fix document extractor node incorrectly processing doc and ppt files (#12902)

api/constants/__init__.py

@@ -15,7 +15,7 @@ AUDIO_EXTENSIONS.extend([ext.upper() for ext in AUDIO_EXTENSIONS])
 
 if dify_config.ETL_TYPE == "Unstructured":
     DOCUMENT_EXTENSIONS = ["txt", "markdown", "md", "mdx", "pdf", "html", "htm", "xlsx", "xls"]
-    DOCUMENT_EXTENSIONS.extend(("docx", "csv", "eml", "msg", "pptx", "xml", "epub"))
+    DOCUMENT_EXTENSIONS.extend(("doc", "docx", "csv", "eml", "msg", "pptx", "xml", "epub"))
     if dify_config.UNSTRUCTURED_API_URL:
         DOCUMENT_EXTENSIONS.append("ppt")
     DOCUMENT_EXTENSIONS.extend([ext.upper() for ext in DOCUMENT_EXTENSIONS])

api/core/workflow/nodes/document_extractor/node.py

@@ -107,8 +107,10 @@ def _extract_text_by_mime_type(*, file_content: bytes, mime_type: str) -> str:
             return _extract_text_from_plain_text(file_content)
         case "application/pdf":
             return _extract_text_from_pdf(file_content)
-        case "application/vnd.openxmlformats-officedocument.wordprocessingml.document" | "application/msword":
+        case "application/msword":
             return _extract_text_from_doc(file_content)
+        case "application/vnd.openxmlformats-officedocument.wordprocessingml.document":
+            return _extract_text_from_docx(file_content)
         case "text/csv":
             return _extract_text_from_csv(file_content)
         case "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet" | "application/vnd.ms-excel":
@@ -142,8 +144,10 @@ def _extract_text_by_file_extension(*, file_content: bytes, file_extension: str)
             return _extract_text_from_yaml(file_content)
         case ".pdf":
             return _extract_text_from_pdf(file_content)
-        case ".doc" | ".docx":
+        case ".doc":
             return _extract_text_from_doc(file_content)
+        case ".docx":
+            return _extract_text_from_docx(file_content)
         case ".csv":
             return _extract_text_from_csv(file_content)
         case ".xls" | ".xlsx":
@@ -203,7 +207,33 @@ def _extract_text_from_pdf(file_content: bytes) -> str:
 
 def _extract_text_from_doc(file_content: bytes) -> str:
     """
-    Extract text from a DOC/DOCX file.
+    Extract text from a DOC file.
+    """
+    from unstructured.partition.api import partition_via_api
+
+    if not (dify_config.UNSTRUCTURED_API_URL and dify_config.UNSTRUCTURED_API_KEY):
+        raise TextExtractionError("UNSTRUCTURED_API_URL and UNSTRUCTURED_API_KEY must be set")
+
+    try:
+        with tempfile.NamedTemporaryFile(suffix=".doc", delete=False) as temp_file:
+            temp_file.write(file_content)
+            temp_file.flush()
+            with open(temp_file.name, "rb") as file:
+                elements = partition_via_api(
+                    file=file,
+                    metadata_filename=temp_file.name,
+                    api_url=dify_config.UNSTRUCTURED_API_URL,
+                    api_key=dify_config.UNSTRUCTURED_API_KEY,
+                )
+            os.unlink(temp_file.name)
+        return "\n".join([getattr(element, "text", "") for element in elements])
+    except Exception as e:
+        raise TextExtractionError(f"Failed to extract text from DOC: {str(e)}") from e
+
+
+def _extract_text_from_docx(file_content: bytes) -> str:
+    """
+    Extract text from a DOCX file.
     For now support only paragraph and table add more if needed
     """
     try:
@@ -255,13 +285,13 @@ def _extract_text_from_doc(file_content: bytes) -> str:
 
                         text.append(markdown_table)
                 except Exception as e:
-                    logger.warning(f"Failed to extract table from DOC/DOCX: {e}")
+                    logger.warning(f"Failed to extract table from DOC: {e}")
                     continue
 
         return "\n".join(text)
 
     except Exception as e:
-        raise TextExtractionError(f"Failed to extract text from DOC/DOCX: {str(e)}") from e
+        raise TextExtractionError(f"Failed to extract text from DOCX: {str(e)}") from e
 
 
 def _download_file_content(file: File) -> bytes:
@@ -329,14 +359,29 @@ def _extract_text_from_excel(file_content: bytes) -> str:
 
 
 def _extract_text_from_ppt(file_content: bytes) -> str:
+    from unstructured.partition.api import partition_via_api
     from unstructured.partition.ppt import partition_ppt
 
     try:
+        if dify_config.UNSTRUCTURED_API_URL and dify_config.UNSTRUCTURED_API_KEY:
+            with tempfile.NamedTemporaryFile(suffix=".ppt", delete=False) as temp_file:
+                temp_file.write(file_content)
+                temp_file.flush()
+                with open(temp_file.name, "rb") as file:
+                    elements = partition_via_api(
+                        file=file,
+                        metadata_filename=temp_file.name,
+                        api_url=dify_config.UNSTRUCTURED_API_URL,
+                        api_key=dify_config.UNSTRUCTURED_API_KEY,
+                    )
+                os.unlink(temp_file.name)
+        else:
             with io.BytesIO(file_content) as file:
                 elements = partition_ppt(file=file)
         return "\n".join([getattr(element, "text", "") for element in elements])
+
     except Exception as e:
-        raise TextExtractionError(f"Failed to extract text from PPT: {str(e)}") from e
+        raise TextExtractionError(f"Failed to extract text from PPTX: {str(e)}") from e
 
 
 def _extract_text_from_pptx(file_content: bytes) -> str:

api/tests/unit_tests/core/workflow/nodes/test_document_extractor_node.py

@@ -8,7 +8,7 @@ from core.variables.variables import StringVariable
 from core.workflow.entities.node_entities import NodeRunResult
 from core.workflow.nodes.document_extractor import DocumentExtractorNode, DocumentExtractorNodeData
 from core.workflow.nodes.document_extractor.node import (
-    _extract_text_from_doc,
+    _extract_text_from_docx,
     _extract_text_from_pdf,
     _extract_text_from_plain_text,
 )
@@ -120,7 +120,7 @@ def test_run_extract_text(
         monkeypatch.setattr("core.workflow.nodes.document_extractor.node._extract_text_from_pdf", mock_pdf_extract)
     elif mime_type.startswith("application/vnd.openxmlformats"):
         mock_docx_extract = Mock(return_value=expected_text[0])
-        monkeypatch.setattr("core.workflow.nodes.document_extractor.node._extract_text_from_doc", mock_docx_extract)
+        monkeypatch.setattr("core.workflow.nodes.document_extractor.node._extract_text_from_docx", mock_docx_extract)
 
     result = document_extractor_node._run()
 
@@ -163,14 +163,14 @@ def test_extract_text_from_pdf(mock_pdf_document):
 
 
 @patch("docx.Document")
-def test_extract_text_from_doc(mock_document):
+def test_extract_text_from_docx(mock_document):
     mock_paragraph1 = Mock()
     mock_paragraph1.text = "Paragraph 1"
     mock_paragraph2 = Mock()
     mock_paragraph2.text = "Paragraph 2"
     mock_document.return_value.paragraphs = [mock_paragraph1, mock_paragraph2]
 
-    text = _extract_text_from_doc(b"PK\x03\x04")
+    text = _extract_text_from_docx(b"PK\x03\x04")
     assert text == "Paragraph 1\nParagraph 2"
 
 

web/app/components/base/prompt-editor/constants.tsx

@@ -52,7 +52,7 @@ export const getInputVars = (text: string): ValueSelector[] => {
 
 export const FILE_EXTS: Record<string, string[]> = {
   [SupportUploadFileTypes.image]: ['JPG', 'JPEG', 'PNG', 'GIF', 'WEBP', 'SVG'],
-  [SupportUploadFileTypes.document]: ['TXT', 'MD', 'MDX', 'MARKDOWN', 'PDF', 'HTML', 'XLSX', 'XLS', 'DOCX', 'CSV', 'EML', 'MSG', 'PPTX', 'PPT', 'XML', 'EPUB'],
+  [SupportUploadFileTypes.document]: ['TXT', 'MD', 'MDX', 'MARKDOWN', 'PDF', 'HTML', 'XLSX', 'XLS', 'DOC', 'DOCX', 'CSV', 'EML', 'MSG', 'PPTX', 'PPT', 'XML', 'EPUB'],
   [SupportUploadFileTypes.audio]: ['MP3', 'M4A', 'WAV', 'WEBM', 'AMR', 'MPGA'],
   [SupportUploadFileTypes.video]: ['MP4', 'MOV', 'MPEG', 'MPGA'],
 }