fix(api): allow PDF extraction from URLs without an upload file (#41950)

api/core/rag/extractor/extract_processor.py

@@ -147,9 +147,11 @@ class ExtractProcessor:
                             upload_file.id,
                         )
                     elif file_extension == ".pdf":
-                        assert upload_file is not None
                         extractor = PdfExtractor(
-                            file_path, upload_file.tenant_id, upload_file.created_by, session=session
+                            file_path,
+                            upload_file.tenant_id if upload_file else None,
+                            upload_file.created_by if upload_file else None,
+                            session=session,
                         )
                     elif file_extension in {".md", ".markdown", ".mdx"}:
                         extractor = (
@@ -195,9 +197,11 @@ class ExtractProcessor:
                             upload_file.id,
                         )
                     elif file_extension == ".pdf":
-                        assert upload_file is not None
                         extractor = PdfExtractor(
-                            file_path, upload_file.tenant_id, upload_file.created_by, session=session
+                            file_path,
+                            upload_file.tenant_id if upload_file else None,
+                            upload_file.created_by if upload_file else None,
+                            session=session,
                         )
                     elif file_extension in {".md", ".markdown", ".mdx"}:
                         extractor = MarkdownExtractor(file_path, autodetect_encoding=True)

api/core/rag/extractor/pdf_extractor.py

@@ -55,8 +55,8 @@ class PdfExtractor(BaseExtractor):
     def __init__(
         self,
         file_path: str,
-        tenant_id: str,
-        user_id: str,
+        tenant_id: str | None = None,
+        user_id: str | None = None,
         file_cache_key: str | None = None,
         *,
         session: Session | None = None,
@@ -127,6 +127,12 @@ class PdfExtractor(BaseExtractor):
         Returns:
             Markdown string containing links to the extracted images.
         """
+        if not self._tenant_id or not self._user_id:
+            # No tenant context (e.g. file loaded from a URL rather than an
+            # uploaded file): extracted images cannot be persisted, so skip
+            # image extraction and return text only.
+            return ""
+
         image_content = []
         upload_files = []
         base_url = dify_config.FILES_URL

api/tests/unit_tests/core/rag/extractor/test_extract_processor.py

@@ -152,6 +152,21 @@ class TestExtractProcessorLoaders:
 
         assert text == content
 
+    def test_load_from_url_extracts_pdf_without_upload_file(self, monkeypatch: pytest.MonkeyPatch):
+        response = SimpleNamespace(headers={"Content-Type": "application/pdf"}, content=b"%PDF-1.1 body")
+        monkeypatch.setattr(processor_module.remote_fetcher, "make_request", lambda *args, **kwargs: response)
+        factory = _patch_all_extractors(monkeypatch)
+        apply_config_overrides(monkeypatch, ETL_TYPE="dify")
+
+        text = ExtractProcessor.load_from_url("https://example.com/report", return_text=True)
+
+        assert text == "extracted-by-PdfExtractor"
+        name, args, kwargs = factory.calls[-1]
+        assert name == "PdfExtractor"
+        # no upload_file for URL-loaded files: tenant/user context must be None
+        assert args[1] is None
+        assert args[2] is None
+
 
 class TestExtractProcessorFileRouting:
     @pytest.fixture(autouse=True)

api/tests/unit_tests/core/rag/extractor/test_pdf_extractor.py

@@ -196,3 +196,15 @@ def test_extract_images_failures(mock_dependencies: _Dependencies):
     assert upload_file is not None
     assert f"![image](http://files.local/files/{upload_file.id}/file-preview)" in result
     assert mock_dependencies.storage.saves == [(upload_file.key, jpeg_bytes)]
+
+
+def test_extract_images_skipped_without_tenant_context():
+    """PDFs loaded from a URL have no tenant/user context; image extraction must be skipped."""
+    mock_page = MagicMock()
+
+    extractor = pe.PdfExtractor(file_path="test.pdf")
+
+    result = extractor._extract_images(mock_page)
+
+    assert result == ""
+    mock_page.get_objects.assert_not_called()