Chore: make the backend test suite pass on PostgreSQL

Remove SQLite-only assumptions from the tests: hard-coded primary keys,
NULL ordering, the database vendor, and the CPU-dependent default process
count. Keep the test connection open inside the test transaction, since
close_old_connections() and connections.close_all() really close it on
PostgreSQL.

src-ui/src/app/components/common/custom-fields-dropdown/custom-fields-dropdown.component.scss

@@ -1,10 +1,3 @@
 .custom-fields-dropdown {
     min-width: 300px;
-
-    // correct position on mobile
-    @media (max-width: 575.98px) {
-        &.show {
-            margin-left: -245px !important;
-        }
-    }
 }

src-ui/src/app/components/document-detail/document-detail.component.html

@@ -308,7 +308,7 @@
                   </tr>
                   <tr class="align-middle">
                     <td i18n>Original SHA256 checksum</td>
-                    <td><span class="badge bg-light font-monospace text-wrap text-break text-start">{{metadata()?.original_checksum}}</span></td>
+                    <td><span class="badge bg-light text-dark font-monospace text-wrap text-break text-start">{{metadata()?.original_checksum}}</span></td>
                   </tr>
                   <tr>
                     <td i18n>Original file size</td>

src/conftest.py

@@ -77,6 +77,30 @@ def _clear_django_caches() -> None:
         cache.clear()
 
 
+@pytest.fixture(autouse=True)
+def _keep_connection_open_in_atomic(monkeypatch: pytest.MonkeyPatch) -> None:
+    """Make ``close_old_connections()`` a no-op inside the test's transaction.
+
+    The Celery signal handlers call it, as a worker would between tasks. Tests
+    run inside an open atomic block, where Postgres really closes the
+    connection, taking the test transaction and every later test on the
+    worker with it. SQLite in-memory ignores the close, which hid this.
+    """
+    from django.db.backends.base.base import BaseDatabaseWrapper
+
+    original = BaseDatabaseWrapper.close_if_unusable_or_obsolete
+
+    def close_unless_atomic(self: BaseDatabaseWrapper) -> None:
+        if not self.in_atomic_block:
+            original(self)
+
+    monkeypatch.setattr(
+        BaseDatabaseWrapper,
+        "close_if_unusable_or_obsolete",
+        close_unless_atomic,
+    )
+
+
 @pytest.fixture
 def paperless_dirs(
     tmp_path: Path,

src/documents/permissions.py

@@ -576,9 +576,12 @@ def annotate_document_count_by_ids(
                        or CustomFieldInstance)
         related_object_field: field on the relation pointing back to queryset pk
         document_ids: the document ids to count against -- a concrete list/set,
-                       or a simple (already resolved) queryset of ids. Keep
-                       the list small: it is embedded in the statement as
-                       literal parameters.
+                       or a simple (already resolved) queryset of ids. Callers
+                       that need this filtered by a complex condition (e.g. a
+                       permission check) should resolve it to a concrete list
+                       first if the same ids will be reused across multiple
+                       calls, rather than passing the complex queryset itself
+                       into each -- see ``_get_selection_data_for_queryset``.
         target_field: field on the relation pointing to Document id
     """
 

src/documents/tests/test_api_custom_fields.py

@@ -1062,7 +1062,7 @@ class TestCustomFieldsAPI(DirectoriesMixin, APITestCase):
                 "custom_fields": [
                     {
                         "field": custom_field_doclink.id,
-                        "value": [2, 3, 4],
+                        "value": [doc2.id, doc3.id, doc4.id],
                     },
                 ],
             },
@@ -1073,13 +1073,13 @@ class TestCustomFieldsAPI(DirectoriesMixin, APITestCase):
         self.assertEqual(CustomFieldInstance.objects.count(), 4)
         _cf_5 = doc2.custom_fields.first()
         assert _cf_5 is not None
-        self.assertEqual(_cf_5.value, [1])
+        self.assertEqual(_cf_5.value, [doc1.id])
         _cf_6 = doc3.custom_fields.first()
         assert _cf_6 is not None
-        self.assertEqual(_cf_6.value, [1])
+        self.assertEqual(_cf_6.value, [doc1.id])
         _cf_7 = doc4.custom_fields.first()
         assert _cf_7 is not None
-        self.assertEqual(_cf_7.value, [1])
+        self.assertEqual(_cf_7.value, [doc1.id])
 
         # Add links appends if necessary
         resp = self.client.patch(
@@ -1088,7 +1088,7 @@ class TestCustomFieldsAPI(DirectoriesMixin, APITestCase):
                 "custom_fields": [
                     {
                         "field": custom_field_doclink.id,
-                        "value": [1, 4],
+                        "value": [doc1.id, doc4.id],
                     },
                 ],
             },
@@ -1098,7 +1098,7 @@ class TestCustomFieldsAPI(DirectoriesMixin, APITestCase):
         self.assertEqual(resp.status_code, status.HTTP_200_OK)
         _cf_8 = doc4.custom_fields.first()
         assert _cf_8 is not None
-        self.assertEqual(_cf_8.value, [1, 3])
+        self.assertEqual(_cf_8.value, [doc1.id, doc3.id])
 
         # Remove one of the links, removed on other doc
         resp = self.client.patch(
@@ -1107,7 +1107,7 @@ class TestCustomFieldsAPI(DirectoriesMixin, APITestCase):
                 "custom_fields": [
                     {
                         "field": custom_field_doclink.id,
-                        "value": [2, 3],
+                        "value": [doc2.id, doc3.id],
                     },
                 ],
             },
@@ -1117,13 +1117,13 @@ class TestCustomFieldsAPI(DirectoriesMixin, APITestCase):
         self.assertEqual(resp.status_code, status.HTTP_200_OK)
         _cf_9 = doc2.custom_fields.first()
         assert _cf_9 is not None
-        self.assertEqual(_cf_9.value, [1])
+        self.assertEqual(_cf_9.value, [doc1.id])
         _cf_10 = doc3.custom_fields.first()
         assert _cf_10 is not None
-        self.assertEqual(_cf_10.value, [1, 4])
+        self.assertEqual(_cf_10.value, [doc1.id, doc4.id])
         _cf_11 = doc4.custom_fields.first()
         assert _cf_11 is not None
-        self.assertEqual(_cf_11.value, [3])
+        self.assertEqual(_cf_11.value, [doc3.id])
 
         # Removes the field entirely
         resp = self.client.patch(
@@ -1140,10 +1140,10 @@ class TestCustomFieldsAPI(DirectoriesMixin, APITestCase):
         self.assertEqual(_cf_12.value, [])
         _cf_13 = doc3.custom_fields.first()
         assert _cf_13 is not None
-        self.assertEqual(_cf_13.value, [4])
+        self.assertEqual(_cf_13.value, [doc4.id])
         _cf_14 = doc4.custom_fields.first()
         assert _cf_14 is not None
-        self.assertEqual(_cf_14.value, [3])
+        self.assertEqual(_cf_14.value, [doc3.id])
 
         # If field exists on target doc but value is None
         doc5 = Document.objects.create(
@@ -1170,7 +1170,7 @@ class TestCustomFieldsAPI(DirectoriesMixin, APITestCase):
         self.assertEqual(resp.status_code, status.HTTP_200_OK)
         _cf_15 = doc5.custom_fields.first()
         assert _cf_15 is not None
-        self.assertEqual(_cf_15.value, [1])
+        self.assertEqual(_cf_15.value, [doc1.id])
 
     def test_documentlink_patch_requires_change_permission_on_target_documents(
         self,

src/documents/tests/test_api_filter_by_custom_fields.py

@@ -628,22 +628,23 @@ class TestCustomFieldsSearch(DirectoriesMixin, APITestCase):
     # Subset check (document link field only)                   #
     # ==========================================================#
     def test_document_link_contains(self) -> None:
+        linked = {self.documents[1].id, self.documents[2].id}
         # Document link field "contains" performs a subset check.
         self._assert_query_match_predicate(
-            ["documentlink_field", "contains", [1, 2]],
+            ["documentlink_field", "contains", sorted(linked)],
             lambda document: (
                 "documentlink_field" in document
                 and document["documentlink_field"] is not None
-                and set(document["documentlink_field"]) >= {1, 2}
+                and set(document["documentlink_field"]) >= linked
             ),
         )
         # The order of IDs don't matter - this is the same as above.
         self._assert_query_match_predicate(
-            ["documentlink_field", "contains", [2, 1]],
+            ["documentlink_field", "contains", sorted(linked, reverse=True)],
             lambda document: (
                 "documentlink_field" in document
                 and document["documentlink_field"] is not None
-                and set(document["documentlink_field"]) >= {1, 2}
+                and set(document["documentlink_field"]) >= linked
             ),
         )
 

src/documents/tests/test_api_search.py

@@ -944,7 +944,7 @@ class TestDocumentSearchApi(DirectoriesMixin, APITestCase):
         self.assertEqual(response.status_code, status.HTTP_200_OK)
 
         results = response.data["results"]
-        self.assertEqual({r["id"] for r in results}, {1, 2})
+        self.assertEqual({r["id"] for r in results}, {d1.id, d2.id})
 
     @mock.patch("documents.search._backend.parse_user_query")
     def test_search_parser_bug_surfaces_as_500_not_400(self, m) -> None:
@@ -1801,6 +1801,7 @@ class TestDocumentSearchApi(DirectoriesMixin, APITestCase):
     def test_search_sorting(self) -> None:
         u1 = UserFactory(username="user1")
         u2 = UserFactory(username="user2")
+        u3 = UserFactory(username="user3")
         c1 = Correspondent.objects.create(name="corres Ax")
         c2 = Correspondent.objects.create(name="corres Cx")
         c3 = Correspondent.objects.create(name="corres Bx")
@@ -1826,6 +1827,7 @@ class TestDocumentSearchApi(DirectoriesMixin, APITestCase):
             content="test",
             archive_serial_number=1,
             title="1",
+            owner=u3,
         )
         Note.objects.create(
             note="This is a note.",
@@ -1878,14 +1880,14 @@ class TestDocumentSearchApi(DirectoriesMixin, APITestCase):
             search_query("&ordering=-num_notes"),
             [d1.id, d3.id, d2.id],
         )
-        # owner sort: ORM orders by owner_id (integer); NULLs first in SQLite ASC
+        # owner sort: ORM orders by owner_id (integer)
         self.assertListEqual(
             search_query("&ordering=owner"),
-            [d3.id, d1.id, d2.id],
+            [d1.id, d2.id, d3.id],
         )
         self.assertListEqual(
             search_query("&ordering=-owner"),
-            [d2.id, d1.id, d3.id],
+            [d3.id, d2.id, d1.id],
         )
 
     def test_search_ordering_by_score(self) -> None:

src/documents/tests/test_api_status.py

@@ -5,6 +5,7 @@ from datetime import timedelta
 from pathlib import Path
 from unittest import mock
 
+from django.db import connection
 from django.test import override_settings
 from django.utils import timezone
 from rest_framework import status
@@ -68,7 +69,7 @@ class TestSystemStatus(APITestCase):
         self.assertEqual(response.data["install_type"], "bare-metal")
         self.assertIsNotNone(response.data["storage"]["total"])
         self.assertIsNotNone(response.data["storage"]["available"])
-        self.assertEqual(response.data["database"]["type"], "sqlite")
+        self.assertEqual(response.data["database"]["type"], connection.vendor)
         self.assertIsNotNone(response.data["database"]["url"])
         self.assertEqual(response.data["database"]["status"], "OK")
         self.assertIsNone(response.data["database"]["error"])

src/documents/tests/test_bulk_edit.py

@@ -1031,9 +1031,7 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         with self.assertLogs("paperless.bulk_edit", level="ERROR") as cm:
             bulk_edit.merge(doc_ids)
             error_str = cm.output[0]
-            expected_str = (
-                "Error merging document 2, it will not be included in the merge"
-            )
+            expected_str = f"Error merging document {self.doc2.id}, it will not be included in the merge"
             self.assertIn(expected_str, error_str)
 
         mock_consume_file.assert_not_called()
@@ -1193,7 +1191,7 @@ class TestPDFActions(DirectoriesMixin, TestCase):
         with self.assertLogs("paperless.bulk_edit", level="ERROR") as cm:
             bulk_edit.split(doc_ids, pages)
             error_str = cm.output[0]
-            expected_str = "Error splitting document 2"
+            expected_str = f"Error splitting document {self.doc2.id}"
             self.assertIn(expected_str, error_str)
 
         mock_consume_file.assert_not_called()

src/documents/tests/test_consumer.py

@@ -248,8 +248,8 @@ class TestConsumer(
         )
         self.assertIsNone(document.correspondent)
         self.assertIsNone(document.document_type)
-        self.assertEqual(document.filename, "0000001.pdf")
-        self.assertEqual(document.archive_filename, "0000001.pdf")
+        self.assertEqual(document.filename, f"{document.pk:07}.pdf")
+        self.assertEqual(document.archive_filename, f"{document.pk:07}.pdf")
 
         self.assertIsFile(document.source_path)
 
@@ -1127,17 +1127,17 @@ class TestConsumer(
         with self.get_consumer(settings.CONSUMPTION_DIR / "simple.png") as consumer:
             consumer.run()
 
-            doc1 = Document.objects.filter(pk=1).first()
+            doc1 = Document.objects.latest("pk")
 
         with self.get_consumer(settings.CONSUMPTION_DIR / "simple.pdf") as consumer:
             consumer.run()
 
-            doc2 = Document.objects.filter(pk=2).first()
+            doc2 = Document.objects.latest("pk")
 
         with self.get_consumer(settings.CONSUMPTION_DIR / "simple.png.pdf") as consumer:
             consumer.run()
 
-            doc3 = Document.objects.filter(pk=3).first()
+            doc3 = Document.objects.latest("pk")
 
         self.assertEqual(doc1.filename, "simple.png")
         self.assertEqual(doc1.archive_filename, "simple.pdf")

src/documents/tests/test_file_handling.py

@@ -760,6 +760,9 @@ class TestFileHandlingWithArchive(DirectoriesMixin, FileSystemAssertsMixin, Test
             archive_checksum="B",
         )
 
+        # Without a format the files are renamed to match the primary key
+        original = doc.source_path
+        archive = doc.archive_path
         self.assertIsFile(original)
         self.assertIsFile(archive)
         self.assertIsFile(doc.source_path)
@@ -978,6 +981,9 @@ class TestFileHandlingWithArchive(DirectoriesMixin, FileSystemAssertsMixin, Test
             archive_filename="0000001.pdf",
         )
 
+        # Without a format the files are renamed to match the primary key
+        original = doc.source_path
+        archive = doc.archive_path
         self.assertIsFile(original)
         self.assertIsFile(archive)
         self.assertIsFile(doc.source_path)

src/documents/tests/test_management_exporter.py

@@ -151,6 +151,20 @@ class TestExportImport(
 
         super().setUp()
 
+    def _install_sample_files(self) -> None:
+        documents_dir = Path(self.dirs.media_dir) / "documents"
+        shutil.rmtree(documents_dir)
+        install_document_samples(documents_dir)
+        # Thumbnails are located by primary key, which the database sequence
+        # does not reset between tests on PostgreSQL
+        for number, document in enumerate(
+            (self.d1, self.d2, self.d3, self.d4),
+            start=1,
+        ):
+            source = documents_dir / "thumbnails" / f"{number:07}.webp"
+            if source != document.thumbnail_path:
+                shutil.copy(source, document.thumbnail_path)
+
     def _get_document_from_manifest(self, manifest, id):
         f = list(
             filter(
@@ -204,8 +218,7 @@ class TestExportImport(
         return manifest
 
     def test_exporter(self, *, use_filename_format=False) -> None:
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         num_permission_objects = Permission.objects.count()
 
@@ -308,8 +321,7 @@ class TestExportImport(
             self.assertEqual(messages.total_issue_count, 0)
 
     def test_exporter_with_filename_format(self) -> None:
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         with override_settings(
             FILENAME_FORMAT="{created_year}/{correspondent}/{title}",
@@ -317,8 +329,7 @@ class TestExportImport(
             self.test_exporter(use_filename_format=True)
 
     def test_exporter_includes_share_links_and_bundles(self) -> None:
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         share_link = ShareLink.objects.create(
             slug="share-link-slug",
@@ -416,8 +427,7 @@ class TestExportImport(
             )
 
     def test_update_export_changed_time(self) -> None:
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         self._do_export()
         self.assertIsFile(self.target / "manifest.json")
@@ -452,8 +462,7 @@ class TestExportImport(
         self.assertEqual(st_mtime_3, st_mtime_4)
 
     def test_update_export_changed_checksum(self) -> None:
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         self._do_export()
 
@@ -479,8 +488,7 @@ class TestExportImport(
         self.assertIsFile(self.target / "manifest.json")
 
     def test_update_export_deleted_document(self) -> None:
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         manifest = self._do_export()
 
@@ -511,8 +519,7 @@ class TestExportImport(
 
     @override_settings(FILENAME_FORMAT="{title}/{correspondent}")
     def test_update_export_changed_location(self) -> None:
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         self._do_export(use_filename_format=True)
         self.assertIsFile(self.target / "wow1" / "c.pdf")
@@ -553,8 +560,7 @@ class TestExportImport(
             - Zipfile is created
             - Zipfile contains exported files
         """
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         args = ["document_exporter", self.target, "--zip"]
 
@@ -582,8 +588,7 @@ class TestExportImport(
             - Zipfile is created
             - Zipfile contains exported files
         """
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         args = ["document_exporter", self.target, "--zip", "--use-filename-format"]
 
@@ -618,8 +623,7 @@ class TestExportImport(
             - Zipfile contains exported files
             - The existing file and directory in target are removed
         """
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         # Create stuff in target directory
         existing_file = self.target / "test.txt"
@@ -714,8 +718,7 @@ class TestExportImport(
             - Manifest.json doesn't contain information about archive files
             - Documents can be imported again
         """
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         manifest = self._do_export()
         has_archive = False
@@ -757,8 +760,7 @@ class TestExportImport(
             - Manifest.json doesn't contain information about thumbnails
             - Documents can be imported again
         """
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         manifest = self._do_export()
         has_thumbnail = False
@@ -802,8 +804,7 @@ class TestExportImport(
             - Main manifest.json file doesn't contain information about documents
             - Documents can be imported again
         """
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         manifest = self._do_export(split_manifest=True)
         has_document = False
@@ -827,8 +828,7 @@ class TestExportImport(
             self.assertEqual(CustomFieldInstance.objects.count(), 1)
 
     def _export_import_barcodes(self, *, split_manifest: bool) -> None:
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
         DocumentBarcodeFactory(document=self.d1, value="https://example.com")
         DocumentBarcodeFactory(document=self.d2, page=2, value="DE8937")
 
@@ -863,8 +863,7 @@ class TestExportImport(
         THEN:
             - Documents can be imported again
         """
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         self._do_export(use_folder_prefix=True)
 
@@ -890,8 +889,7 @@ class TestExportImport(
         THEN:
             - Documents can be imported again
         """
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         self._do_export(use_folder_prefix=True, split_manifest=True)
 
@@ -917,8 +915,7 @@ class TestExportImport(
             - ContentType & Permission objects are not deleted, db transaction rolled back
         """
 
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         num_content_type_objects = ContentType.objects.count()
         num_permission_objects = Permission.objects.count()
@@ -954,8 +951,7 @@ class TestExportImport(
             self.assertEqual(Permission.objects.count(), num_permission_objects + 1)
 
     def test_exporter_with_auditlog_disabled(self) -> None:
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         with override_settings(
             AUDIT_LOG_ENABLED=False,
@@ -974,8 +970,7 @@ class TestExportImport(
             - The soft-deleted document, note, and custom field instance
               survive the round-trip with deleted_at preserved
         """
-        shutil.rmtree(Path(self.dirs.media_dir) / "documents")
-        install_document_samples(Path(self.dirs.media_dir) / "documents")
+        self._install_sample_files()
 
         # d1 has self.note and self.cfi1 attached via setUp
         self.d1.delete()

src/documents/tests/test_management_fuzzy.py

@@ -19,6 +19,10 @@ class TestFuzzyMatchCommand(TestCase):
         call_command(
             "document_fuzzy_match",
             "--no-progress-bar",
+            # The default scales with the CPU count and forks, which closes
+            # the test's database connection on PostgreSQL
+            "--processes",
+            "1",
             *args,
             stdout=stdout,
             stderr=stderr,

src/documents/tests/test_management_retagger.py

@@ -425,26 +425,29 @@ class TestRetaggerStoragePath(DirectoriesMixin):
 @pytest.mark.management
 @pytest.mark.django_db
 class TestRetaggerIdRange(DirectoriesMixin):
-    @pytest.mark.usefixtures("documents")
     @pytest.mark.parametrize(
-        ("id_range_args", "expected_count"),
+        ("range_end", "expected_count"),
         [
-            pytest.param(["1", "2"], 1, id="narrow_range_limits_scope"),
-            pytest.param(["1", "9999"], 2, id="wide_range_tags_all_matches"),
+            pytest.param("second", 1, id="narrow_range_limits_scope"),
+            pytest.param("extra", 2, id="wide_range_tags_all_matches"),
         ],
     )
     def test_id_range_limits_scope(
         self,
         tags: TagTuple,
-        id_range_args: list[str],
+        documents: DocumentTuple,
+        range_end: str,
         expected_count: int,
     ) -> None:
-        DocumentFactory(content="NOT the first document")
+        first, second, *_ = documents
+        extra = DocumentFactory(content="NOT the first document")
+        end = {"second": second, "extra": extra}[range_end]
         call_command(
             "document_retagger",
             "--tags",
             "--id-range",
-            *id_range_args,
+            str(first.pk),
+            str(end.pk),
             skip_checks=True,
         )
         tag_first, *_ = tags

src/documents/tests/test_workflows.py

@@ -174,8 +174,8 @@ class TestWorkflows(
         w.save()
 
         self.assertEqual(w.__str__(), "Workflow: Workflow 1")
-        self.assertEqual(trigger.__str__(), "WorkflowTrigger 1")
-        self.assertEqual(action.__str__(), "WorkflowAction 1")
+        self.assertEqual(trigger.__str__(), f"WorkflowTrigger {trigger.pk}")
+        self.assertEqual(action.__str__(), f"WorkflowAction {action.pk}")
 
         test_file = shutil.copy(
             self.SIMPLE_PDF,
@@ -433,7 +433,7 @@ class TestWorkflows(
                 list(document.tags.all()),
                 [self.t1, self.t2, self.t3],
             )
-            self.assertEqual(
+            self.assertCountEqual(
                 list(
                     get_users_with_perms(
                         document,

src/documents/views.py

@@ -168,6 +168,7 @@ from documents.permissions import PaperlessNotePermissions
 from documents.permissions import PaperlessObjectPermissions
 from documents.permissions import TrashPermissions
 from documents.permissions import ViewDocumentsPermissions
+from documents.permissions import annotate_document_count_by_ids
 from documents.permissions import annotate_document_count_for_related_queryset
 from documents.permissions import get_document_count_filter_for_user
 from documents.permissions import get_permission_grantees
@@ -1022,61 +1023,70 @@ class DocumentViewSet(
     )
 
     def _get_selection_data_for_queryset(self, queryset):
-        # The filtered queryset (user filters plus the permission filter) is
-        # kept as SQL and handed to the database, never resolved to a Python id
-        # list: at hundreds of thousands of documents that list becomes a
-        # multi-MB IN (...) clause per statement, which is slow to plan and, on
-        # SQLite, exceeds the bound-parameter limit outright. Each relation
-        # instead gets one GROUP BY with the filter as a plain WHERE
-        # (foreign keys) or a semi-join subquery (link tables).
-        # order_by() drops the default/user ordering, which is irrelevant for
-        # counting and would otherwise force a sort.
-        documents = queryset.order_by()
-        document_ids = documents.values("pk")
-
-        def foreign_key_counts(field_name: str) -> dict[int, int]:
-            # distinct: filters can join (and so repeat) a document row.
-            return dict(
-                documents.filter(**{f"{field_name}__isnull": False})
-                .values_list(field_name)
-                .annotate(document_count=Count("pk", distinct=True)),
-            )
-
-        def link_table_counts(link_model, field_name: str) -> dict[int, int]:
-            return dict(
-                link_model.objects.filter(document_id__in=document_ids)
-                .values_list(field_name)
-                .annotate(document_count=Count("document_id")),
-            )
-
-        def selection(model, counts: dict[int, int]) -> list[dict[str, int]]:
-            # Every object is listed, with 0 for those with no matching document.
-            return [
-                {"id": pk, "document_count": counts.get(pk, 0)}
-                for pk in model.objects.values_list("pk", flat=True)
-            ]
+        # Resolve once instead of once per model below. `queryset` can carry an
+        # arbitrarily expensive WHERE clause (user filters plus the permission
+        # filter); re-embedding it as a subquery inside 5 separate Count(...)
+        # calls forces the database to re-evaluate that whole thing 5 times, and
+        # -- for FK relations especially -- can defeat semi-join planning
+        # entirely at scale. A concrete id list is cheap to reuse.
+        # order_by() drops the default/user ordering -- irrelevant for a plain
+        # id list, but left in place it forces a sort over the full filtered
+        # set before the ids can even be collected.
+        document_ids = list(queryset.order_by().values_list("pk", flat=True))
 
+        correspondents = Correspondent.objects.annotate(
+            document_count=Count(
+                "documents",
+                filter=Q(documents__id__in=document_ids),
+                distinct=True,
+            ),
+        )
+        document_types = DocumentType.objects.annotate(
+            document_count=Count(
+                "documents",
+                filter=Q(documents__id__in=document_ids),
+                distinct=True,
+            ),
+        )
+        storage_paths = StoragePath.objects.annotate(
+            document_count=Count(
+                "documents",
+                filter=Q(documents__id__in=document_ids),
+                distinct=True,
+            ),
+        )
+        # Tag and CustomField reach Document through an M2M/through-model table;
+        # a plain Count(filter=...) there is a much more expensive plan than the
+        # FK relations above once the bridge table is large -- see
+        # annotate_document_count_by_ids() for why.
+        tags = annotate_document_count_by_ids(
+            Tag.objects.all(),
+            through_model=Document.tags.through,
+            related_object_field="tag_id",
+            document_ids=document_ids,
+        )
+        custom_fields = annotate_document_count_by_ids(
+            CustomField.objects.all(),
+            through_model=CustomFieldInstance,
+            related_object_field="field_id",
+            document_ids=document_ids,
+        )
         return {
-            "selected_correspondents": selection(
-                Correspondent,
-                foreign_key_counts("correspondent_id"),
-            ),
-            "selected_tags": selection(
-                Tag,
-                link_table_counts(Document.tags.through, "tag_id"),
-            ),
-            "selected_document_types": selection(
-                DocumentType,
-                foreign_key_counts("document_type_id"),
-            ),
-            "selected_storage_paths": selection(
-                StoragePath,
-                foreign_key_counts("storage_path_id"),
-            ),
-            "selected_custom_fields": selection(
-                CustomField,
-                link_table_counts(CustomFieldInstance, "field_id"),
-            ),
+            "selected_correspondents": [
+                {"id": t.id, "document_count": t.document_count} for t in correspondents
+            ],
+            "selected_tags": [
+                {"id": t.id, "document_count": t.document_count} for t in tags
+            ],
+            "selected_document_types": [
+                {"id": t.id, "document_count": t.document_count} for t in document_types
+            ],
+            "selected_storage_paths": [
+                {"id": t.id, "document_count": t.document_count} for t in storage_paths
+            ],
+            "selected_custom_fields": [
+                {"id": t.id, "document_count": t.document_count} for t in custom_fields
+            ],
         }
 
     @classmethod

src/paperless_ai/tests/test_chat.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import json
+from contextlib import nullcontext
 from typing import TYPE_CHECKING
 from typing import Any
 from unittest.mock import MagicMock
@@ -37,6 +38,13 @@ def patch_embed_model():
     llama_settings.Settings.embed_model = None
 
 
+@pytest.fixture(autouse=True)
+def keep_db_connection():
+    # Closing the connection would roll back the test's transaction on PostgreSQL
+    with patch("paperless_ai.chat.db_connection_released", nullcontext):
+        yield
+
+
 @pytest.fixture(autouse=True)
 def patch_embed_nodes():
     with patch(