Route uploaded images to vision LLM with document-parser fallback

surfsense_backend/tests/unit/etl_pipeline/test_etl_pipeline_service.py

@@ -549,8 +549,11 @@ def test_unsupported_extensions_classified_correctly(filename):
         ("doc.docx", "document"),
         ("slides.pptx", "document"),
         ("sheet.xlsx", "document"),
-        ("photo.png", "document"),
-        ("photo.jpg", "document"),
+        ("photo.png", "image"),
+        ("photo.jpg", "image"),
+        ("photo.webp", "image"),
+        ("photo.gif", "image"),
+        ("photo.heic", "image"),
         ("book.epub", "document"),
         ("letter.odt", "document"),
         ("readme.md", "plaintext"),
@@ -680,3 +683,57 @@ async def test_extract_eml_with_docling_raises_unsupported(tmp_path, mocker):
         await EtlPipelineService().extract(
             EtlRequest(file_path=str(eml_file), filename="mail.eml")
         )
+
+
+# ---------------------------------------------------------------------------
+# Image extraction via vision LLM
+# ---------------------------------------------------------------------------
+
+
+async def test_extract_image_with_vision_llm(tmp_path):
+    """An image file is analyzed by the vision LLM when provided."""
+    from unittest.mock import AsyncMock, MagicMock
+
+    img_file = tmp_path / "photo.png"
+    img_file.write_bytes(b"\x89PNG\r\n\x1a\n" + b"\x00" * 50)
+
+    fake_response = MagicMock()
+    fake_response.content = "# A photo of a sunset over the ocean"
+    fake_llm = AsyncMock()
+    fake_llm.ainvoke.return_value = fake_response
+
+    service = EtlPipelineService(vision_llm=fake_llm)
+    result = await service.extract(
+        EtlRequest(file_path=str(img_file), filename="photo.png")
+    )
+
+    assert result.markdown_content == "# A photo of a sunset over the ocean"
+    assert result.etl_service == "VISION_LLM"
+    assert result.content_type == "image"
+    fake_llm.ainvoke.assert_called_once()
+
+
+async def test_extract_image_falls_back_to_document_without_vision_llm(
+    tmp_path, mocker
+):
+    """Without a vision LLM, image files fall back to the document parser."""
+    mocker.patch("app.config.config.ETL_SERVICE", "DOCLING")
+
+    fake_docling = mocker.AsyncMock()
+    fake_docling.process_document.return_value = {"content": "# OCR text from image"}
+    mocker.patch(
+        "app.services.docling_service.create_docling_service",
+        return_value=fake_docling,
+    )
+
+    img_file = tmp_path / "scan.png"
+    img_file.write_bytes(b"\x89PNG\r\n\x1a\n" + b"\x00" * 50)
+
+    service = EtlPipelineService()
+    result = await service.extract(
+        EtlRequest(file_path=str(img_file), filename="scan.png")
+    )
+
+    assert result.markdown_content == "# OCR text from image"
+    assert result.etl_service == "DOCLING"
+    assert result.content_type == "document"

surfsense_backend/tests/unit/utils/test_file_extensions.py

@@ -154,3 +154,40 @@ def test_get_extensions_for_none_returns_union():
     )
 
     assert get_document_extensions_for_service(None) == DOCUMENT_EXTENSIONS
+
+
+# ---------------------------------------------------------------------------
+# IMAGE_EXTENSIONS
+# ---------------------------------------------------------------------------
+
+
+@pytest.mark.parametrize(
+    "ext",
+    [
+        ".png",
+        ".jpg",
+        ".jpeg",
+        ".gif",
+        ".bmp",
+        ".tiff",
+        ".tif",
+        ".webp",
+        ".svg",
+        ".heic",
+        ".heif",
+    ],
+)
+def test_image_extensions_contains_expected(ext):
+    from app.utils.file_extensions import IMAGE_EXTENSIONS
+
+    assert ext in IMAGE_EXTENSIONS
+
+
+def test_image_extensions_are_subset_of_document_extensions():
+    """Image extensions used for routing should also be in DOCUMENT_EXTENSIONS for fallback."""
+    from app.utils.file_extensions import DOCUMENT_EXTENSIONS, IMAGE_EXTENSIONS
+
+    missing = IMAGE_EXTENSIONS - DOCUMENT_EXTENSIONS
+    assert not missing, (
+        f"Image extensions missing from document sets (breaks fallback): {missing}"
+    )