Squash Odysseus development history

tests/test_document_pdf_marker.py

@@ -5,6 +5,7 @@ The old call sites used ``str.lstrip("\\n[PDF content]:")``, which treats the
 argument as a *set of characters* and keeps stripping leading characters that
 happen to be in that set — corrupting the start of the extracted document.
 """
+from src import document_processor
 from src.document_processor import strip_pdf_content_marker, _PDF_CONTENT_MARKER
 
 
@@ -28,3 +29,28 @@ def test_text_without_marker_is_only_stripped():
 
 def test_handles_none():
     assert strip_pdf_content_marker(None) == ""
+
+
+def test_local_document_extraction_disables_embedded_image_analysis(monkeypatch, tmp_path):
+    source = tmp_path / "brief.pdf"
+    source.write_bytes(b"%PDF-1.4")
+    observed = {}
+
+    def fake_process(path, owner=None, *, analyze_embedded_images=True):
+        observed.update(
+            path=path,
+            owner=owner,
+            analyze_embedded_images=analyze_embedded_images,
+        )
+        return "extracted"
+
+    monkeypatch.setattr(document_processor, "_process_pdf", fake_process)
+
+    result = document_processor.extract_local_document(str(source))
+
+    assert result == "extracted"
+    assert observed == {
+        "path": str(source),
+        "owner": None,
+        "analyze_embedded_images": False,
+    }