentCollectionMapper;
 import tech.easyflow.ai.mapper.DocumentMapper;
@@ -406,6 +407,14 @@ public class DocumentCollectionServiceImpl extends ServiceImpl<DocumentCollectio
                     return false;
                 }
                 item.setContent(content);
+                String renderMarkdown = hitSnapshot.findChunkRenderMarkdown(item.getId());
+                if (StringUtil.hasText(renderMarkdown)) {
+                    item.addMetadata("renderMarkdown", renderMarkdown);
+                }
+                String sourceFileName = hitSnapshot.findSourceFileName(item.getId());
+                if (StringUtil.hasText(sourceFileName)) {
+                    item.addMetadata("sourceFileName", sourceFileName);
+                }
                 return true;
             })
             .collect(Collectors.toList());
@@ -596,6 +605,30 @@ public class DocumentCollectionServiceImpl extends ServiceImpl<DocumentCollectio
             }
             return StringUtil.noText(documentChunk.getContent()) ? null : documentChunk.getContent();
         }
+
+        private String findChunkRenderMarkdown(Object chunkId) {
+            DocumentChunk documentChunk = chunkMap.get(String.valueOf(chunkId));
+            if (documentChunk == null || documentChunk.getDocumentId() == null || documentChunk.getOptions() == null) {
+                return null;
+            }
+            if (!documentMap.containsKey(String.valueOf(documentChunk.getDocumentId()))) {
+                return null;
+            }
+            Object renderMarkdown = documentChunk.getOptions().get(DocumentImportKeys.KEY_DOCUMENT_RENDER_MARKDOWN);
+            return renderMarkdown == null ? null : String.valueOf(renderMarkdown);
+        }
+
+        private String findSourceFileName(Object chunkId) {
+            DocumentChunk documentChunk = chunkMap.get(String.valueOf(chunkId));
+            if (documentChunk == null || documentChunk.getDocumentId() == null) {
+                return null;
+            }
+            tech.easyflow.ai.entity.Document sourceDocument = documentMap.get(String.valueOf(documentChunk.getDocumentId()));
+            if (sourceDocument == null || StringUtil.noText(sourceDocument.getTitle())) {
+                return null;
+            }
+            return sourceDocument.getTitle();
+        }
     }
 
     private String buildFaqPromptContent(FaqItem faqItem, List<Map<String, String>> images) {

easyflow-modules/easyflow-module-ai/src/test/java/tech/easyflow/ai/document/service/impl/DocumentParseBridgeServiceImplTest.java

@@ -6,6 +6,9 @@ import com.easyagents.document.core.entity.ParseResponse;
 import com.easyagents.document.core.entity.ParseResult;
 import com.easyagents.document.core.entity.ParseTaskInfo;
 import com.easyagents.document.core.entity.ParseTaskStatus;
+import com.easyagents.document.pdf.PdfDocumentParseService;
+import com.easyagents.document.pptx.PptxDocumentParseService;
+import com.easyagents.document.xlsx.XlsxDocumentParseService;
 import org.junit.Assert;
 import org.junit.Test;
 import tech.easyflow.ai.document.exception.DocumentParseBridgeException;
@@ -37,8 +40,8 @@ public class DocumentParseBridgeServiceImplTest {
      */
     @Test
     public void shouldParseSuccessfully() {
-        FakeDocumentParseService parseService = new FakeDocumentParseService();
+        FakePdfDocumentParseService parseService = new FakePdfDocumentParseService();
-        DocumentParseBridgeServiceImpl bridgeService = buildBridgeService(parseService);
+        DocumentParseBridgeServiceImpl bridgeService = buildBridgeService(parseService, null, null, parseService);
 
         DocumentParsedResult document = bridgeService.parse(buildSource(), DocumentParseScenario.WORKFLOW_TEXT);
 
@@ -52,8 +55,8 @@ public class DocumentParseBridgeServiceImplTest {
      */
     @Test
     public void shouldSupportAsyncFlow() {
-        FakeDocumentParseService parseService = new FakeDocumentParseService();
+        FakePdfDocumentParseService parseService = new FakePdfDocumentParseService();
-        DocumentParseBridgeServiceImpl bridgeService = buildBridgeService(parseService);
+        DocumentParseBridgeServiceImpl bridgeService = buildBridgeService(parseService, null, null, parseService);
 
         DocumentParseTaskStatus taskStatus = bridgeService.submit(buildSource(), DocumentParseScenario.KNOWLEDGE_IMPORT);
         DocumentParseTaskStatus queriedStatus = bridgeService.queryTask("task-1");
@@ -69,9 +72,9 @@ public class DocumentParseBridgeServiceImplTest {
      */
     @Test
     public void shouldQueryTaskInfoSuccessfully() {
-        FakeDocumentParseService parseService = new FakeDocumentParseService();
+        FakePdfDocumentParseService parseService = new FakePdfDocumentParseService();
         parseService.taskStatusValue = "completed";
-        DocumentParseBridgeServiceImpl bridgeService = buildBridgeService(parseService);
+        DocumentParseBridgeServiceImpl bridgeService = buildBridgeService(parseService, null, null, parseService);
 
         DocumentParseTaskInfo taskInfo = bridgeService.queryTaskInfo("task-1");
 
@@ -85,7 +88,7 @@ public class DocumentParseBridgeServiceImplTest {
      */
     @Test
     public void shouldThrowWhenServiceDisabled() {
-        DocumentParseBridgeServiceImpl bridgeService = buildBridgeService(null);
+        DocumentParseBridgeServiceImpl bridgeService = buildBridgeService(null, null, null, null);
 
         try {
             bridgeService.parse(buildSource(), DocumentParseScenario.WORKFLOW_TEXT);
@@ -95,9 +98,29 @@ public class DocumentParseBridgeServiceImplTest {
         }
     }
 
-    private DocumentParseBridgeServiceImpl buildBridgeService(DocumentParseService parseService) {
+    @Test
+    public void shouldRoutePptxToDedicatedService() {
+        FakePptxDocumentParseService pptxService = new FakePptxDocumentParseService();
+        FakePdfDocumentParseService defaultService = new FakePdfDocumentParseService();
+        DocumentParseBridgeServiceImpl bridgeService = buildBridgeService(null, pptxService, null, defaultService);
+
+        DocumentParsedResult result = bridgeService.parse(buildSource("slides.pptx",
+            "application/vnd.openxmlformats-officedocument.presentationml.presentation"), DocumentParseScenario.KNOWLEDGE_IMPORT);
+
+        Assert.assertEquals("# pptx", result.getPreferredText());
+        Assert.assertEquals(1, pptxService.parseCallCount);
+        Assert.assertEquals(0, defaultService.parseCallCount);
+    }
+
+    private DocumentParseBridgeServiceImpl buildBridgeService(PdfDocumentParseService pdfDocumentParseService,
+                                                              PptxDocumentParseService pptxDocumentParseService,
+                                                              XlsxDocumentParseService xlsxDocumentParseService,
+                                                              DocumentParseService parseService) {
         return new DocumentParseBridgeServiceImpl(
             parseService,
+            pdfDocumentParseService,
+            pptxDocumentParseService,
+            xlsxDocumentParseService,
             new DocumentSourceLoader(new InMemoryFileStorageService()),
             new DocumentParseRequestFactory(),
             new DocumentParseResultMapper()
@@ -105,8 +128,12 @@ public class DocumentParseBridgeServiceImplTest {
     }
 
     private DocumentSourceRef buildSource() {
-        DocumentSourceRef sourceRef = DocumentSourceRef.ofBytes("demo.pdf", "pdf-data".getBytes(StandardCharsets.UTF_8));
+        return buildSource("demo.pdf", "application/pdf");
-        sourceRef.setContentType("application/pdf");
+    }
+
+    private DocumentSourceRef buildSource(String fileName, String contentType) {
+        DocumentSourceRef sourceRef = DocumentSourceRef.ofBytes(fileName, "pdf-data".getBytes(StandardCharsets.UTF_8));
+        sourceRef.setContentType(contentType);
         sourceRef.setSize(8L);
         return sourceRef;
     }
@@ -133,13 +160,15 @@ public class DocumentParseBridgeServiceImplTest {
         }
     }
 
-    private static class FakeDocumentParseService implements DocumentParseService {
+    private static class FakePdfDocumentParseService implements PdfDocumentParseService {
 
         private ParseRequest lastParseRequest;
         private String taskStatusValue = "running";
+        private int parseCallCount;
 
         @Override
         public ParseResponse parse(ParseRequest request) {
+            parseCallCount++;
             this.lastParseRequest = request;
             return buildResponse();
         }
@@ -187,4 +216,36 @@ public class DocumentParseBridgeServiceImplTest {
             return response;
         }
     }
+
+    private static class FakePptxDocumentParseService implements PptxDocumentParseService {
+
+        private int parseCallCount;
+
+        @Override
+        public ParseResponse parse(ParseRequest request) {
+            parseCallCount++;
+            ParseResult result = new ParseResult();
+            result.setFileName("slides.pptx");
+            result.setMarkdown("# pptx");
+            result.setPlainText("pptx");
+            ParseResponse response = new ParseResponse();
+            response.setResults(Collections.singletonList(result));
+            return response;
+        }
+
+        @Override
+        public ParseTaskStatus submit(ParseRequest request) {
+            throw new UnsupportedOperationException();
+        }
+
+        @Override
+        public ParseTaskStatus queryTask(String taskId) {
+            throw new UnsupportedOperationException();
+        }
+
+        @Override
+        public ParseResponse queryResult(String taskId) {
+            throw new UnsupportedOperationException();
+        }
+    }
 }

easyflow-modules/easyflow-module-ai/src/test/java/tech/easyflow/ai/document/support/DocumentParseRequestFactoryTest.java

@@ -1,6 +1,9 @@
 package tech.easyflow.ai.document.support;
 
 import com.easyagents.document.core.entity.ParseRequest;
+import com.easyagents.document.core.entity.PdfParseRequest;
+import com.easyagents.document.core.entity.PptxParseRequest;
+import com.easyagents.document.core.entity.XlsxParseRequest;
 import org.junit.Assert;
 import org.junit.Test;
 import tech.easyflow.ai.document.model.DocumentParseScenario;
@@ -26,6 +29,7 @@ public class DocumentParseRequestFactoryTest {
         Assert.assertFalse(request.getReturnMiddleJson());
         Assert.assertFalse(request.getReturnContentList());
         Assert.assertFalse(request.getReturnImages());
+        Assert.assertTrue(request instanceof PdfParseRequest);
     }
 
     /**
@@ -41,12 +45,33 @@ public class DocumentParseRequestFactoryTest {
         Assert.assertTrue(request.getReturnMiddleJson());
         Assert.assertTrue(request.getReturnContentList());
         Assert.assertTrue(request.getReturnImages());
+        Assert.assertTrue(request instanceof PdfParseRequest);
+    }
+
+    /**
+     * 验证 PPTX / XLSX 会构建对应的强类型请求。
+     */
+    @Test
+    public void shouldBuildOfficeTypedRequests() {
+        DocumentParseRequestFactory factory = new DocumentParseRequestFactory();
+
+        ParseRequest pptxRequest = factory.build(buildSource("slides.pptx",
+            "application/vnd.openxmlformats-officedocument.presentationml.presentation"), DocumentParseScenario.KNOWLEDGE_IMPORT);
+        ParseRequest xlsxRequest = factory.build(buildSource("table.xlsx",
+            "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"), DocumentParseScenario.KNOWLEDGE_IMPORT);
+
+        Assert.assertTrue(pptxRequest instanceof PptxParseRequest);
+        Assert.assertTrue(xlsxRequest instanceof XlsxParseRequest);
     }
 
     private LoadedDocumentSource buildSource() {
+        return buildSource("demo.pdf", "application/pdf");
+    }
+
+    private LoadedDocumentSource buildSource(String fileName, String contentType) {
         LoadedDocumentSource source = new LoadedDocumentSource();
-        source.setFileName("demo.pdf");
+        source.setFileName(fileName);
-        source.setContentType("application/pdf");
+        source.setContentType(contentType);
         source.setContentBytes("pdf-data".getBytes());
         source.setSize(8L);
         return source;

easyflow-modules/easyflow-module-ai/src/test/java/tech/easyflow/ai/document/support/DocumentParseResultMapperTest.java

@@ -4,6 +4,7 @@ import com.easyagents.document.core.entity.ParseArtifacts;
 import com.easyagents.document.core.entity.ParseResult;
 import com.easyagents.document.core.entity.ParseResponse;
 import com.easyagents.document.core.entity.ParseTaskInfo;
+import com.easyagents.document.core.entity.ParseTaskStatus;
 import org.junit.Assert;
 import org.junit.Test;
 import tech.easyflow.ai.document.model.DocumentParseTaskInfo;
@@ -65,6 +66,8 @@ public class DocumentParseResultMapperTest {
         ParseTaskInfo taskInfo = new ParseTaskInfo();
         taskInfo.setTaskId("task-1");
         taskInfo.setStatus("completed");
+        taskInfo.setProgressPercent(100);
+        taskInfo.setCurrentStage("completed");
 
         ParseResult result = new ParseResult();
         result.setFileName("demo.pdf");
@@ -76,7 +79,33 @@ public class DocumentParseResultMapperTest {
         DocumentParseTaskInfo mapped = mapper.map(taskInfo);
 
         Assert.assertEquals("task-1", mapped.getTaskId());
+        Assert.assertEquals(Integer.valueOf(100), mapped.getProgressPercent());
+        Assert.assertEquals("completed", mapped.getCurrentStage());
         Assert.assertNotNull(mapped.getResult());
         Assert.assertEquals("# title", mapped.getResult().getPreferredText());
     }
+
+    /**
+     * 验证异步进度字段被完整透传。
+     */
+    @Test
+    public void shouldMapTaskStatusProgressFields() {
+        DocumentParseResultMapper mapper = new DocumentParseResultMapper();
+        ParseTaskStatus status = new ParseTaskStatus();
+        status.setTaskId("task-2");
+        status.setStatus("running");
+        status.setProgressPercent(45);
+        status.setCurrentStage("ocr");
+        status.setProcessedItems(9);
+        status.setTotalItems(20);
+        status.setStatusMessage("正在识别图片");
+
+        tech.easyflow.ai.document.model.DocumentParseTaskStatus mapped = mapper.map(status);
+
+        Assert.assertEquals(Integer.valueOf(45), mapped.getProgressPercent());
+        Assert.assertEquals("ocr", mapped.getCurrentStage());
+        Assert.assertEquals(Integer.valueOf(9), mapped.getProcessedItems());
+        Assert.assertEquals(Integer.valueOf(20), mapped.getTotalItems());
+        Assert.assertEquals("正在识别图片", mapped.getStatusMessage());
+    }
 }

easyflow-modules/easyflow-module-ai/src/test/java/tech/easyflow/ai/documentimport/task/KnowledgeDocumentImportTaskAppServiceTest.java

@@ -1,17 +1,33 @@
 package tech.easyflow.ai.documentimport.task;
 
+import com.easyagents.document.core.entity.DocumentBlock;
+import com.easyagents.document.core.entity.DocumentImage;
+import com.easyagents.document.core.entity.DocumentTable;
+import com.easyagents.rag.ingestion.model.StrategyConfig;
 import org.junit.Assert;
 import org.junit.Test;
+import org.springframework.web.multipart.MultipartFile;
+import tech.easyflow.ai.document.model.DocumentParseArtifacts;
+import tech.easyflow.ai.document.model.DocumentParsedResult;
+import tech.easyflow.ai.documentimport.DocumentImportKeys;
+import tech.easyflow.ai.entity.DocumentChunk;
 import tech.easyflow.ai.entity.DocumentImportTask;
 import tech.easyflow.ai.enums.DocumentImportTaskStatus;
 import tech.easyflow.ai.enums.DocumentProcessStatus;
 import tech.easyflow.ai.mapper.DocumentMapper;
 import tech.easyflow.ai.service.DocumentImportTaskService;
+import tech.easyflow.common.filestorage.FileStorageService;
 
 import java.lang.reflect.Field;
 import java.lang.reflect.Method;
 import java.lang.reflect.Proxy;
 import java.math.BigInteger;
+import java.nio.charset.StandardCharsets;
+import java.util.ArrayList;
+import java.util.Base64;
+import java.util.LinkedHashMap;
+import java.util.List;
+import java.util.Map;
 import java.util.concurrent.atomic.AtomicReference;
 
 /**
@@ -84,6 +100,258 @@ public class KnowledgeDocumentImportTaskAppServiceTest {
         Assert.assertEquals("新错误", updatedTask.getErrorSummary());
     }
 
+    /**
+     * 验证知识库导入会把解析图片上传到对象存储，并同步改写 Markdown 与结构化引用。
+     *
+     * @throws Exception 反射调用异常
+     */
+    @Test
+    public void normalizeParsedImagesForKnowledgeImportShouldUploadAndRewriteReferences() throws Exception {
+        KnowledgeDocumentImportTaskAppService service = new KnowledgeDocumentImportTaskAppService();
+        AtomicReference<String> savedPrePathRef = new AtomicReference<String>();
+        AtomicReference<String> savedFilenameRef = new AtomicReference<String>();
+        setField(service, "storageService", mockFileStorageService(savedPrePathRef, savedFilenameRef));
+
+        tech.easyflow.ai.entity.Document document = new tech.easyflow.ai.entity.Document();
+        document.setId(BigInteger.valueOf(88));
+        document.setTitle("产品说明书（终版）.pdf");
+
+        DocumentParsedResult parsedResult = new DocumentParsedResult();
+        parsedResult.setMarkdown("图例如下：\n![](images/sample-image.png)");
+        parsedResult.setPreferredText(parsedResult.getMarkdown());
+        parsedResult.setPlainText(parsedResult.getMarkdown());
+
+        DocumentImage image = new DocumentImage();
+        image.setName("sample-image.png");
+        image.setSourcePath("images/sample-image.png");
+        image.setMimeType("image/png");
+        image.setDataUrl("data:image/png;base64," + Base64.getEncoder().encodeToString("demo".getBytes(StandardCharsets.UTF_8)));
+        parsedResult.setImages(new ArrayList<DocumentImage>(List.of(image)));
+
+        DocumentBlock block = new DocumentBlock();
+        block.setImagePath("images/sample-image.png");
+        parsedResult.setBlocks(new ArrayList<DocumentBlock>(List.of(block)));
+
+        DocumentTable table = new DocumentTable();
+        table.setImagePath("images/sample-image.png");
+        parsedResult.setTables(new ArrayList<DocumentTable>(List.of(table)));
+
+        DocumentParseArtifacts artifacts = new DocumentParseArtifacts();
+        List<Map<String, Object>> contentList = new ArrayList<Map<String, Object>>();
+        Map<String, Object> contentItem = new LinkedHashMap<String, Object>();
+        contentItem.put("img_path", "images/sample-image.png");
+        contentList.add(contentItem);
+        artifacts.setContentList(contentList);
+        Map<String, Object> xlsxArtifact = new LinkedHashMap<String, Object>();
+        List<Map<String, Object>> sheetImages = new ArrayList<Map<String, Object>>();
+        sheetImages.add(new LinkedHashMap<String, Object>() {{
+            put("sheetName", "Sheet1");
+            put("sourcePaths", new ArrayList<String>(List.of("images/sample-image.png")));
+        }});
+        xlsxArtifact.put("sheetImages", sheetImages);
+        artifacts.setExtraJsonArtifacts(new LinkedHashMap<String, Object>() {{
+            put("xlsx", xlsxArtifact);
+        }});
+        parsedResult.setArtifacts(artifacts);
+
+        Method method = KnowledgeDocumentImportTaskAppService.class.getDeclaredMethod(
+            "normalizeParsedImagesForKnowledgeImport",
+            tech.easyflow.ai.entity.Document.class,
+            DocumentParsedResult.class
+        );
+        method.setAccessible(true);
+        DocumentParsedResult normalized = (DocumentParsedResult) method.invoke(service, document, parsedResult);
+
+        Assert.assertNotNull(normalized);
+        Assert.assertEquals("knowledge-parse/88_产品说明书_终版/images", savedPrePathRef.get());
+        Assert.assertEquals("sample-image.png", savedFilenameRef.get());
+
+        String expectedUrl = "http://localhost:39000/easyflow/attachment/knowledge-parse/88_产品说明书_终版/images/sample-image.png";
+        Assert.assertTrue(normalized.getMarkdown().contains(expectedUrl));
+        Assert.assertEquals(expectedUrl, normalized.getBlocks().get(0).getImagePath());
+        Assert.assertEquals(expectedUrl, normalized.getTables().get(0).getImagePath());
+        Assert.assertEquals(expectedUrl, normalized.getImages().get(0).getSourcePath());
+        Assert.assertNull(normalized.getImages().get(0).getDataUrl());
+        Object rewrittenContentList = normalized.getArtifacts().getContentList();
+        Assert.assertTrue(rewrittenContentList instanceof List<?>);
+        Assert.assertEquals(expectedUrl, ((Map<?, ?>) ((List<?>) rewrittenContentList).get(0)).get("img_path"));
+        Object rewrittenSheetImages = ((Map<?, ?>) normalized.getArtifacts().getExtraJsonArtifacts().get("xlsx")).get("sheetImages");
+        Assert.assertTrue(rewrittenSheetImages instanceof List<?>);
+        Object sourcePaths = ((Map<?, ?>) ((List<?>) rewrittenSheetImages).get(0)).get("sourcePaths");
+        Assert.assertEquals(expectedUrl, ((List<?>) sourcePaths).get(0));
+    }
+
+    /**
+     * 验证 PPTX 会基于页级工件生成稳定的知识库分块。
+     *
+     * @throws Exception 反射调用异常
+     */
+    @Test
+    public void buildOfficeDocumentChunksShouldSplitPptxBySlide() throws Exception {
+        KnowledgeDocumentImportTaskAppService service = new KnowledgeDocumentImportTaskAppService();
+        tech.easyflow.ai.entity.Document document = new tech.easyflow.ai.entity.Document();
+        document.setId(BigInteger.valueOf(101));
+        document.setCollectionId(BigInteger.valueOf(201));
+        document.setTitle("季度汇报.pptx");
+
+        Map<String, Object> parseArtifactSummary = new LinkedHashMap<String, Object>();
+        List<Map<String, Object>> slides = new ArrayList<Map<String, Object>>();
+        slides.add(new LinkedHashMap<String, Object>() {{
+            put("slideIndex", 0);
+            put("title", "封面");
+            put("ocrMarkdown", "本页介绍季度目标。");
+            put("imagePath", "https://example.com/slides/slide-001.png");
+            put("imageName", "slide-001-page");
+        }});
+        slides.add(new LinkedHashMap<String, Object>() {{
+            put("slideIndex", 1);
+            put("title", "经营分析");
+            put("ocrMarkdown", "收入同比增长 18%。");
+            put("imagePath", "https://example.com/slides/slide-002.png");
+            put("imageName", "slide-002-page");
+        }});
+        parseArtifactSummary.put("slides", slides);
+
+        Method method = KnowledgeDocumentImportTaskAppService.class.getDeclaredMethod(
+            "buildOfficeDocumentChunks",
+            tech.easyflow.ai.entity.Document.class,
+            String.class,
+            StrategyConfig.class,
+            Map.class
+        );
+        method.setAccessible(true);
+
+        @SuppressWarnings("unchecked")
+        List<DocumentChunk> chunks = (List<DocumentChunk>) method.invoke(
+            service,
+            document,
+            "pptx",
+            null,
+            parseArtifactSummary
+        );
+
+        Assert.assertEquals(2, chunks.size());
+        DocumentChunk firstChunk = chunks.get(0);
+        Assert.assertTrue(firstChunk.getContent().contains("Slide 1"));
+        Assert.assertTrue(firstChunk.getContent().contains("本页介绍季度目标"));
+        Assert.assertEquals("https://example.com/slides/slide-001.png",
+            ((List<?>) firstChunk.getOptions().get(DocumentImportKeys.KEY_DOCUMENT_IMAGE_REFS)).get(0));
+        Assert.assertEquals(1, firstChunk.getOptions().get(DocumentImportKeys.KEY_DOCUMENT_PAGE_INDEX));
+        Assert.assertTrue(String.valueOf(firstChunk.getOptions().get(DocumentImportKeys.KEY_DOCUMENT_RENDER_MARKDOWN))
+            .contains("slide-001.png"));
+    }
+
+    /**
+     * 验证 XLSX 纯图片 Sheet 不会退化为空内容，并会输出稳定图片引用。
+     *
+     * @throws Exception 反射调用异常
+     */
+    @Test
+    public void buildOfficeDocumentChunksShouldKeepImageOnlyXlsxSheetReferences() throws Exception {
+        KnowledgeDocumentImportTaskAppService service = new KnowledgeDocumentImportTaskAppService();
+        tech.easyflow.ai.entity.Document document = new tech.easyflow.ai.entity.Document();
+        document.setId(BigInteger.valueOf(102));
+        document.setCollectionId(BigInteger.valueOf(202));
+        document.setTitle("巡检记录.xlsx");
+
+        Map<String, Object> parseArtifactSummary = new LinkedHashMap<String, Object>();
+        List<Map<String, Object>> sheets = new ArrayList<Map<String, Object>>();
+        sheets.add(new LinkedHashMap<String, Object>() {{
+            put("sheetName", "图片页");
+            put("sheetIndex", 0);
+            put("rows", new ArrayList<Map<String, Object>>());
+        }});
+        parseArtifactSummary.put("sheets", sheets);
+
+        List<Map<String, Object>> cellImages = new ArrayList<Map<String, Object>>();
+        cellImages.add(new LinkedHashMap<String, Object>() {{
+            put("sheetName", "图片页");
+            put("referenceKey", "image-sheet-r2c2-001");
+            put("sourcePath", "https://example.com/xlsx/sheet/image-001.jpeg");
+            put("anchorCell", "B2");
+            put("ocrText", "设备状态正常");
+            put("fromRow", 1);
+        }});
+        parseArtifactSummary.put("cellImages", cellImages);
+
+        StrategyConfig strategyConfig = new StrategyConfig();
+        strategyConfig.setRowsPerChunk(10);
+
+        Method method = KnowledgeDocumentImportTaskAppService.class.getDeclaredMethod(
+            "buildOfficeDocumentChunks",
+            tech.easyflow.ai.entity.Document.class,
+            String.class,
+            StrategyConfig.class,
+            Map.class
+        );
+        method.setAccessible(true);
+
+        @SuppressWarnings("unchecked")
+        List<DocumentChunk> chunks = (List<DocumentChunk>) method.invoke(
+            service,
+            document,
+            "xlsx",
+            strategyConfig,
+            parseArtifactSummary
+        );
+
+        Assert.assertEquals(1, chunks.size());
+        DocumentChunk onlyChunk = chunks.get(0);
+        Assert.assertTrue(onlyChunk.getContent().contains("图片 OCR"));
+        Assert.assertTrue(onlyChunk.getContent().contains("设备状态正常"));
+        Assert.assertEquals("图片页", onlyChunk.getOptions().get(DocumentImportKeys.KEY_DOCUMENT_SHEET_NAME));
+        Assert.assertEquals("https://example.com/xlsx/sheet/image-001.jpeg",
+            ((List<?>) onlyChunk.getOptions().get(DocumentImportKeys.KEY_DOCUMENT_IMAGE_REFS)).get(0));
+        String renderMarkdown = String.valueOf(onlyChunk.getOptions().get(DocumentImportKeys.KEY_DOCUMENT_RENDER_MARKDOWN));
+        Assert.assertTrue(renderMarkdown.contains("[IMG:image-sheet-r2c2-001]"));
+        Assert.assertTrue(renderMarkdown.contains("![image-sheet-r2c2-001](https://example.com/xlsx/sheet/image-001.jpeg)"));
+    }
+
+    /**
+     * 验证空白 Sheet 不会被误判成纯图片分块。
+     *
+     * @throws Exception 反射调用异常
+     */
+    @Test
+    public void buildOfficeDocumentChunksShouldSkipBlankXlsxSheetWithoutImages() throws Exception {
+        KnowledgeDocumentImportTaskAppService service = new KnowledgeDocumentImportTaskAppService();
+        tech.easyflow.ai.entity.Document document = new tech.easyflow.ai.entity.Document();
+        document.setId(BigInteger.valueOf(103));
+        document.setCollectionId(BigInteger.valueOf(203));
+        document.setTitle("空白工作簿.xlsx");
+
+        Map<String, Object> parseArtifactSummary = new LinkedHashMap<String, Object>();
+        parseArtifactSummary.put("sheets", new ArrayList<Map<String, Object>>(List.of(new LinkedHashMap<String, Object>() {{
+            put("sheetName", "空白页");
+            put("sheetIndex", 0);
+            put("rows", new ArrayList<Map<String, Object>>());
+        }})));
+        parseArtifactSummary.put("cellImages", new ArrayList<Map<String, Object>>());
+
+        StrategyConfig strategyConfig = new StrategyConfig();
+        strategyConfig.setRowsPerChunk(10);
+
+        Method method = KnowledgeDocumentImportTaskAppService.class.getDeclaredMethod(
+            "buildOfficeDocumentChunks",
+            tech.easyflow.ai.entity.Document.class,
+            String.class,
+            StrategyConfig.class,
+            Map.class
+        );
+        method.setAccessible(true);
+
+        @SuppressWarnings("unchecked")
+        List<DocumentChunk> chunks = (List<DocumentChunk>) method.invoke(
+            service,
+            document,
+            "xlsx",
+            strategyConfig,
+            parseArtifactSummary
+        );
+
+        Assert.assertTrue(chunks.isEmpty());
+    }
+
     private static DocumentMapper mockDocumentMapper(tech.easyflow.ai.entity.Document persistedDocument,
                                                      AtomicReference<tech.easyflow.ai.entity.Document> updatedDocumentRef) {
         return (DocumentMapper) Proxy.newProxyInstance(
@@ -116,6 +384,22 @@ public class KnowledgeDocumentImportTaskAppServiceTest {
         );
     }
 
+    private static FileStorageService mockFileStorageService(AtomicReference<String> savedPrePathRef,
+                                                             AtomicReference<String> savedFilenameRef) {
+        return (FileStorageService) Proxy.newProxyInstance(
+            FileStorageService.class.getClassLoader(),
+            new Class<?>[]{FileStorageService.class},
+            (proxy, method, args) -> {
+                if ("save".equals(method.getName()) && args != null && args.length == 2 && args[0] instanceof MultipartFile file) {
+                    savedPrePathRef.set((String) args[1]);
+                    savedFilenameRef.set(file.getOriginalFilename());
+                    return "http://localhost:39000/easyflow/attachment/" + args[1] + "/" + file.getOriginalFilename();
+                }
+                return defaultValue(method.getReturnType());
+            }
+        );
+    }
+
     private static void setField(Object target, String fieldName, Object value) throws Exception {
         Field field = KnowledgeDocumentImportTaskAppService.class.getDeclaredField(fieldName);
         field.setAccessible(true);

easyflow-starter/easyflow-starter-all/src/main/resources/application.yml

@@ -166,7 +166,7 @@ dromara:
 # easy-agents 文档解析统一配置
 easy-agents:
   document:
-    pdf:
+    ocr:
       provider: mineru
       mineru:
         # 统一文档解析桥接层直接复用 easy-agents 的 provider 配置，不在 easyflow 再复制一套配置体系

easyflow-ui-admin/app/src/views/ai/documentCollection/ChunkDocumentTable.vue

@@ -1,10 +1,11 @@
 <script setup lang="ts">
 import { ref } from 'vue';
+import ElXMarkdown from 'vue-element-plus-x/es/XMarkdown/index.js';
 
 import { EasyFlowFormModal } from '@easyflow/common-ui';
 import { $t } from '@easyflow/locales';
 
-import { Delete, MoreFilled } from '@element-plus/icons-vue';
+import { Delete, EditPen, MoreFilled } from '@element-plus/icons-vue';
 import {
   ElButton,
   ElDropdown,
@@ -12,6 +13,7 @@ import {
   ElDropdownMenu,
   ElForm,
   ElFormItem,
+  ElIcon,
   ElInput,
   ElMessage,
   ElMessageBox,
@@ -21,6 +23,10 @@ import {
 
 import { api } from '#/api/request';
 import PageData from '#/components/page/PageData.vue';
+import {
+  markdownRenderProps,
+  resolveMarkdownContent,
+} from '#/views/ai/documentCollection/markdown-content';
 import { buildKnowledgePath } from '#/views/ai/documentCollection/share-path';
 
 const props = defineProps({
@@ -122,6 +128,85 @@ const form = ref({
   id: '',
   content: '',
 });
+
+const getChunkOptions = (row: any) => row?.options || {};
+
+const getMarkdown = (row: any) =>
+  resolveMarkdownContent(getChunkOptions(row)?.renderMarkdown || row?.content);
+
+const isExcelChunk = (row: any) => {
+  const options = getChunkOptions(row);
+  const sourceFileExt = String(
+    options?.sourceFileExt || options?.['splitter.sourceFileExt'] || '',
+  ).toLowerCase();
+
+  return Boolean(
+    sourceFileExt === 'xlsx' ||
+      options?.sheetName ||
+      options?.rowStart ||
+      options?.rowEnd,
+  );
+};
+
+const shouldUseExcelChunkCards = (pageList: any[] = []) =>
+  pageList.length > 0 && pageList.every((row) => isExcelChunk(row));
+
+const getSheetName = (row: any) =>
+  String(getChunkOptions(row)?.sheetName || '');
+
+const getRowStart = (row: any) => {
+  const rowStart = Number(getChunkOptions(row)?.rowStart || 0);
+  return Math.max(rowStart, 0);
+};
+
+const getRowEnd = (row: any) => {
+  const rowEnd = Number(getChunkOptions(row)?.rowEnd || 0);
+  return Math.max(rowEnd, 0);
+};
+
+const getRowRangeLabel = (row: any) => {
+  const rowStart = getRowStart(row);
+  const rowEnd = getRowEnd(row);
+  if (rowStart > 0 && rowEnd > 0) {
+    return rowStart === rowEnd
+      ? `第 ${rowStart} 行`
+      : `第 ${rowStart}-${rowEnd} 行`;
+  }
+  if (rowStart > 0) {
+    return `第 ${rowStart} 行起`;
+  }
+  return '';
+};
+
+const getChunkTitle = (row: any) => {
+  const options = getChunkOptions(row);
+  if (options?.sourceLabel) {
+    return String(options.sourceLabel);
+  }
+  const sheetName = getSheetName(row);
+  const rowRangeLabel = getRowRangeLabel(row);
+  if (sheetName && rowRangeLabel) {
+    return `${sheetName} · ${rowRangeLabel}`;
+  }
+  return sheetName || row?.id || '-';
+};
+
+const getChunkIndexLabel = (row: any) => {
+  const sorting = Number(row?.sorting || 0);
+  if (sorting <= 0) {
+    return '';
+  }
+  return String(sorting).padStart(2, '0');
+};
+
+const getChunkHeaderLabel = (row: any) => {
+  const chunkIndexLabel = getChunkIndexLabel(row);
+  const chunkTitle = getChunkTitle(row);
+  if (chunkIndexLabel) {
+    return `分块 ${chunkIndexLabel} · ${chunkTitle}`;
+  }
+  return chunkTitle;
+};
 </script>
 
 <template>
@@ -136,12 +221,76 @@ const form = ref({
       :extra-query-params="queryParams"
     >
       <template #default="{ pageList }">
-        <ElTable :data="pageList" style="width: 100%" size="large">
+        <div v-if="shouldUseExcelChunkCards(pageList)" class="chunk-board">
+          <article v-for="row in pageList" :key="row.id" class="chunk-card">
+            <div v-if="props.manageable" class="chunk-card__toolbar">
+              <ElButton
+                circle
+                text
+                type="primary"
+                class="chunk-card__action"
+                @click="handleEdit(row)"
+              >
+                <ElIcon><EditPen /></ElIcon>
+              </ElButton>
+              <ElDropdown>
+                <ElButton
+                  circle
+                  text
+                  class="chunk-card__action chunk-card__action--ghost"
+                >
+                  <ElIcon><MoreFilled /></ElIcon>
+                </ElButton>
+
+                <template #dropdown>
+                  <ElDropdownMenu>
+                    <ElDropdownItem @click="handleDelete(row)">
+                      <ElButton link type="danger" :icon="Delete">
+                        {{ $t('button.delete') }}
+                      </ElButton>
+                    </ElDropdownItem>
+                  </ElDropdownMenu>
+                </template>
+              </ElDropdown>
+            </div>
+
+            <div class="chunk-card__header">
+              <div class="chunk-card__eyebrow">
+                <span class="chunk-card__eyebrow-dot"></span>
+                <span>{{ getChunkHeaderLabel(row) }}</span>
+              </div>
+            </div>
+
+            <div v-if="getMarkdown(row)" class="chunk-card__content">
+              <div class="chunk-rich-content chunk-rich-content--card">
+                <ElXMarkdown
+                  :markdown="getMarkdown(row)"
+                  :allow-html="markdownRenderProps.allowHtml"
+                  :sanitize="markdownRenderProps.sanitize"
+                />
+              </div>
+            </div>
+            <span v-else class="chunk-table__empty">-</span>
+          </article>
+        </div>
+
+        <ElTable v-else :data="pageList" style="width: 100%" size="large">
           <ElTableColumn
             prop="content"
             :label="$t('documentCollection.content')"
             min-width="240"
-          />
+          >
+            <template #default="{ row }">
+              <div v-if="getMarkdown(row)" class="chunk-rich-content">
+                <ElXMarkdown
+                  :markdown="getMarkdown(row)"
+                  :allow-html="markdownRenderProps.allowHtml"
+                  :sanitize="markdownRenderProps.sanitize"
+                />
+              </div>
+              <span v-else class="chunk-table__empty">-</span>
+            </template>
+          </ElTableColumn>
           <ElTableColumn
             v-if="props.manageable"
             :label="$t('common.handle')"
@@ -198,4 +347,251 @@ const form = ref({
   </div>
 </template>
 
-<style scoped></style>
+<style scoped>
+.chunk-board {
+  display: grid;
+  gap: 14px;
+}
+
+.chunk-card {
+  position: relative;
+  padding: 20px 20px 18px;
+  overflow: hidden;
+  background: linear-gradient(
+    135deg,
+    color-mix(in srgb, var(--el-color-primary-light-9) 80%, white) 0%,
+    var(--el-fill-color-blank) 38%
+  );
+  border: 1px solid color-mix(in srgb, var(--el-border-color-light) 78%, white);
+  border-radius: 18px;
+  box-shadow: 0 18px 40px rgb(15 23 42 / 6%);
+}
+
+.chunk-card::before {
+  position: absolute;
+  inset: 0 auto 0 0;
+  width: 4px;
+  content: '';
+  background: linear-gradient(
+    180deg,
+    var(--el-color-primary),
+    color-mix(in srgb, var(--el-color-primary) 44%, white)
+  );
+}
+
+.chunk-card__toolbar {
+  position: absolute;
+  top: 14px;
+  right: 14px;
+  z-index: 1;
+  display: flex;
+  gap: 6px;
+  align-items: center;
+}
+
+.chunk-card__action {
+  width: 34px;
+  height: 34px;
+  color: var(--el-color-primary);
+  background: color-mix(in srgb, var(--el-color-primary-light-9) 66%, white);
+  border: 1px solid
+    color-mix(in srgb, var(--el-color-primary-light-8) 72%, white);
+  box-shadow: 0 8px 18px rgb(37 99 235 / 10%);
+}
+
+.chunk-card__action--ghost {
+  color: var(--el-text-color-secondary);
+  background: rgb(255 255 255 / 86%);
+  border-color: color-mix(in srgb, var(--el-border-color-light) 86%, white);
+  box-shadow: none;
+}
+
+.chunk-card__header {
+  display: flex;
+  flex-direction: column;
+  min-width: 0;
+  padding-right: 92px;
+}
+
+.chunk-card__eyebrow {
+  display: inline-flex;
+  gap: 8px;
+  align-items: center;
+  min-height: 28px;
+  font-size: 12px;
+  font-weight: 600;
+  line-height: 1.5;
+  color: var(--el-text-color-secondary);
+}
+
+.chunk-card__eyebrow-dot {
+  width: 8px;
+  height: 8px;
+  border-radius: 999px;
+  background: var(--el-color-primary);
+  box-shadow: 0 0 0 4px
+    color-mix(in srgb, var(--el-color-primary-light-8) 50%, transparent);
+}
+
+.chunk-card__content {
+  padding-top: 14px;
+}
+
+.chunk-rich-content {
+  min-width: 0;
+  padding: 4px 0;
+  font-size: 14px;
+  line-height: 1.72;
+  color: var(--el-text-color-regular);
+}
+
+.chunk-rich-content--card {
+  padding: 14px 16px;
+  overflow-x: auto;
+  background: rgb(255 255 255 / 82%);
+  border: 1px solid rgb(15 23 42 / 6%);
+  border-radius: 16px;
+}
+
+.chunk-rich-content :deep(.markdown-body) {
+  font-size: inherit;
+  line-height: inherit;
+  color: inherit;
+  background: transparent;
+}
+
+.chunk-rich-content :deep(.markdown-body > :first-child) {
+  margin-top: 0;
+}
+
+.chunk-rich-content :deep(.markdown-body > :last-child) {
+  margin-bottom: 0;
+}
+
+.chunk-rich-content :deep(*) {
+  overflow-wrap: anywhere;
+}
+
+.chunk-rich-content :deep(p) {
+  margin: 0 0 10px;
+}
+
+.chunk-rich-content :deep(p:last-child) {
+  margin-bottom: 0;
+}
+
+.chunk-rich-content :deep(h1),
+.chunk-rich-content :deep(h2),
+.chunk-rich-content :deep(h3),
+.chunk-rich-content :deep(h4),
+.chunk-rich-content :deep(h5),
+.chunk-rich-content :deep(h6) {
+  margin: 14px 0 10px;
+  font-weight: 600;
+  line-height: 1.45;
+  color: var(--el-text-color-primary);
+}
+
+.chunk-rich-content :deep(h1:first-child),
+.chunk-rich-content :deep(h2:first-child),
+.chunk-rich-content :deep(h3:first-child),
+.chunk-rich-content :deep(h4:first-child),
+.chunk-rich-content :deep(h5:first-child),
+.chunk-rich-content :deep(h6:first-child) {
+  margin-top: 0;
+}
+
+.chunk-rich-content :deep(ul),
+.chunk-rich-content :deep(ol) {
+  padding-left: 20px;
+  margin: 0 0 12px;
+}
+
+.chunk-rich-content :deep(li + li) {
+  margin-top: 4px;
+}
+
+.chunk-rich-content :deep(a) {
+  color: var(--el-color-primary);
+  text-decoration: underline;
+  text-underline-offset: 2px;
+}
+
+.chunk-rich-content :deep(img) {
+  display: block;
+  max-width: min(100%, 560px);
+  height: auto;
+  margin: 12px 0;
+  border: 1px solid rgb(15 23 42 / 8%);
+  border-radius: 12px;
+  box-shadow: 0 10px 24px rgb(15 23 42 / 8%);
+}
+
+.chunk-rich-content :deep(table) {
+  width: 100%;
+  margin: 12px 0;
+  overflow: hidden;
+  border-collapse: collapse;
+  background: rgb(255 255 255 / 92%);
+  border: 1px solid rgb(15 23 42 / 8%);
+  border-radius: 12px;
+}
+
+.chunk-rich-content :deep(th),
+.chunk-rich-content :deep(td) {
+  padding: 10px 12px;
+  text-align: left;
+  vertical-align: top;
+  border: 1px solid rgb(15 23 42 / 8%);
+}
+
+.chunk-rich-content :deep(th) {
+  font-weight: 600;
+  color: var(--el-text-color-primary);
+  background: rgb(37 99 235 / 4%);
+}
+
+.chunk-rich-content :deep(pre) {
+  max-width: 100%;
+  padding: 12px 14px;
+  overflow: auto;
+  background: rgb(15 23 42 / 4%);
+  border: 1px solid rgb(15 23 42 / 6%);
+  border-radius: 12px;
+}
+
+.chunk-rich-content :deep(blockquote) {
+  padding-left: 12px;
+  margin: 12px 0;
+  color: var(--el-text-color-secondary);
+  border-left: 3px solid rgb(37 99 235 / 24%);
+}
+
+.chunk-table__empty {
+  color: var(--el-text-color-placeholder);
+}
+
+@media (max-width: 960px) {
+  .chunk-card {
+    padding: 18px 16px 16px;
+  }
+
+  .chunk-card__toolbar {
+    position: static;
+    justify-content: flex-end;
+    padding-bottom: 12px;
+  }
+
+  .chunk-card__header {
+    padding-right: 0;
+  }
+
+  .chunk-rich-content--card {
+    padding: 12px;
+  }
+
+  .chunk-card__eyebrow {
+    font-size: 12px;
+  }
+}
+</style>

easyflow-ui-admin/app/src/views/ai/documentCollection/DocumentTable.vue

@@ -40,6 +40,8 @@ interface DocumentStatusPayload {
   failedChunks?: number;
   knowledgeId?: number | string;
   lastTaskError?: string;
+  parseCurrentStage?: string;
+  parseStatusMessage?: string;
   processStatus?: string;
   progressPercent?: number;
   taskModifiedAt?: string;
@@ -154,11 +156,24 @@ const statusMetaMap: Record<
   },
 };
 
+const defaultStatusMeta: {
+  icon: Component;
+  toneClass: string;
+} = statusMetaMap.UPLOADED!;
+
 const getStatusLabel = (status?: string) =>
   $t(`documentCollection.taskStatus.${status || 'UPLOADED'}`);
 
-const getStatusMeta = (status?: string) =>
+const getStatusMeta = (
-  statusMetaMap[status || 'UPLOADED'] || statusMetaMap.UPLOADED;
+  status?: string,
+): {
+  icon: Component;
+  toneClass: string;
+} => statusMetaMap[status || 'UPLOADED'] ?? defaultStatusMeta;
+
+const getStatusToneClass = (status?: string) => getStatusMeta(status).toneClass;
+
+const getStatusIcon = (status?: string) => getStatusMeta(status).icon;
 
 const getChunkCount = (row: any) => {
   const totalChunks = Number(row.totalChunks || 0);
@@ -171,12 +186,28 @@ const getChunkCount = (row: any) => {
 const getProgressText = (row: any) => {
   const completed = Number(row.completedChunks || 0);
   const total = Number(row.totalChunks || 0);
+  if (row.processStatus === 'PARSING') {
+    return `${Number(row.progressPercent || 0)}%`;
+  }
   if (total <= 0) {
     return `${Number(row.progressPercent || 0)}%`;
   }
   return `${Number(row.progressPercent || 0)}% · ${completed}/${total}`;
 };
 
+const parseStageLabels: Record<string, string> = {
+  assembling: '汇总中',
+  extracting: '提取中',
+  ocr: 'OCR 中',
+  preparing: '准备中',
+  queued: '排队中',
+};
+
+const getProcessingHint = (row: any) =>
+  row.parseStatusMessage ||
+  parseStageLabels[row.parseCurrentStage || ''] ||
+  '';
+
 const clearReconnectTimer = () => {
   if (!reconnectTimer) {
     return;
@@ -211,6 +242,8 @@ const patchDocumentRow = (payload: DocumentStatusPayload) => {
     completedChunks: payload.completedChunks,
     failedChunks: payload.failedChunks,
     lastTaskError: payload.lastTaskError,
+    parseCurrentStage: payload.parseCurrentStage,
+    parseStatusMessage: payload.parseStatusMessage,
     processStatus: payload.processStatus,
     progressPercent: payload.progressPercent,
     taskModifiedAt: payload.taskModifiedAt,
@@ -529,7 +562,7 @@ watch(
             <div class="status-cell">
               <div
                 class="status-pill"
-                :class="getStatusMeta(row.processStatus).toneClass"
+                :class="getStatusToneClass(row.processStatus)"
               >
                 <span class="status-pill__icon-shell">
                   <ElIcon
@@ -540,7 +573,7 @@ watch(
                         : ''
                     "
                   >
-                    <component :is="getStatusMeta(row.processStatus).icon" />
+                    <component :is="getStatusIcon(row.processStatus)" />
                   </ElIcon>
                 </span>
                 <span class="status-pill__label">
@@ -548,7 +581,10 @@ watch(
                 </span>
               </div>
               <div
-                v-if="row.processStatus === 'INDEXING'"
+                v-if="
+                  row.processStatus === 'INDEXING' ||
+                  row.processStatus === 'PARSING'
+                "
                 class="status-progress"
               >
                 <ElProgress
@@ -558,6 +594,12 @@ watch(
                 <span class="status-progress__text">
                   {{ getProgressText(row) }}
                 </span>
+                <span
+                  v-if="row.processStatus === 'PARSING' && getProcessingHint(row)"
+                  class="status-progress__hint"
+                >
+                  {{ getProcessingHint(row) }}
+                </span>
               </div>
               <div
                 v-else-if="row.lastTaskError"
@@ -663,6 +705,12 @@ watch(
   text-align: left;
 }
 
+.status-progress__hint {
+  font-size: 12px;
+  color: var(--el-text-color-secondary);
+  text-align: left;
+}
+
 .status-error {
   max-width: 176px;
   font-size: 12px;

easyflow-ui-admin/app/src/views/ai/documentCollection/KnowledgeSearch.vue

@@ -21,6 +21,8 @@ type RetrievalMode = 'HYBRID' | 'KEYWORD' | 'VECTOR';
 interface SearchResultItem {
   sorting: number;
   content: string;
+  renderMarkdown?: string;
+  sourceFileName?: string;
   score?: number;
   hitSource?: 'BOTH' | 'KEYWORD' | 'VECTOR';
   vectorScore?: number;

easyflow-ui-admin/app/src/views/ai/documentCollection/PreviewSearchKnowledge.vue

@@ -1,17 +1,27 @@
 <script setup lang="ts">
+import type { PropType } from 'vue';
+
 import { ref } from 'vue';
+import ElXMarkdown from 'vue-element-plus-x/es/XMarkdown/index.js';
 
 import { $t } from '@easyflow/locales';
 
 import { Document } from '@element-plus/icons-vue';
 import { ElButton, ElEmpty, ElIcon, ElTag } from 'element-plus';
 
+import {
+  markdownRenderProps,
+  resolveMarkdownContent,
+} from '#/views/ai/documentCollection/markdown-content';
+
 type RetrievalMode = 'HYBRID' | 'KEYWORD' | 'VECTOR';
 type HitSource = 'BOTH' | 'KEYWORD' | 'VECTOR';
 
 interface PreviewItem {
   sorting: number | string;
   content: string;
+  renderMarkdown?: string;
+  sourceFileName?: string;
   score?: number | string;
   hitSource?: HitSource;
 }
@@ -42,12 +52,12 @@ const props = defineProps({
     default: false,
   },
   onCancel: {
-    type: Function,
+    type: Function as PropType<() => void>,
-    default: () => {},
+    default: () => undefined,
   },
   onConfirm: {
-    type: Function,
+    type: Function as PropType<() => void>,
-    default: () => {},
+    default: () => undefined,
   },
   isSearching: {
     type: Boolean,
@@ -100,18 +110,31 @@ const resolveHitSourceType = (hitSource?: HitSource) => {
   return 'info';
 };
 
-const normalizePreviewContent = (content?: string) => {
+const resolvePreviewMarkdown = (item: PreviewItem) =>
-  if (!content) {
+  resolveMarkdownContent(item.renderMarkdown || item.content);
-    return '';
+
+const resolveScoreLine = (item: PreviewItem) => {
+  const pieces: string[] = [];
+  if (
+    !props.hideScore &&
+    item.score !== undefined &&
+    item.score !== null &&
+    `${item.score}` !== ''
+  ) {
+    pieces.push(`${$t('documentCollection.similarityScore')}: ${item.score}`);
   }
-  if (typeof window !== 'undefined' && typeof DOMParser !== 'undefined') {
+  if (item.sourceFileName) {
-    const doc = new DOMParser().parseFromString(content, 'text/html');
+    pieces.push(`来源: ${item.sourceFileName}`);
-    return (doc.body.textContent || '').replaceAll(/\n\s*\n/g, '\n').trim();
   }
-  return content
+  return pieces.join(' · ');
-    .replaceAll(/<[^>]+>/g, ' ')
+};
-    .replaceAll(/\s+/g, ' ')
+
-    .trim();
+const handleCancel = () => {
+  props.onCancel?.();
+};
+
+const handleConfirm = () => {
+  props.onConfirm?.();
 };
 
 defineExpose({
@@ -149,8 +172,8 @@ defineExpose({
               <div class="segment-badge">
                 {{ item.sorting ?? index + 1 }}
               </div>
-              <div v-if="!hideScore" class="score-text">
+              <div v-if="resolveScoreLine(item)" class="score-text">
-                {{ $t('documentCollection.similarityScore') }}: {{ item.score }}
+                {{ resolveScoreLine(item) }}
               </div>
             </div>
             <div
@@ -174,8 +197,18 @@ defineExpose({
               </ElTag>
             </div>
           </div>
-          <div class="content-desc">
+          <div
-            {{ normalizePreviewContent(item.content) }}
+            v-if="resolvePreviewMarkdown(item)"
+            class="content-desc content-desc--markdown"
+          >
+            <ElXMarkdown
+              :markdown="resolvePreviewMarkdown(item)"
+              :allow-html="markdownRenderProps.allowHtml"
+              :sanitize="markdownRenderProps.sanitize"
+            />
+          </div>
+          <div v-else class="content-desc">
+            {{ item.content }}
           </div>
         </div>
       </div>
@@ -193,17 +226,17 @@ defineExpose({
       <div class="action-buttons">
         <ElButton
           :style="{ minWidth: '100px', height: '36px' }"
-          @click="onCancel"
+          @click="handleCancel"
         >
-          {{ $t('documentCollection.actions.confirmImport') }}
+          {{ $t('documentCollection.actions.cancelImport') }}
         </ElButton>
         <ElButton
           type="primary"
           :style="{ minWidth: '100px', height: '36px' }"
           :loading="disabledConfirm"
-          @click="onConfirm"
+          @click="handleConfirm"
         >
-          {{ $t('documentCollection.actions.cancelImport') }}
+          {{ $t('documentCollection.actions.confirmImport') }}
         </ElButton>
       </div>
     </div>
@@ -276,12 +309,26 @@ defineExpose({
         padding: 14px 16px;
         font-size: 14px;
         line-height: 1.6;
-        white-space: pre-wrap;
         word-break: break-word;
         background: rgb(248 250 252 / 90%);
         border-radius: 14px;
       }
 
+      .content-desc--markdown :deep(.markdown-body) {
+        font-size: inherit;
+        line-height: inherit;
+        color: inherit;
+        background: transparent;
+      }
+
+      .content-desc--markdown :deep(.markdown-body > :first-child) {
+        margin-top: 0;
+      }
+
+      .content-desc--markdown :deep(.markdown-body > :last-child) {
+        margin-bottom: 0;
+      }
+
       .score-text {
         font-size: 13px;
         font-weight: 500;

easyflow-ui-admin/app/src/views/ai/documentCollection/SegmenterDoc.vue

@@ -1,5 +1,5 @@
 <script setup lang="ts">
-import { reactive, ref, watch } from 'vue';
+import { computed, reactive, ref, watch } from 'vue';
 
 import { $t } from '@easyflow/locales';
 
@@ -8,6 +8,7 @@ import {
   ElForm,
   ElFormItem,
   ElInput,
+  ElInputNumber,
   ElMessage,
   ElOption,
   ElSelect,
@@ -85,7 +86,7 @@ const createDefaultFormState = () => ({
   mdSplitterLevel: 2,
   overlapSize: 128,
   regex: '',
-  rowsPerChunk: 1,
+  rowsPerChunk: 10,
   strategyCode: 'AUTO',
 });
 
@@ -126,6 +127,17 @@ const strategyOptions = [
   },
 ];
 
+const fileExt = computed(() =>
+  String(props.documentTitle || '')
+    .split('.')
+    .pop()
+    ?.toLowerCase() || '',
+);
+
+const isPptx = computed(() => fileExt.value === 'pptx');
+const isXlsx = computed(() => fileExt.value === 'xlsx');
+const showStrategySelector = computed(() => !isPptx.value && !isXlsx.value);
+
 const mdLevels = [1, 2, 3, 4, 5, 6];
 
 const showLengthSettings = (strategyCode?: string) =>
@@ -150,9 +162,22 @@ const resetPreviewState = () => {
   previewError.value = '';
 };
 
-const buildStrategyConfig = () => ({
+const buildStrategyConfig = () => {
-  ...formState,
+  if (isPptx.value) {
-});
+    return {
+      strategyCode: 'OFFICE_PPTX_PAGE',
+    };
+  }
+  if (isXlsx.value) {
+    return {
+      rowsPerChunk: formState.rowsPerChunk,
+      strategyCode: 'OFFICE_XLSX_ROW_WINDOW',
+    };
+  }
+  return {
+    ...formState,
+  };
+};
 
 const normalizeSourceRanges = (ranges?: SourceRange[]) =>
   Array.isArray(ranges)
@@ -292,6 +317,13 @@ watch(
     previewSequence += 1;
     clearPreviewTimer();
     Object.assign(formState, createDefaultFormState());
+    if (isPptx.value) {
+      formState.strategyCode = 'OFFICE_PPTX_PAGE';
+    }
+    if (isXlsx.value) {
+      formState.strategyCode = 'OFFICE_XLSX_ROW_WINDOW';
+      formState.rowsPerChunk = 10;
+    }
     resetPreviewState();
     if (activeDocumentId.value) {
       schedulePreviewGeneration();
@@ -317,6 +349,7 @@ watch(
     <ElForm :model="formState" label-position="top" class="workbench__form">
       <div class="workbench__form-grid">
         <ElFormItem
+          v-if="showStrategySelector"
           :label="$t('documentCollection.importDoc.strategySelection')"
           class="workbench__form-full"
         >
@@ -330,6 +363,20 @@ watch(
           </ElSelect>
         </ElFormItem>
 
+        <ElFormItem
+          v-if="isXlsx"
+          label="每多少行分一块"
+          class="workbench__form-full"
+        >
+          <ElInputNumber
+            v-model="formState.rowsPerChunk"
+            :min="1"
+            :max="200"
+            :step="1"
+            class="workbench__rows-input"
+          />
+        </ElFormItem>
+
         <ElFormItem
           v-if="showLengthSettings(formState.strategyCode)"
           :label="$t('documentCollection.splitterDoc.chunkSize')"
@@ -478,6 +525,10 @@ watch(
   box-shadow: 0 0 0 1px rgb(15 23 42 / 7%) inset;
 }
 
+.workbench__rows-input {
+  width: 100%;
+}
+
 :deep(.workbench__form .el-slider__runway) {
   background: rgb(15 23 42 / 8%);
 }

easyflow-ui-admin/app/src/views/ai/documentCollection/SplitterDocPreview.vue

@@ -1,10 +1,16 @@
 <script setup lang="ts">
 import { computed, ref, watch } from 'vue';
+import ElXMarkdown from 'vue-element-plus-x/es/XMarkdown/index.js';
 
 import { $t } from '@easyflow/locales';
 
 import { ElEmpty, ElSkeleton, ElTag } from 'element-plus';
 
+import {
+  markdownRenderProps,
+  resolveMarkdownContent,
+} from '#/views/ai/documentCollection/markdown-content';
+
 interface SourceRange {
   end: number;
   start: number;
@@ -21,6 +27,7 @@ interface ChunkItem {
   partNo?: number;
   partTotal?: number;
   question?: string;
+  renderMarkdown?: string;
   sourceLabel?: string;
   sourceRanges?: SourceRange[];
   tokenEstimate?: number;
@@ -180,15 +187,38 @@ const isActiveChunk = (chunk: ChunkItem) =>
           <div v-if="chunk.chunkType === 'qa_pair'" class="qa-block">
             <div class="qa-block__item">
               <span class="qa-block__label">Q</span>
-              <span class="qa-block__text">{{ chunk.question }}</span>
+              <div class="qa-block__text">
+                <ElXMarkdown
+                  :markdown="resolveMarkdownContent(chunk.question)"
+                  :allow-html="markdownRenderProps.allowHtml"
+                  :sanitize="markdownRenderProps.sanitize"
+                />
+              </div>
             </div>
             <div class="qa-block__item">
               <span class="qa-block__label">A</span>
-              <span class="qa-block__text">{{ chunk.answer }}</span>
+              <div class="qa-block__text">
+                <ElXMarkdown
+                  :markdown="resolveMarkdownContent(chunk.answer)"
+                  :allow-html="markdownRenderProps.allowHtml"
+                  :sanitize="markdownRenderProps.sanitize"
+                />
+              </div>
             </div>
           </div>
 
-          <pre class="chunk-card__content">{{ chunk.content }}</pre>
+          <div
+            v-if="resolveMarkdownContent(chunk.renderMarkdown || chunk.content)"
+            class="chunk-card__content chunk-card__content--markdown"
+          >
+            <ElXMarkdown
+              :markdown="
+                resolveMarkdownContent(chunk.renderMarkdown || chunk.content)
+              "
+              :allow-html="markdownRenderProps.allowHtml"
+              :sanitize="markdownRenderProps.sanitize"
+            />
+          </div>
 
           <div
             v-if="chunk.warnings && chunk.warnings.length > 0"
@@ -342,14 +372,129 @@ const isActiveChunk = (chunk: ChunkItem) =>
 
 .chunk-card__content {
   margin: 14px 0 0;
-  font-family: inherit;
   font-size: 13px;
   line-height: 1.75;
   color: var(--el-text-color-regular);
-  white-space: pre-wrap;
+}
+
+.chunk-card__content--markdown {
+  min-width: 0;
+}
+
+.chunk-card__content--markdown :deep(.markdown-body) {
+  font-size: inherit;
+  line-height: inherit;
+  color: inherit;
+  background: transparent;
+}
+
+.chunk-card__content--markdown :deep(.markdown-body > :first-child) {
+  margin-top: 0;
+}
+
+.chunk-card__content--markdown :deep(.markdown-body > :last-child) {
+  margin-bottom: 0;
+}
+
+.chunk-card__content--markdown :deep(*) {
   overflow-wrap: anywhere;
 }
 
+.chunk-card__content--markdown :deep(p) {
+  margin: 0 0 10px;
+}
+
+.chunk-card__content--markdown :deep(p:last-child) {
+  margin-bottom: 0;
+}
+
+.chunk-card__content--markdown :deep(h1),
+.chunk-card__content--markdown :deep(h2),
+.chunk-card__content--markdown :deep(h3),
+.chunk-card__content--markdown :deep(h4),
+.chunk-card__content--markdown :deep(h5),
+.chunk-card__content--markdown :deep(h6) {
+  margin: 14px 0 10px;
+  font-weight: 600;
+  line-height: 1.45;
+  color: var(--el-text-color-primary);
+}
+
+.chunk-card__content--markdown :deep(h1:first-child),
+.chunk-card__content--markdown :deep(h2:first-child),
+.chunk-card__content--markdown :deep(h3:first-child),
+.chunk-card__content--markdown :deep(h4:first-child),
+.chunk-card__content--markdown :deep(h5:first-child),
+.chunk-card__content--markdown :deep(h6:first-child) {
+  margin-top: 0;
+}
+
+.chunk-card__content--markdown :deep(ul),
+.chunk-card__content--markdown :deep(ol) {
+  padding-left: 20px;
+  margin: 0 0 12px;
+}
+
+.chunk-card__content--markdown :deep(li + li) {
+  margin-top: 4px;
+}
+
+.chunk-card__content--markdown :deep(a) {
+  color: var(--el-color-primary);
+  text-decoration: underline;
+  text-underline-offset: 2px;
+}
+
+.chunk-card__content--markdown :deep(img) {
+  display: block;
+  max-width: min(100%, 520px);
+  height: auto;
+  margin: 12px 0;
+  border: 1px solid rgb(15 23 42 / 8%);
+  border-radius: 12px;
+  box-shadow: 0 10px 24px rgb(15 23 42 / 8%);
+}
+
+.chunk-card__content--markdown :deep(table) {
+  width: 100%;
+  margin: 12px 0;
+  overflow: hidden;
+  border-collapse: collapse;
+  background: rgb(255 255 255 / 92%);
+  border: 1px solid rgb(15 23 42 / 8%);
+  border-radius: 12px;
+}
+
+.chunk-card__content--markdown :deep(th),
+.chunk-card__content--markdown :deep(td) {
+  padding: 10px 12px;
+  text-align: left;
+  vertical-align: top;
+  border: 1px solid rgb(15 23 42 / 8%);
+}
+
+.chunk-card__content--markdown :deep(th) {
+  font-weight: 600;
+  color: var(--el-text-color-primary);
+  background: rgb(37 99 235 / 4%);
+}
+
+.chunk-card__content--markdown :deep(pre) {
+  max-width: 100%;
+  padding: 12px 14px;
+  overflow: auto;
+  background: rgb(15 23 42 / 4%);
+  border: 1px solid rgb(15 23 42 / 6%);
+  border-radius: 12px;
+}
+
+.chunk-card__content--markdown :deep(blockquote) {
+  padding-left: 12px;
+  margin: 12px 0;
+  color: var(--el-text-color-secondary);
+  border-left: 3px solid rgb(37 99 235 / 24%);
+}
+
 .chunk-card__warnings {
   display: flex;
   flex-wrap: wrap;
@@ -387,10 +532,32 @@ const isActiveChunk = (chunk: ChunkItem) =>
 }
 
 .qa-block__text {
+  min-width: 0;
   font-size: 13px;
   line-height: 1.7;
   color: var(--el-text-color-regular);
-  white-space: pre-wrap;
+}
+
+.qa-block__text :deep(.markdown-body) {
+  font-size: inherit;
+  line-height: inherit;
+  background: transparent;
+}
+
+.qa-block__text :deep(.markdown-body > :first-child) {
+  margin-top: 0;
+}
+
+.qa-block__text :deep(.markdown-body > :last-child) {
+  margin-bottom: 0;
+}
+
+.qa-block__text :deep(img) {
+  display: block;
+  max-width: min(100%, 420px);
+  height: auto;
+  margin: 10px 0;
+  border-radius: 10px;
 }
 
 @media (max-width: 960px) {

easyflow-ui-admin/app/src/views/ai/documentCollection/markdown-content.ts

@@ -0,0 +1,40 @@
+const ESCAPED_TABLE_HTML_TAG_PATTERN =
+  /<\/?(?:table|thead|tbody|tfoot|tr|th|td|caption|colgroup|col)\b/i;
+
+/**
+ * 将知识库分块内容规整为可直接交给 Markdown 组件的文本。
+ * 这里额外兼容被转义的 HTML 片段，例如 `<table>...</table>`。
+ */
+export const resolveMarkdownContent = (content?: string) => {
+  const markdown = String(content || '').trim();
+  if (!markdown) {
+    return '';
+  }
+
+  if (ESCAPED_TABLE_HTML_TAG_PATTERN.test(markdown)) {
+    return decodeHtmlEntities(markdown);
+  }
+
+  return markdown;
+};
+
+/**
+ * 统一控制 Markdown 中原生 HTML 的开启策略。
+ */
+export const markdownRenderProps = {
+  allowHtml: true,
+  sanitize: true,
+} as const;
+
+/**
+ * 解码后端可能返回的 HTML 实体，便于 Markdown 组件继续处理原生标签。
+ */
+function decodeHtmlEntities(content: string) {
+  if (typeof window === 'undefined' || window.DOMParser === undefined) {
+    return content;
+  }
+
+  const parser = new window.DOMParser();
+  const doc = parser.parseFromString(content, 'text/html');
+  return doc.documentElement.textContent || content;
+}