Feature: Remote OCR (Azure AI) (#10320)

src/paperless/settings.py

@@ -321,6 +321,7 @@ INSTALLED_APPS = [
     "paperless_tesseract.apps.PaperlessTesseractConfig",
     "paperless_text.apps.PaperlessTextConfig",
     "paperless_mail.apps.PaperlessMailConfig",
+    "paperless_remote.apps.PaperlessRemoteParserConfig",
     "django.contrib.admin",
     "rest_framework",
     "rest_framework.authtoken",
@@ -1396,3 +1397,10 @@ WEBHOOKS_ALLOW_INTERNAL_REQUESTS = __get_boolean(
     "PAPERLESS_WEBHOOKS_ALLOW_INTERNAL_REQUESTS",
     "true",
 )
+
+###############################################################################
+# Remote Parser                                                               #
+###############################################################################
+REMOTE_OCR_ENGINE = os.getenv("PAPERLESS_REMOTE_OCR_ENGINE")
+REMOTE_OCR_API_KEY = os.getenv("PAPERLESS_REMOTE_OCR_API_KEY")
+REMOTE_OCR_ENDPOINT = os.getenv("PAPERLESS_REMOTE_OCR_ENDPOINT")

src/paperless_remote/__init__.py

@@ -0,0 +1,4 @@
+# this is here so that django finds the checks.
+from paperless_remote.checks import check_remote_parser_configured
+
+__all__ = ["check_remote_parser_configured"]

src/paperless_remote/apps.py

@@ -0,0 +1,14 @@
+from django.apps import AppConfig
+
+from paperless_remote.signals import remote_consumer_declaration
+
+
+class PaperlessRemoteParserConfig(AppConfig):
+    name = "paperless_remote"
+
+    def ready(self):
+        from documents.signals import document_consumer_declaration
+
+        document_consumer_declaration.connect(remote_consumer_declaration)
+
+        AppConfig.ready(self)

src/paperless_remote/checks.py

@@ -0,0 +1,17 @@
+from django.conf import settings
+from django.core.checks import Error
+from django.core.checks import register
+
+
+@register()
+def check_remote_parser_configured(app_configs, **kwargs):
+    if settings.REMOTE_OCR_ENGINE == "azureai" and not (
+        settings.REMOTE_OCR_ENDPOINT and settings.REMOTE_OCR_API_KEY
+    ):
+        return [
+            Error(
+                "Azure AI remote parser requires endpoint and API key to be configured.",
+            ),
+        ]
+
+    return []

src/paperless_remote/parsers.py

@@ -0,0 +1,118 @@
+from pathlib import Path
+
+from django.conf import settings
+
+from paperless_tesseract.parsers import RasterisedDocumentParser
+
+
+class RemoteEngineConfig:
+    def __init__(
+        self,
+        engine: str,
+        api_key: str | None = None,
+        endpoint: str | None = None,
+    ):
+        self.engine = engine
+        self.api_key = api_key
+        self.endpoint = endpoint
+
+    def engine_is_valid(self):
+        valid = self.engine in ["azureai"] and self.api_key is not None
+        if self.engine == "azureai":
+            valid = valid and self.endpoint is not None
+        return valid
+
+
+class RemoteDocumentParser(RasterisedDocumentParser):
+    """
+    This parser uses a remote OCR engine to parse documents. Currently, it supports Azure AI Vision
+    as this is the only service that provides a remote OCR API with text-embedded PDF output.
+    """
+
+    logging_name = "paperless.parsing.remote"
+
+    def get_settings(self) -> RemoteEngineConfig:
+        """
+        Returns the configuration for the remote OCR engine, loaded from Django settings.
+        """
+        return RemoteEngineConfig(
+            engine=settings.REMOTE_OCR_ENGINE,
+            api_key=settings.REMOTE_OCR_API_KEY,
+            endpoint=settings.REMOTE_OCR_ENDPOINT,
+        )
+
+    def supported_mime_types(self):
+        if self.settings.engine_is_valid():
+            return {
+                "application/pdf": ".pdf",
+                "image/png": ".png",
+                "image/jpeg": ".jpg",
+                "image/tiff": ".tiff",
+                "image/bmp": ".bmp",
+                "image/gif": ".gif",
+                "image/webp": ".webp",
+            }
+        else:
+            return {}
+
+    def azure_ai_vision_parse(
+        self,
+        file: Path,
+    ) -> str | None:
+        """
+        Uses Azure AI Vision to parse the document and return the text content.
+        It requests a searchable PDF output with embedded text.
+        The PDF is saved to the archive_path attribute.
+        Returns the text content extracted from the document.
+        If the parsing fails, it returns None.
+        """
+        from azure.ai.documentintelligence import DocumentIntelligenceClient
+        from azure.ai.documentintelligence.models import AnalyzeDocumentRequest
+        from azure.ai.documentintelligence.models import AnalyzeOutputOption
+        from azure.ai.documentintelligence.models import DocumentContentFormat
+        from azure.core.credentials import AzureKeyCredential
+
+        client = DocumentIntelligenceClient(
+            endpoint=self.settings.endpoint,
+            credential=AzureKeyCredential(self.settings.api_key),
+        )
+
+        try:
+            with file.open("rb") as f:
+                analyze_request = AnalyzeDocumentRequest(bytes_source=f.read())
+                poller = client.begin_analyze_document(
+                    model_id="prebuilt-read",
+                    body=analyze_request,
+                    output_content_format=DocumentContentFormat.TEXT,
+                    output=[AnalyzeOutputOption.PDF],  # request searchable PDF output
+                    content_type="application/json",
+                )
+
+            poller.wait()
+            result_id = poller.details["operation_id"]
+            result = poller.result()
+
+            # Download the PDF with embedded text
+            self.archive_path = self.tempdir / "archive.pdf"
+            with self.archive_path.open("wb") as f:
+                for chunk in client.get_analyze_result_pdf(
+                    model_id="prebuilt-read",
+                    result_id=result_id,
+                ):
+                    f.write(chunk)
+            return result.content
+        except Exception as e:
+            self.log.error(f"Azure AI Vision parsing failed: {e}")
+        finally:
+            client.close()
+
+        return None
+
+    def parse(self, document_path: Path, mime_type, file_name=None):
+        if not self.settings.engine_is_valid():
+            self.log.warning(
+                "No valid remote parser engine is configured, content will be empty.",
+            )
+            self.text = ""
+        elif self.settings.engine == "azureai":
+            self.text = self.azure_ai_vision_parse(document_path)

src/paperless_remote/signals.py

@@ -0,0 +1,18 @@
+def get_parser(*args, **kwargs):
+    from paperless_remote.parsers import RemoteDocumentParser
+
+    return RemoteDocumentParser(*args, **kwargs)
+
+
+def get_supported_mime_types():
+    from paperless_remote.parsers import RemoteDocumentParser
+
+    return RemoteDocumentParser(None).supported_mime_types()
+
+
+def remote_consumer_declaration(sender, **kwargs):
+    return {
+        "parser": get_parser,
+        "weight": 5,
+        "mime_types": get_supported_mime_types(),
+    }

src/paperless_remote/tests/test_checks.py

@@ -0,0 +1,24 @@
+from unittest import TestCase
+
+from django.test import override_settings
+
+from paperless_remote import check_remote_parser_configured
+
+
+class TestChecks(TestCase):
+    @override_settings(REMOTE_OCR_ENGINE=None)
+    def test_no_engine(self):
+        msgs = check_remote_parser_configured(None)
+        self.assertEqual(len(msgs), 0)
+
+    @override_settings(REMOTE_OCR_ENGINE="azureai")
+    @override_settings(REMOTE_OCR_API_KEY="somekey")
+    @override_settings(REMOTE_OCR_ENDPOINT=None)
+    def test_azure_no_endpoint(self):
+        msgs = check_remote_parser_configured(None)
+        self.assertEqual(len(msgs), 1)
+        self.assertTrue(
+            msgs[0].msg.startswith(
+                "Azure AI remote parser requires endpoint and API key to be configured.",
+            ),
+        )

src/paperless_remote/tests/test_parser.py

@@ -0,0 +1,128 @@
+import uuid
+from pathlib import Path
+from unittest import mock
+
+from django.test import TestCase
+from django.test import override_settings
+
+from documents.tests.utils import DirectoriesMixin
+from documents.tests.utils import FileSystemAssertsMixin
+from paperless_remote.parsers import RemoteDocumentParser
+from paperless_remote.signals import get_parser
+
+
+class TestParser(DirectoriesMixin, FileSystemAssertsMixin, TestCase):
+    SAMPLE_FILES = Path(__file__).resolve().parent / "samples"
+
+    def assertContainsStrings(self, content: str, strings: list[str]):
+        # Asserts that all strings appear in content, in the given order.
+        indices = []
+        for s in strings:
+            if s in content:
+                indices.append(content.index(s))
+            else:
+                self.fail(f"'{s}' is not in '{content}'")
+        self.assertListEqual(indices, sorted(indices))
+
+    @mock.patch("paperless_tesseract.parsers.run_subprocess")
+    @mock.patch("azure.ai.documentintelligence.DocumentIntelligenceClient")
+    def test_get_text_with_azure(self, mock_client_cls, mock_subprocess):
+        # Arrange mock Azure client
+        mock_client = mock.Mock()
+        mock_client_cls.return_value = mock_client
+
+        # Simulate poller result and its `.details`
+        mock_poller = mock.Mock()
+        mock_poller.wait.return_value = None
+        mock_poller.details = {"operation_id": "fake-op-id"}
+        mock_client.begin_analyze_document.return_value = mock_poller
+        mock_poller.result.return_value.content = "This is a test document."
+
+        # Return dummy PDF bytes
+        mock_client.get_analyze_result_pdf.return_value = [
+            b"%PDF-",
+            b"1.7 ",
+            b"FAKEPDF",
+        ]
+
+        # Simulate pdftotext by writing dummy text to sidecar file
+        def fake_run(cmd, *args, **kwargs):
+            with Path(cmd[-1]).open("w", encoding="utf-8") as f:
+                f.write("This is a test document.")
+
+        mock_subprocess.side_effect = fake_run
+
+        with override_settings(
+            REMOTE_OCR_ENGINE="azureai",
+            REMOTE_OCR_API_KEY="somekey",
+            REMOTE_OCR_ENDPOINT="https://endpoint.cognitiveservices.azure.com",
+        ):
+            parser = get_parser(uuid.uuid4())
+            parser.parse(
+                self.SAMPLE_FILES / "simple-digital.pdf",
+                "application/pdf",
+            )
+
+            self.assertContainsStrings(
+                parser.text.strip(),
+                ["This is a test document."],
+            )
+
+    @mock.patch("azure.ai.documentintelligence.DocumentIntelligenceClient")
+    def test_get_text_with_azure_error_logged_and_returns_none(self, mock_client_cls):
+        mock_client = mock.Mock()
+        mock_client.begin_analyze_document.side_effect = RuntimeError("fail")
+        mock_client_cls.return_value = mock_client
+
+        with override_settings(
+            REMOTE_OCR_ENGINE="azureai",
+            REMOTE_OCR_API_KEY="somekey",
+            REMOTE_OCR_ENDPOINT="https://endpoint.cognitiveservices.azure.com",
+        ):
+            parser = get_parser(uuid.uuid4())
+            with mock.patch.object(parser.log, "error") as mock_log_error:
+                parser.parse(
+                    self.SAMPLE_FILES / "simple-digital.pdf",
+                    "application/pdf",
+                )
+
+        self.assertIsNone(parser.text)
+        mock_client.begin_analyze_document.assert_called_once()
+        mock_client.close.assert_called_once()
+        mock_log_error.assert_called_once()
+        self.assertIn(
+            "Azure AI Vision parsing failed",
+            mock_log_error.call_args[0][0],
+        )
+
+    @override_settings(
+        REMOTE_OCR_ENGINE="azureai",
+        REMOTE_OCR_API_KEY="key",
+        REMOTE_OCR_ENDPOINT="https://endpoint.cognitiveservices.azure.com",
+    )
+    def test_supported_mime_types_valid_config(self):
+        parser = RemoteDocumentParser(uuid.uuid4())
+        expected_types = {
+            "application/pdf": ".pdf",
+            "image/png": ".png",
+            "image/jpeg": ".jpg",
+            "image/tiff": ".tiff",
+            "image/bmp": ".bmp",
+            "image/gif": ".gif",
+            "image/webp": ".webp",
+        }
+        self.assertEqual(parser.supported_mime_types(), expected_types)
+
+    def test_supported_mime_types_invalid_config(self):
+        parser = get_parser(uuid.uuid4())
+        self.assertEqual(parser.supported_mime_types(), {})
+
+    @override_settings(
+        REMOTE_OCR_ENGINE=None,
+        REMOTE_OCR_API_KEY=None,
+        REMOTE_OCR_ENDPOINT=None,
+    )
+    def test_parse_with_invalid_config(self):
+        parser = get_parser(uuid.uuid4())
+        parser.parse(self.SAMPLE_FILES / "simple-digital.pdf", "application/pdf")
+        self.assertEqual(parser.text, "")