Assert permission assignment does not scale with selection size

The two batching tests only checked that permissions came out correct at
5 and 50 objects, so reverting to the old per-object loop would still
have passed.  Check sizes as well to prevent that

src/documents/bulk_edit.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import logging
 import tempfile
+import uuid
 from pathlib import Path
 from typing import TYPE_CHECKING
 from typing import Literal
@@ -27,7 +28,7 @@ from documents.models import DocumentType
 from documents.models import PaperlessTask
 from documents.models import StoragePath
 from documents.models import Tag
-from documents.permissions import set_permissions_for_object
+from documents.permissions import set_permissions_for_objects
 from documents.plugins.helpers import DocumentsStatusManager
 from documents.tasks import bulk_update_documents
 from documents.tasks import consume_file
@@ -379,7 +380,7 @@ def delete(doc_ids: list[int]) -> Literal["OK"]:
         )
         delete_ids = list({*doc_ids, *version_ids})
 
-        Document.objects.filter(id__in=delete_ids).delete()
+        Document.objects.filter(id__in=delete_ids).delete(transaction_id=uuid.uuid4())
 
         from documents.search import get_backend
 
@@ -430,10 +431,13 @@ def set_permissions(
     else:
         qs.update(owner=owner)
 
-    for doc in qs:
-        set_permissions_for_object(permissions=set_permissions, object=doc, merge=merge)
-
     affected_docs = list(qs.values_list("pk", flat=True))
+    set_permissions_for_objects(
+        permissions=set_permissions,
+        model=Document,
+        pks=affected_docs,
+        merge=merge,
+    )
 
     bulk_update_documents.apply_async(
         kwargs={"document_ids": affected_docs},

src/documents/management/commands/document_consumer.py

@@ -156,6 +156,15 @@ class FileStabilityTracker:
                     logger.debug(f"File disappeared during stability check: {path}")
                 continue
 
+            # Stable, but empty: some scanners create a zero byte placeholder
+            # and only write the page some time later. Consuming it now can
+            # only fail  so drop it and let the writer's next event
+            # (or the periodic rescan) bring it back once it has content
+            if not tracked.last_size:
+                to_remove.append(path)
+                logger.debug("Ignoring stable but empty file: %s", path)
+                continue
+
             # File is stable, we can return it
             to_yield.append(path)
             logger.info(f"File is stable: {path}")

src/documents/models.py

@@ -1,4 +1,5 @@
 import datetime
+import uuid
 from pathlib import Path
 from typing import Final
 
@@ -514,13 +515,20 @@ class Document(SoftDeleteModel, ModelWithOwner):  # type: ignore[django-manager-
     def delete(
         self,
         *args,
+        transaction_id=None,
         **kwargs,
     ):
-        # If deleting a root document, move all its versions to trash as well.
+        # Versions must share the root's transaction ID so they are restored
+        # together by django-softdelete.
+        if transaction_id is None:
+            transaction_id = uuid.uuid4()
         if self.root_document_id is None:
-            Document.objects.filter(root_document=self).delete()
+            Document.objects.filter(root_document=self).delete(
+                transaction_id=transaction_id,
+            )
         return super().delete(
             *args,
+            transaction_id=transaction_id,
             **kwargs,
         )
 

src/documents/permissions.py

@@ -173,6 +173,179 @@ def set_permissions_for_object(
                         )
 
 
+def _resolve_permissions(codenames: set[str], ctype: ContentType) -> list[Permission]:
+    """
+    Resolves `codenames` to Permission rows, raising like the single-object
+    assign_perm() this bulk path replaces does (via a `.get()` internally)
+    if any codename doesn't exist -- e.g. a client-supplied action name that
+    was never validated (BulkEditObjectsSerializer._validate_permissions
+    calls validate_set_permissions() only for its side-effecting id checks
+    and discards the filtered dict it returns, so an unrecognized action key
+    reaches this function as-is). A plain `.filter()` with no existence
+    check would otherwise silently build zero rows and no-op instead of
+    reporting the bad input.
+    """
+    permission_objs = list(
+        Permission.objects.filter(content_type=ctype, codename__in=codenames),
+    )
+    missing = codenames - {p.codename for p in permission_objs}
+    if missing:
+        raise Permission.DoesNotExist(
+            f"Permission matching query does not exist for codename(s): "
+            f"{', '.join(sorted(missing))}",
+        )
+    return permission_objs
+
+
+def _apply_bulk_permission_entry(
+    *,
+    perm_model: type[UserObjectPermission] | type[GroupObjectPermission],
+    identity_model: type[User] | type[Group],
+    identity_field: str,
+    ids: list[int],
+    codename: str,
+    permission_objs: list[Permission],
+    ctype: ContentType,
+    object_pks: list[str],
+    merge: bool,
+) -> None:
+    # Only the ids are needed to build permission rows (via `<field>_id=`),
+    # so avoid fetching full User/Group rows for identities that may not
+    # even end up being granted anything new.
+    add_ids = set(
+        identity_model.objects.filter(id__in=ids).values_list("id", flat=True),
+    )
+
+    if not merge:
+        existing_ids = set(
+            perm_model.objects.filter(
+                content_type=ctype,
+                object_pk__in=object_pks,
+                permission__codename=codename,
+            )
+            .values_list(f"{identity_field}_id", flat=True)
+            .distinct(),
+        )
+        remove_ids = existing_ids - add_ids
+        if remove_ids:
+            perm_model.objects.filter(
+                content_type=ctype,
+                object_pk__in=object_pks,
+                permission__codename=codename,
+                **{f"{identity_field}_id__in": remove_ids},
+            ).delete()
+
+    if not add_ids:
+        return
+
+    rows = [
+        perm_model(
+            content_type=ctype,
+            object_pk=pk,
+            permission=permission_obj,
+            **{f"{identity_field}_id": identity_id},
+        )
+        for permission_obj in permission_objs
+        for pk in object_pks
+        for identity_id in add_ids
+    ]
+    # ignore_conflicts skips only rows that already exist as an exact
+    # (identity, permission, object) match -- the same de-dup the
+    # underlying (user|group, permission, object_pk) unique constraint
+    # already enforces for the single-object assign_perm() this replaces,
+    # so it doesn't change what counts as "already granted". batch_size
+    # caps how many rows go into a single INSERT statement.
+    perm_model.objects.bulk_create(rows, ignore_conflicts=True, batch_size=1000)
+
+
+def set_permissions_for_objects(
+    permissions: dict,
+    model: type[Model],
+    pks: QuerySet | list,
+    *,
+    merge: bool = False,
+) -> None:
+    """
+    Bulk equivalent of set_permissions_for_object: applies the same
+    permission changes to every object identified by `pks` at once.
+
+    Takes a model + pks (rather than model instances) deliberately -- the
+    permission rows built below only ever need `pk`, `content_type`, and
+    identity ids, so callers shouldn't have to fetch full rows (with every
+    other field) just to hand them to this function.
+
+    Deliberately does not use guardian's queryset/list-aware assign_perm:
+    passing a list as the object routes to bulk_assign_perm, which skips
+    creating a direct permission row for anyone who already has the
+    permission via ANY group membership (it checks
+    ObjectPermissionChecker.has_perm, which is group-inheritance-aware) --
+    unlike the single-object assign_perm this replaces, which always
+    ensures a direct row via get_or_create regardless of group-derived
+    access. Losing that guarantee would mean a later revocation of the
+    group's grant silently strips access an admin explicitly asked to be
+    direct. Bulk-creating rows straight against the permission models
+    instead (see _apply_bulk_permission_entry) preserves the original
+    always-create-a-direct-row semantics while still batching every object
+    and every identity into one query per action, rather than one query per
+    (object, user) pair.
+    """
+    object_pks = [str(pk) for pk in pks]
+    if not object_pks:  # pragma: no cover
+        return
+
+    model_name = model.__name__.lower()
+    ctype = ContentType.objects.get_for_model(model)
+
+    # Every action is resolved up front, before anything is written, so an
+    # unrecognized action name (see _resolve_permissions) aborts the whole
+    # call instead of leaving the actions ahead of it already applied --
+    # BulkEditObjectsSerializer lets unknown keys through and its view turns
+    # the exception into a 400, so a half-applied change would otherwise be
+    # reported to the client as a failure.
+    permissions_by_action: dict[str, list[Permission]] = {}
+    for action, entry in permissions.items():
+        if "users" not in entry and "groups" not in entry:
+            continue
+        implied_codenames = {f"{action}_{model_name}"}
+        if action == "change":
+            # change gives view too
+            implied_codenames.add(f"view_{model_name}")
+        permissions_by_action[action] = _resolve_permissions(
+            implied_codenames,
+            ctype,
+        )
+
+    for action, entry in permissions.items():
+        codename = f"{action}_{model_name}"
+        permission_objs = permissions_by_action.get(action, [])
+
+        if "users" in entry:
+            _apply_bulk_permission_entry(
+                perm_model=UserObjectPermission,
+                identity_model=User,
+                identity_field="user",
+                ids=entry["users"],
+                codename=codename,
+                permission_objs=permission_objs,
+                ctype=ctype,
+                object_pks=object_pks,
+                merge=merge,
+            )
+
+        if "groups" in entry:
+            _apply_bulk_permission_entry(
+                perm_model=GroupObjectPermission,
+                identity_model=Group,
+                identity_field="group",
+                ids=entry["groups"],
+                codename=codename,
+                permission_objs=permission_objs,
+                ctype=ctype,
+                object_pks=object_pks,
+                merge=merge,
+            )
+
+
 def permitted_object_ids(
     user: User | None,
     model: type[Model],

src/documents/tests/test_api_objects.py

@@ -2,10 +2,15 @@ import datetime
 import json
 from unittest import mock
 
+from django.contrib.auth.models import Group
 from django.contrib.auth.models import Permission
 from django.contrib.auth.models import User
+from django.db import connection
 from django.test import override_settings
+from django.test.utils import CaptureQueriesContext
 from guardian.shortcuts import assign_perm
+from guardian.shortcuts import get_groups_with_perms
+from guardian.shortcuts import get_users_with_perms
 from rest_framework import status
 from rest_framework.test import APITestCase
 
@@ -842,6 +847,66 @@ class TestBulkEditObjects(APITestCase):
         self.assertEqual(response.status_code, status.HTTP_200_OK)
         self.assertEqual(StoragePath.objects.count(), 0)
 
+    def test_bulk_objects_set_permissions_batched_across_object_count(
+        self,
+    ) -> None:
+        """
+        GIVEN:
+            - Many tags are being bulk-edited to set permissions at once
+        WHEN:
+            - bulk_edit_objects API endpoint is called with set_permissions
+              operation over a small batch vs. a much larger one
+        THEN:
+            - Permissions are applied correctly at both scales
+            - Query count does not grow with the number of tags, i.e. each
+              user/group is applied across all tags with one batched call
+              rather than one call per (tag, identity) pair
+        """
+        group1 = Group.objects.create(name="perm-group")
+        permissions = {
+            "view": {"users": [self.user1.id, self.user2.id], "groups": [group1.id]},
+            "change": {"users": [self.user1.id], "groups": [group1.id]},
+        }
+
+        def run_with_n_tags(n: int) -> int:
+            tags = [Tag.objects.create(name=f"perm-tag-{n}-{i}") for i in range(n)]
+            with CaptureQueriesContext(connection) as ctx:
+                response = self.client.post(
+                    "/api/bulk_edit_objects/",
+                    json.dumps(
+                        {
+                            "objects": [t.id for t in tags],
+                            "object_type": "tags",
+                            "operation": "set_permissions",
+                            "permissions": permissions,
+                            "merge": False,
+                        },
+                    ),
+                    content_type="application/json",
+                )
+            self.assertEqual(response.status_code, status.HTTP_200_OK)
+            for tag in tags:
+                self.assertEqual(get_users_with_perms(tag).count(), 2)
+                self.assertEqual(get_groups_with_perms(tag).count(), 1)
+            return len(ctx.captured_queries)
+
+        small_batch_queries = run_with_n_tags(5)
+        large_batch_queries = run_with_n_tags(50)
+
+        # A tolerance rather than equality, matching the N+1 check in
+        # test_views.py: bulk_create's batch_size caps rows per INSERT, so a
+        # large enough selection does legitimately add statements, and the
+        # per-process ContentType cache makes the first run carry an extra
+        # query. Neither can hide a regression to per-object assignment,
+        # which would be ~10x the small-batch count here.
+        self.assertLessEqual(
+            large_batch_queries,
+            small_batch_queries + 5,
+            "Permission assignment appears to scale with object count: "
+            f"{small_batch_queries} queries for 5 tags vs. "
+            f"{large_batch_queries} for 50",
+        )
+
     def test_bulk_objects_delete_all_filtered(self) -> None:
         """
         GIVEN:

src/documents/tests/test_api_trash.py

@@ -207,3 +207,65 @@ class TestTrashAPI(DirectoriesMixin, APITestCase):
         )
         self.assertEqual(resp.status_code, status.HTTP_400_BAD_REQUEST)
         self.assertIn("have not yet been deleted", resp.data["documents"][0])
+
+    def _make_versioned_document(self) -> tuple[Document, list[Document]]:
+        root = Document.objects.create(
+            title="root",
+            content="root-content",
+            checksum="root",
+            mime_type="application/pdf",
+        )
+        versions = [
+            Document.objects.create(
+                title=f"v{index}",
+                content=f"v{index}-content",
+                checksum=f"v{index}",
+                mime_type="application/pdf",
+                root_document=root,
+                version_index=index,
+            )
+            for index in range(1, 3)
+        ]
+        return root, versions
+
+    def test_api_trash_restore_document_restores_its_versions(self) -> None:
+        """
+        GIVEN:
+            - Existing document with two versions
+        WHEN:
+            - API request to delete the document
+            - API request to restore it from the trash
+        THEN:
+            - Only the document itself is listed in the trash
+            - A version cannot be restored without its root
+            - The document is restored together with all of its versions
+        """
+        root, versions = self._make_versioned_document()
+
+        self.client.force_login(user=self.user)
+        self.client.delete(f"/api/documents/{root.pk}/")
+        self.assertEqual(Document.deleted_objects.count(), 3)
+
+        resp = self.client.get("/api/trash/")
+        self.assertEqual(resp.status_code, status.HTTP_200_OK)
+        self.assertEqual(resp.data["count"], 1)
+        self.assertEqual(resp.data["results"][0]["id"], root.pk)
+
+        # A version cannot be restored while its root remains in the trash.
+        resp = self.client.post(
+            "/api/trash/",
+            {"action": "restore", "documents": [versions[0].pk]},
+        )
+        self.assertEqual(resp.status_code, status.HTTP_400_BAD_REQUEST)
+        self.assertIn("Restore the root document", resp.data["documents"][0])
+
+        resp = self.client.post(
+            "/api/trash/",
+            {"action": "restore", "documents": [root.pk]},
+        )
+        self.assertEqual(resp.status_code, status.HTTP_200_OK)
+        self.assertEqual(Document.deleted_objects.count(), 0)
+        self.assertCountEqual(
+            Document.objects.filter(root_document=root).values_list("id", flat=True),
+            [version.pk for version in versions],
+        )

src/documents/tests/test_bulk_edit.py

@@ -5,8 +5,11 @@ from unittest import mock
 
 import pikepdf
 from django.contrib.auth.models import Group
+from django.contrib.auth.models import Permission
 from django.contrib.auth.models import User
+from django.db import connection
 from django.test import TestCase
+from django.test.utils import CaptureQueriesContext
 from guardian.shortcuts import assign_perm
 from guardian.shortcuts import get_groups_with_perms
 from guardian.shortcuts import get_users_with_perms
@@ -19,6 +22,7 @@ from documents.models import Document
 from documents.models import DocumentType
 from documents.models import StoragePath
 from documents.models import Tag
+from documents.permissions import set_permissions_for_objects
 from documents.tests.utils import DirectoriesMixin
 
 
@@ -392,6 +396,11 @@ class TestBulkEdit(DirectoriesMixin, TestCase):
         self.assertFalse(Document.objects.filter(id=self.doc1.id).exists())
         self.assertFalse(Document.objects.filter(id=version.id).exists())
 
+        Document.deleted_objects.get(id=self.doc1.id).restore(strict=False)
+
+        self.assertTrue(Document.objects.filter(id=self.doc1.id).exists())
+        self.assertTrue(Document.objects.filter(id=version.id).exists())
+
     def test_delete_version_document_keeps_root(self) -> None:
         version = Document.objects.create(
             checksum="A-v1",
@@ -510,6 +519,178 @@ class TestBulkEdit(DirectoriesMixin, TestCase):
         )
         self.assertEqual(groups_with_perms.count(), 2)
 
+    @mock.patch("documents.tasks.bulk_update_documents.apply_async")
+    def test_set_permissions_batched_across_document_count(
+        self,
+        m,
+    ) -> None:
+        """
+        GIVEN:
+            - Many documents are being bulk-edited to set permissions at once
+        WHEN:
+            - set_permissions runs over a small batch vs. a much larger one
+        THEN:
+            - Permissions are applied correctly at both scales
+            - Query count does not grow with the number of documents, i.e.
+              each user/group is applied across all documents with one
+              batched call rather than one call per (document, identity)
+              pair
+        """
+        permissions = {
+            "view": {
+                "users": [self.user1.id, self.user2.id],
+                "groups": [self.group2.id],
+            },
+            "change": {
+                "users": [self.user1.id],
+                "groups": [self.group2.id],
+            },
+        }
+
+        def run_with_n_documents(n: int) -> int:
+            docs = [
+                Document.objects.create(checksum=f"perm-{n}-{i}", title=f"perm-{n}-{i}")
+                for i in range(n)
+            ]
+            with CaptureQueriesContext(connection) as ctx:
+                bulk_edit.set_permissions(
+                    [doc.id for doc in docs],
+                    set_permissions=permissions,
+                    owner=self.owner,
+                    merge=False,
+                )
+            for doc in docs:
+                self.assertEqual(get_users_with_perms(doc).count(), 2)
+                self.assertEqual(get_groups_with_perms(doc).count(), 1)
+            return len(ctx.captured_queries)
+
+        small_batch_queries = run_with_n_documents(5)
+        large_batch_queries = run_with_n_documents(50)
+
+        # A tolerance rather than equality, matching the N+1 check in
+        # test_views.py: bulk_create's batch_size caps rows per INSERT, so a
+        # large enough selection does legitimately add statements, and the
+        # per-process ContentType cache makes the first run carry an extra
+        # query. Neither can hide a regression to per-document assignment,
+        # which would be ~10x the small-batch count here.
+        self.assertLessEqual(
+            large_batch_queries,
+            small_batch_queries + 5,
+            "Permission assignment appears to scale with document count: "
+            f"{small_batch_queries} queries for 5 documents vs. "
+            f"{large_batch_queries} for 50",
+        )
+
+    @mock.patch("documents.tasks.bulk_update_documents.apply_async")
+    def test_set_permissions_grants_direct_perm_even_if_already_granted_via_group(
+        self,
+        m,
+    ) -> None:
+        """
+        GIVEN:
+            - A user already has view access to a document via group
+              membership, with no direct grant of their own
+        WHEN:
+            - set_permissions explicitly grants that same user direct view
+              access via bulk_edit
+        THEN:
+            - A direct permission grant is created for the user, not skipped
+              because they already have equivalent access via the group
+
+        Regression test: guardian's queryset-aware assign_perm() (routed to
+        when the target is a list/queryset) skips creating a direct row for
+        anyone whose ObjectPermissionChecker.has_perm() already returns True
+        -- which includes group-derived access. The single-object assign_perm
+        this bulk path replaces has no such check; it always ensures a
+        direct row via get_or_create. Losing that guarantee would mean
+        revoking the group's grant later silently strips access that was
+        supposed to be explicit.
+        """
+        self.doc1.owner = self.user1
+        self.doc1.save()
+        self.user1.groups.add(self.group1)
+        assign_perm("view_document", self.group1, self.doc1)
+
+        bulk_edit.set_permissions(
+            [self.doc1.id],
+            set_permissions={
+                "view": {"users": [self.user1.id], "groups": []},
+            },
+            merge=True,
+        )
+
+        direct_users = get_users_with_perms(
+            self.doc1,
+            only_with_perms_in=["view_document"],
+            with_group_users=False,
+        )
+        self.assertIn(self.user1, direct_users)
+
+    def test_set_permissions_for_objects_raises_for_unknown_action(self) -> None:
+        """
+        GIVEN:
+            - An unrecognized permission action name with users to grant it
+              to
+        WHEN:
+            - set_permissions_for_objects is called
+        THEN:
+            - Permission.DoesNotExist is raised, not a silent no-op
+
+        Regression test: the endpoint that calls this
+        (BulkEditObjectPermissionsView) never actually validates action
+        names against the raw client-supplied permissions dict --
+        BulkEditObjectsSerializer._validate_permissions calls
+        validate_set_permissions() only for its side-effecting user/group id
+        checks and discards the filtered dict it returns -- so a bogus
+        action key reaches this function as-is. Resolving the Permission via
+        a bare `.filter()` (which returns empty instead of raising) would
+        silently drop the grant and report success.
+        """
+        with self.assertRaises(Permission.DoesNotExist):
+            set_permissions_for_objects(
+                {"not_a_real_action": {"users": [self.user1.id], "groups": []}},
+                Document,
+                [self.doc1.pk],
+            )
+
+    def test_set_permissions_for_objects_unknown_action_applies_nothing(
+        self,
+    ) -> None:
+        """
+        GIVEN:
+            - A permissions dict with a valid action ordered ahead of an
+              unrecognized one
+        WHEN:
+            - set_permissions_for_objects is called
+        THEN:
+            - Permission.DoesNotExist is raised
+            - The valid action ahead of it is not applied either
+
+        Every action is resolved before any row is written, so a bad action
+        name cannot leave a half-applied change behind. That matters because
+        BulkEditObjectsView turns this exception into a 400: without the
+        up-front resolution the client would be told the request failed
+        while the leading action had already been committed.
+        """
+        with self.assertRaises(Permission.DoesNotExist):
+            set_permissions_for_objects(
+                {
+                    "view": {"users": [self.user1.id], "groups": []},
+                    "not_a_real_action": {"users": [self.user1.id], "groups": []},
+                },
+                Document,
+                [self.doc1.pk],
+            )
+
+        self.assertNotIn(
+            self.user1,
+            get_users_with_perms(
+                self.doc1,
+                only_with_perms_in=["view_document"],
+                with_group_users=False,
+            ),
+        )
+
     @mock.patch("documents.models.Document.delete")
     def test_delete_documents_old_uuid_field(self, m) -> None:
         m.side_effect = Exception("Data too long for column 'transaction_id' at row 1")

src/documents/tests/test_document_model.py

@@ -110,7 +110,7 @@ class TestDocument(TestCase):
             checksum="checksum",
             mime_type="application/pdf",
         )
-        Document.objects.create(
+        version = Document.objects.create(
             root_document=root,
             correspondent=root.correspondent,
             title="Version",
@@ -124,6 +124,10 @@ class TestDocument(TestCase):
         self.assertEqual(Document.objects.count(), 0)
         self.assertEqual(Document.deleted_objects.count(), 2)
 
+        root.restore(strict=False)
+
+        self.assertTrue(Document.objects.filter(pk=version.pk).exists())
+
     def test_file_name(self) -> None:
         doc = Document(
             mime_type="application/pdf",

src/documents/tests/test_management_consumer.py

@@ -136,6 +136,23 @@ def wait_for_mock_call(
     return False
 
 
+def sleep_past_stability(
+    owner: FileStabilityTracker | ConsumerThread,
+    *,
+    windows: float = 1.5,
+) -> None:
+    """
+    Block until a tracked file's stability window has certainly elapsed.
+
+    Args:
+        owner: The tracker, or the consumer thread running one, whose
+               configured stability delay sets the wait.
+        windows: How many stability windows to wait, giving slop for a slow
+                 or loaded test runner.
+    """
+    sleep(owner.stability_delay * windows)
+
+
 class TestTrackedFile:
     """Tests for the TrackedFile dataclass."""
 
@@ -261,6 +278,56 @@ class TestFileStabilityTracker:
         assert len(stable) == 0
         assert stability_tracker.pending_count == 1
 
+    def test_get_stable_files_skips_empty_file(
+        self,
+        stability_tracker: FileStabilityTracker,
+        tmp_path: Path,
+    ) -> None:
+        """
+        GIVEN:
+            - A zero byte file, tracked and past its stability delay
+        WHEN:
+            - Stable files are collected
+        THEN:
+            - The file is not yielded for consumption
+            - The file is dropped from tracking rather than held, so an
+              abandoned placeholder does not keep the watch loop awake
+        """
+        empty = tmp_path / "scan.pdf"
+        empty.write_bytes(b"")
+        stability_tracker.track(empty, Change.added)
+        sleep_past_stability(stability_tracker)
+
+        stable = list(stability_tracker.get_stable_files())
+
+        assert stable == []
+        assert stability_tracker.pending_count == 0
+
+    def test_empty_file_is_yielded_once_content_arrives(
+        self,
+        stability_tracker: FileStabilityTracker,
+        tmp_path: Path,
+    ) -> None:
+        """
+        GIVEN:
+            - A zero byte file which was dropped from tracking while empty
+        WHEN:
+            - The writer fills the file and a new event re-tracks it
+        THEN:
+            - The file is yielded for consumption once it is stable
+        """
+        target = tmp_path / "scan.pdf"
+        target.write_bytes(b"")
+        stability_tracker.track(target, Change.added)
+        sleep_past_stability(stability_tracker)
+        assert list(stability_tracker.get_stable_files()) == []
+
+        target.write_bytes(b"%PDF-1.4 content")
+        stability_tracker.track(target, Change.modified)
+        sleep_past_stability(stability_tracker)
+
+        assert list(stability_tracker.get_stable_files()) == [target]
+
     def test_get_stable_files_deleted_during_check(self, temp_file: Path) -> None:
         """Test deleted file is not returned during stability check."""
         tracker = FileStabilityTracker(stability_delay=0.1)
@@ -879,6 +946,51 @@ class TestCommandWatch:
 
         mock_consume_file_delay.apply_async.assert_called()
 
+    def test_scanner_placeholder_is_not_consumed_while_empty(
+        self,
+        consumption_dir: Path,
+        sample_pdf: Path,
+        mock_consume_file_delay: MagicMock,
+        start_consumer: Callable[..., ConsumerThread],
+    ) -> None:
+        """
+        GIVEN:
+            - A scanner which creates a zero byte placeholder and only writes
+              the page some time later (GH discussion #13969)
+        WHEN:
+            - The placeholder sits untouched well past the stability delay
+            - The scanner then writes the real content
+        THEN:
+            - The empty placeholder is never queued, as it could only fail
+              with "Unsupported mime type inode/x-empty"
+            - The file is queued exactly once, when the content lands
+        """
+        thread = start_consumer(stability_delay=0.2)
+
+        target = consumption_dir / "scan.pdf"
+        target.write_bytes(b"")  # the scanner's placeholder
+
+        # Well past the stability delay: the old behaviour queued it here.
+        sleep_past_stability(thread, windows=5)
+        if thread.exception:
+            raise thread.exception
+        assert mock_consume_file_delay.apply_async.call_count == 0
+
+        shutil.copy(sample_pdf, target)  # the scanner finishes the page
+
+        assert wait_for_mock_call(
+            mock_consume_file_delay.apply_async,
+            timeout_s=5.0,
+        )
+        if thread.exception:
+            raise thread.exception
+
+        assert mock_consume_file_delay.apply_async.call_count == 1
+        queued_doc = mock_consume_file_delay.apply_async.call_args.kwargs["kwargs"][
+            "input_doc"
+        ]
+        assert queued_doc.original_file.name == "scan.pdf"
+
     def test_ignores_macos_files(
         self,
         consumption_dir: Path,

src/documents/tests/test_views.py

@@ -32,6 +32,7 @@ from documents.signals.handlers import update_llm_suggestions_cache
 from documents.tests.utils import DirectoriesMixin
 from documents.tests.utils import read_streaming_response
 from paperless.models import ApplicationConfiguration
+from paperless_ai.exceptions import LLMProviderError
 from paperless_ai.exceptions import LLMTimeoutError
 
 
@@ -737,6 +738,38 @@ class TestAISuggestions(DirectoriesMixin, TestCase):
             get_llm_suggestion_cache(self.document.pk, backend="openai-like"),
         )
 
+    @patch("documents.views.get_ai_document_classification")
+    @override_settings(
+        AI_ENABLED=True,
+        LLM_BACKEND="openai-like",
+    )
+    def test_ai_suggestions_with_llm_provider_error(
+        self,
+        mock_get_ai_classification,
+    ) -> None:
+        mock_get_ai_classification.side_effect = LLMProviderError(
+            "confidential provider response",
+        )
+
+        self.client.force_login(user=self.user)
+        response = self.client.get(
+            f"/api/documents/{self.document.pk}/ai_suggestions/",
+        )
+
+        self.assertEqual(response.status_code, status.HTTP_502_BAD_GATEWAY)
+        self.assertEqual(
+            response.json(),
+            {
+                "ai": [
+                    "AI backend rejected the request. Check logs for details.",
+                ],
+            },
+        )
+        self.assertNotIn("confidential provider response", response.content.decode())
+        self.assertIsNone(
+            get_llm_suggestion_cache(self.document.pk, backend="openai-like"),
+        )
+
     @patch("documents.views.get_ai_document_classification")
     @override_settings(
         AI_ENABLED=True,

src/documents/views.py

@@ -179,7 +179,7 @@ from documents.permissions import has_perms_owner_aware
 from documents.permissions import has_system_status_permission
 from documents.permissions import permitted_document_ids
 from documents.permissions import permitted_object_ids
-from documents.permissions import set_permissions_for_object
+from documents.permissions import set_permissions_for_objects
 from documents.permissions import user_is_unrestricted
 from documents.plugins.date_parsing import get_date_parser
 from documents.schema import generate_object_with_permissions_schema
@@ -252,6 +252,7 @@ from paperless.views import StandardPagination
 from paperless_ai.ai_classifier import get_ai_document_classification
 from paperless_ai.ai_classifier import get_llm_output_language
 from paperless_ai.chat import stream_chat_with_documents
+from paperless_ai.exceptions import LLMProviderError
 from paperless_ai.exceptions import LLMTimeoutError
 from paperless_ai.matching import extract_unmatched_names
 from paperless_ai.matching import match_correspondents_by_name
@@ -1603,6 +1604,22 @@ class DocumentViewSet(
                     {"ai": [_("AI backend request timed out.")]},
                     status=status.HTTP_503_SERVICE_UNAVAILABLE,
                 )
+            except LLMProviderError:
+                logger.exception(
+                    "AI backend rejected the request for document %s",
+                    doc.pk,
+                )
+                return Response(
+                    {
+                        "ai": [
+                            _(
+                                "AI backend rejected the request. "
+                                "Check logs for details.",
+                            ),
+                        ],
+                    },
+                    status=status.HTTP_502_BAD_GATEWAY,
+                )
             set_llm_suggestions_cache(
                 doc.pk,
                 llm_suggestions,
@@ -4950,10 +4967,10 @@ class BulkEditObjectsView(PassUserMixin):
                     qs_owner_update.update(owner=owner)
 
                 if "permissions" in serializer.validated_data:
-                    for obj in qs:
+                    set_permissions_for_objects(
-                        set_permissions_for_object(
                         permissions=permissions,
-                            object=obj,
+                        model=object_class,
+                        pks=qs.values_list("pk", flat=True),
                         merge=merge,
                     )
 
@@ -5437,7 +5454,10 @@ class TrashView(ListModelMixin, PassUserMixin):
 
     model = Document
 
-    queryset = Document.deleted_objects.all()
+    # A version is listed separately only when its root is not in the trash.
+    queryset = Document.deleted_objects.exclude(
+        root_document_id__in=Document.deleted_objects.values("id"),
+    )
 
     def get(self, request: Request, format: str | None = None) -> Response:
         self.serializer_class = DocumentSerializer
@@ -5468,7 +5488,15 @@ class TrashView(ListModelMixin, PassUserMixin):
             return HttpResponseForbidden("Insufficient permissions")
         action = serializer.validated_data.get("action")
         if action == "restore":
-            restored = list(Document.deleted_objects.filter(id__in=doc_ids))
+            restored = list(self.get_queryset().filter(id__in=doc_ids))
+            if len(restored) != len(doc_ids):
+                raise ValidationError(
+                    {
+                        "documents": [
+                            "Restore the root document instead of one of its versions.",
+                        ],
+                    },
+                )
             for doc in restored:
                 doc.restore(strict=False)
             if restored:

src/paperless_ai/ai_classifier.py

@@ -4,21 +4,24 @@ from django.conf import settings
 from django.contrib.auth.models import User
 
 from documents.models import Document
-from documents.permissions import get_objects_for_user_owner_aware
+from documents.permissions import permitted_object_ids
+from documents.permissions import restrict_queryset_to_visible
+from documents.permissions import user_is_unrestricted
 from paperless.config import AIConfig
 from paperless_ai.base_model import ClassificationSuggestions
 from paperless_ai.base_model import TaxonomyChoiceDict
 from paperless_ai.base_model import classification_suggestions_to_model
 from paperless_ai.client import AIClient
 from paperless_ai.db import db_connection_released
-from paperless_ai.indexing import _node_document_ids
 from paperless_ai.indexing import retrieve_similar_nodes
 from paperless_ai.indexing import truncate_content
 from paperless_ai.prompts.context import ClassificationPromptContext
 from paperless_ai.prompts.context import LocalizationPromptContext
 from paperless_ai.prompts.context import RagContextPromptContext
 from paperless_ai.prompts.render import render_prompt
+from paperless_ai.taxonomy import SimilarDocument
 from paperless_ai.taxonomy import TaxonomyCandidates
+from paperless_ai.taxonomy import _node_document_weights
 from paperless_ai.taxonomy import build_taxonomy_candidates
 from paperless_ai.taxonomy import empty_taxonomy_candidates
 from paperless_ai.taxonomy import format_taxonomy_for_prompt
@@ -37,6 +40,48 @@ logger = logging.getLogger("paperless_ai.rag_classifier")
 TAXONOMY_CANDIDATE_TOP_K = 15
 
 
+def _fulltext_similar_documents(
+    document: Document,
+    user: User | None,
+    top_k: int,
+) -> list[SimilarDocument]:
+    """Rank-based fallback when no embedding backend is configured. Uses
+    Tantivy's "More Like This" (term-overlap similarity) instead of vector
+    similarity - cruder, but far better than no candidates at all.
+    more_like_this_ids returns only a ranked ID list, no scores, so weight is
+    synthesized from rank (descending from top_k) rather than claiming a
+    similarity magnitude that doesn't exist. An unrestricted user (none, or an
+    active superuser - see user_is_unrestricted) is normalized to ``None``
+    before calling, since the backend's permission filter has no superuser
+    short-circuit of its own. Results are re-checked with
+    restrict_queryset_to_visible() since Tantivy's indexed permission fields
+    lag the DB via async reindexing.
+    """
+    from documents.search import get_backend
+
+    unrestricted = user_is_unrestricted(user)
+    search_user = None if unrestricted else user
+    backend = get_backend()
+    similar_ids = backend.more_like_this_ids(
+        document.pk,
+        user=search_user,
+        limit=top_k,
+    )
+    if not unrestricted:
+        allowed_ids = set(
+            restrict_queryset_to_visible(
+                Document.objects.filter(pk__in=similar_ids),
+                user,
+                "view_document",
+            ).values_list("pk", flat=True),
+        )
+        similar_ids = [doc_id for doc_id in similar_ids if doc_id in allowed_ids]
+    return [
+        SimilarDocument(document_id=doc_id, weight=float(top_k - rank))
+        for rank, doc_id in enumerate(similar_ids)
+    ]
+
+
 def get_language_name(language_code: str) -> str:
     normalized_language_code = language_code.lower()
     for code, name in settings.LANGUAGES:
@@ -136,43 +181,52 @@ def get_taxonomy_context(
     user: User | None = None,
     max_docs: int = 5,
 ) -> tuple[TaxonomyCandidates, str]:
-    """One retrieval feeds both taxonomy candidates and RAG text context.
+    """One retrieval feeds both taxonomy candidates and RAG text context. Uses
-    On any retrieval failure, degrades to empty candidates/context rather than
+    vector similarity when an embedding backend is configured, otherwise
-    propagating the exception - a vector-store outage should not block
+    falls back to Tantivy full-text "More Like This" similarity - see
-    classification, only its RAG-assisted enrichment.
+    _fulltext_similar_documents. On any retrieval failure, degrades to empty
+    candidates/context rather than propagating the exception - neither a
+    vector-store outage nor a search-index issue should block classification,
+    only its context-assisted enrichment.
     """
+    ai_config = AIConfig()
     try:
-        # None means "no restriction" to retrieve_similar_nodes. A superuser
+        if ai_config.llm_embedding_backend:
-        # (like no user at all) can see every document, so skip materializing
+            # None means "no restriction" to retrieve_similar_nodes. An
-        # every visible pk into a Python list and passing it through as an IN
+            # unrestricted user (no user at all, or an active superuser -- see
-        # filter: for a large library that is a wasted quadratic scan in the
+            # user_is_unrestricted) can see every document, so skip
-        # vector store at best, and past ~32,763 documents a hard
+            # materializing every visible pk into a Python list and passing it
-        # sqlite3.OperationalError (SQLite's bound-parameter limit) at worst.
+            # through as an IN filter: for a large library that is a wasted
-        # get_objects_for_user_owner_aware() would return every Document for a
+            # quadratic scan in the vector store at best, and past ~32,763
-        # superuser anyway (guardian's own with_superuser shortcut), so this
+            # documents a hard sqlite3.OperationalError (SQLite's
-        # changes nothing about which documents are considered -- only how we
+            # bound-parameter limit) at worst.
-        # get there.
+            # permitted_object_ids() has its own superuser shortcut that would
+            # return every Document's id anyway, so this changes nothing about
+            # which documents are considered -- only how we get there.
             visible_document_ids = (
                 None
-            if user is None or user.is_superuser
+                if user_is_unrestricted(user)
-            else list(
+                else list(permitted_object_ids(user, Document, "view_document"))
-                get_objects_for_user_owner_aware(
-                    user,
-                    "view_document",
-                    Document,
-                ).values_list("pk", flat=True),
-            )
             )
             nodes = retrieve_similar_nodes(
                 document,
                 top_k=TAXONOMY_CANDIDATE_TOP_K,
                 document_ids=visible_document_ids,
             )
+            similar_documents = _node_document_weights(nodes)
+        else:
+            # See _fulltext_similar_documents: it applies its own permission
+            # filter via `user`, so no visible-document-id list is needed here.
+            similar_documents = _fulltext_similar_documents(
+                document,
+                user,
+                top_k=TAXONOMY_CANDIDATE_TOP_K,
+            )
 
-        candidates = build_taxonomy_candidates(nodes, user)
+        candidates = build_taxonomy_candidates(similar_documents, user)
 
-        # ``nodes`` are already ordered by descending vector similarity; don't lose it.
+        # similar_documents is already ordered by descending weight; don't lose it.
-        similar_document_ids = list(dict.fromkeys(_node_document_ids(nodes)))
+        similar_document_ids = [s["document_id"] for s in similar_documents]
         similar_documents_by_id = Document.objects.in_bulk(similar_document_ids)
         similar_docs = [
             similar_documents_by_id[document_id]
@@ -186,8 +240,8 @@ def get_taxonomy_context(
             context_blocks.append(f"TITLE: {title}\n{text}")
     except Exception:
         logger.exception(
-            "Failed to retrieve RAG neighbours for document %s; continuing "
+            "Failed to retrieve similar-document context for document %s; "
-            "without taxonomy candidates or similar-document context.",
+            "continuing without taxonomy candidates or similar-document context.",
             document.pk,
         )
         return empty_taxonomy_candidates(), ""
@@ -241,7 +295,6 @@ def get_ai_document_classification(
 ) -> ClassificationSuggestions:
     ai_config = AIConfig()
 
-    if ai_config.llm_embedding_backend:
     candidates, context = get_taxonomy_context(document, user)
     prompt = build_prompt_with_rag(
         document,
@@ -249,9 +302,6 @@ def get_ai_document_classification(
         candidates=candidates,
         context=context,
     )
-    else:
-        candidates = empty_taxonomy_candidates()
-        prompt = build_prompt_without_rag(document, ai_config, candidates=candidates)
 
     client = AIClient()
     # Hand the pooled DB connection back while the (slow) LLM query runs so it

src/paperless_ai/client.py

@@ -22,6 +22,7 @@ from paperless.network import validate_outbound_http_url
 from paperless_ai.base_model import ClassificationSuggestions
 from paperless_ai.base_model import DocumentClassifierSchema
 from paperless_ai.base_model import model_to_classification_suggestions
+from paperless_ai.exceptions import LLMProviderError
 from paperless_ai.exceptions import LLMTimeoutError
 
 logger = logging.getLogger("paperless_ai.client")
@@ -132,7 +133,7 @@ class AIClient:
         from llama_index.core.llms import ChatMessage
 
         if self.settings.llm_backend == LLMBackend.OLLAMA:
-            with self._normalize_timeouts():
+            with self._normalize_errors():
                 result = self.llm.chat(
                     [ChatMessage(role="user", content=prompt)],
                     format=DocumentClassifierSchema.model_json_schema(),
@@ -153,7 +154,7 @@ class AIClient:
             content=f"{prompt}\n\n"
             f"Answer by calling the {tool.metadata.name} tool. Do not write the answer as text.",
         )
-        with self._normalize_timeouts():
+        with self._normalize_errors():
             result = self.llm.chat_with_tools(
                 tools=[tool],
                 user_msg=user_msg,
@@ -173,7 +174,7 @@ class AIClient:
         )
 
     @contextmanager
-    def _normalize_timeouts(self) -> Iterator[None]:
+    def _normalize_errors(self) -> Iterator[None]:
         try:
             yield
         except httpx.TimeoutException as exc:
@@ -181,8 +182,23 @@ class AIClient:
         except Exception as exc:
             if self._is_openai_timeout(exc):
                 raise LLMTimeoutError from exc
+            if self._is_provider_error(exc):
+                raise LLMProviderError from exc
             raise
 
+    def _is_provider_error(self, exc: Exception) -> bool:
+        if self.settings.llm_backend == LLMBackend.OLLAMA:
+            from ollama import ResponseError
+
+            return isinstance(exc, ResponseError)
+
+        if self.settings.llm_backend == LLMBackend.OPENAI_LIKE:
+            from openai import APIStatusError
+
+            return isinstance(exc, APIStatusError)
+
+        return False
+
     def _is_openai_timeout(self, exc: Exception) -> bool:
         if self.settings.llm_backend != LLMBackend.OPENAI_LIKE:
             return False

src/paperless_ai/exceptions.py

@@ -1,2 +1,6 @@
 class LLMTimeoutError(Exception):
     pass
+
+
+class LLMProviderError(Exception):
+    """The LLM backend rejected the request."""

src/paperless_ai/indexing.py

@@ -721,20 +721,3 @@ def retrieve_similar_nodes(
             continue
         filtered.append(node)
     return filtered
-
-
-def _node_document_ids(nodes: list["NodeWithScore"]) -> list[int]:
-    document_ids: list[int] = []
-    for node in nodes:
-        document_id = node.metadata.get("document_id")
-        if document_id is None:  # pragma: no cover
-            # See the matching guard in retrieve_similar_nodes() above.
-            continue
-        try:
-            document_ids.append(int(document_id))
-        except ValueError:  # pragma: no cover
-            logger.warning(
-                "Skipping LLM index result with invalid document_id %r.",
-                document_id,
-            )
-    return document_ids

src/paperless_ai/taxonomy.py

@@ -31,6 +31,11 @@ class TaxonomyCandidate(TypedDict):
     weight: float
 
 
+class SimilarDocument(TypedDict):
+    document_id: int
+    weight: float
+
+
 class TaxonomyCandidates(TypedDict):
     tags: list[TaxonomyCandidate]
     document_types: list[TaxonomyCandidate]
@@ -49,10 +54,10 @@ def empty_taxonomy_candidates() -> TaxonomyCandidates:
     )
 
 
-def _node_document_weights(nodes: list["NodeWithScore"]) -> dict[int, float]:
+def _node_document_weights(nodes: list["NodeWithScore"]) -> list[SimilarDocument]:
-    """document_id -> that node's similarity score, summed if a document_id
+    """Sum each node's similarity score into its document_id (a document can
-    appears more than once across the retrieved nodes (e.g. multiple chunks
+    appear via multiple chunks/nodes) and return one SimilarDocument per
-    of the same source document)."""
+    distinct document_id."""
     weights: dict[int, float] = defaultdict(float)
     for node in nodes:
         document_id = node.metadata.get("document_id")
@@ -65,7 +70,14 @@ def _node_document_weights(nodes: list["NodeWithScore"]) -> dict[int, float]:
             weights[int(document_id)] += float(node.score or 0.0)
         except (TypeError, ValueError):  # pragma: no cover
             continue
-    return weights
+    return sorted(
+        (
+            SimilarDocument(document_id=document_id, weight=weight)
+            for document_id, weight in weights.items()
+        ),
+        key=lambda similar: similar["weight"],
+        reverse=True,
+    )
 
 
 def _visible_ranked_candidates(
@@ -101,21 +113,26 @@ def _visible_ranked_candidates(
 
 
 def build_taxonomy_candidates(
-    nodes: list["NodeWithScore"],
+    similar_documents: list[SimilarDocument],
     user: User | None,
 ) -> TaxonomyCandidates:
-    """Resolve each neighbour node's document_id to a live Document, read its
+    """Resolve each similar document's id to a live Document, read its
-    *current* tags/type/correspondent/storage_path via the ORM (never the
+    *current* tags/type/correspondent/storage_path via the ORM (never any
-    possibly-stale names cached in vector-index node metadata), weight each
+    possibly-stale names an adapter's source might have cached), weight each
-    distinct taxonomy object by aggregate neighbour similarity, permission-filter
+    distinct taxonomy object by aggregate similarity weight, permission-filter
     against what ``user`` can see, and return each category ranked by weight
-    and capped.
+    and capped. ``similar_documents`` may come from either the vector-RAG
+    adapter or the full-text fallback adapter - both produce this same shape.
     """
-
+    if not similar_documents:
-    document_weights = _node_document_weights(nodes)
-    if not document_weights:
         return empty_taxonomy_candidates()
 
+    # Both adapters guarantee at most one SimilarDocument per document_id, so
+    # this never silently drops a duplicate's weight.
+    document_weights: dict[int, float] = {
+        s["document_id"]: s["weight"] for s in similar_documents
+    }
+
     # Only .tags.all() needs prefetching (a reverse M2M, one extra query for
     # the whole batch). document_type/correspondent/storage_path are read
     # below via their *_id columns (neighbour.document_type_id, etc.), which

src/paperless_ai/tests/test_ai_classifier.py

@@ -1,4 +1,5 @@
 import datetime
+from collections.abc import Generator
 from types import SimpleNamespace
 from unittest.mock import MagicMock
 from unittest.mock import patch
@@ -6,18 +7,24 @@ from unittest.mock import patch
 import pytest
 import pytest_mock
 from django.test import override_settings
+from guardian.shortcuts import assign_perm
+from guardian.shortcuts import remove_perm
 
 from documents.models import Document
+from documents.search import TantivyBackend
 from documents.tests.factories import DocumentFactory
 from documents.tests.factories import TagFactory
 from documents.tests.factories import UserFactory
 from paperless.config import AIConfig
+from paperless_ai.ai_classifier import TAXONOMY_CANDIDATE_TOP_K
+from paperless_ai.ai_classifier import _fulltext_similar_documents
 from paperless_ai.ai_classifier import build_localization_prompt
 from paperless_ai.ai_classifier import build_prompt_with_rag
 from paperless_ai.ai_classifier import build_prompt_without_rag
 from paperless_ai.ai_classifier import get_ai_document_classification
 from paperless_ai.ai_classifier import get_language_name
 from paperless_ai.ai_classifier import get_taxonomy_context
+from paperless_ai.taxonomy import SimilarDocument
 from paperless_ai.taxonomy import TaxonomyCandidate
 from paperless_ai.taxonomy import TaxonomyCandidates
 
@@ -220,12 +227,10 @@ def test_use_rag_if_configured(
 
 @pytest.mark.django_db
 @patch("paperless_ai.client.AIClient.run_llm_query")
-@patch("paperless_ai.ai_classifier.build_prompt_without_rag")
+@patch("paperless_ai.ai_classifier.build_prompt_with_rag")
-@patch("paperless_ai.ai_classifier.AIConfig")
 @override_settings(LLM_BACKEND="ollama", LLM_MODEL="some_model")
-def test_use_without_rag_if_not_configured(
+def test_use_rag_prompt_even_without_embedding_backend(
-    mock_ai_config,
+    mock_build_prompt_with_rag,
-    mock_build_prompt_without_rag,
     mock_run_llm_query,
     mock_document,
 ):
@@ -235,13 +240,13 @@ def test_use_without_rag_if_not_configured(
     WHEN:
         - get_ai_document_classification() is called
     THEN:
-        - The non-RAG prompt builder is used
+        - The RAG-context prompt builder is still used (fed by the full-text
+          fallback's context/candidates instead of the vector store's)
     """
-    mock_ai_config.return_value.llm_embedding_backend = None
+    mock_build_prompt_with_rag.return_value = "Prompt with RAG"
-    mock_build_prompt_without_rag.return_value = "Prompt without RAG"
     mock_run_llm_query.return_value = NESTED_SUGGESTIONS
     get_ai_document_classification(mock_document)
-    mock_build_prompt_without_rag.assert_called_once()
+    mock_build_prompt_with_rag.assert_called_once()
 
 
 @pytest.mark.django_db
@@ -320,6 +325,7 @@ def test_build_localization_prompt_preserves_unicode_characters():
 
 
 @pytest.mark.django_db
+@override_settings(LLM_EMBEDDING_BACKEND="huggingface")
 def test_get_taxonomy_context_assembles_rag_text_and_candidates():
     """
     GIVEN:
@@ -354,6 +360,7 @@ def test_get_taxonomy_context_assembles_rag_text_and_candidates():
 
 
 @pytest.mark.django_db
+@override_settings(LLM_EMBEDDING_BACKEND="huggingface")
 def test_get_taxonomy_context_preserves_similarity_order_and_distinct_documents():
     """
     GIVEN:
@@ -424,6 +431,7 @@ def test_get_taxonomy_context_preserves_similarity_order_and_distinct_documents(
 
 
 @pytest.mark.django_db
+@override_settings(LLM_EMBEDDING_BACKEND="huggingface")
 def test_get_taxonomy_context_no_similar_docs():
     """
     GIVEN:
@@ -447,6 +455,67 @@ def test_get_taxonomy_context_no_similar_docs():
     }
 
 
+@pytest.mark.django_db
+def test_get_taxonomy_context_uses_fulltext_fallback_when_no_embedding_backend(
+    mocker: pytest_mock.MockerFixture,
+) -> None:
+    """
+    GIVEN:
+        - No LLM embedding backend is configured (the default test settings)
+    WHEN:
+        - get_taxonomy_context() is called
+    THEN:
+        - _fulltext_similar_documents() is called with the document, the user
+          and TAXONOMY_CANDIDATE_TOP_K
+        - retrieve_similar_nodes() (the vector path) is never called
+    """
+    document = DocumentFactory.create(content="Some content")
+    mock_fulltext = mocker.patch(
+        "paperless_ai.ai_classifier._fulltext_similar_documents",
+        return_value=[],
+    )
+    mock_retrieve = mocker.patch("paperless_ai.ai_classifier.retrieve_similar_nodes")
+
+    get_taxonomy_context(document, user=None)
+
+    mock_fulltext.assert_called_once_with(
+        document,
+        None,
+        top_k=TAXONOMY_CANDIDATE_TOP_K,
+    )
+    mock_retrieve.assert_not_called()
+
+
+@pytest.mark.django_db
+@override_settings(LLM_EMBEDDING_BACKEND="huggingface")
+def test_get_taxonomy_context_uses_vector_path_when_embedding_backend_configured(
+    mocker: pytest_mock.MockerFixture,
+) -> None:
+    """
+    GIVEN:
+        - An LLM embedding backend is configured
+    WHEN:
+        - get_taxonomy_context() is called
+    THEN:
+        - retrieve_similar_nodes() (the vector path) is called
+        - _fulltext_similar_documents() (the no-embedding-backend fallback)
+          is never called
+    """
+    document = DocumentFactory.create(content="Some content")
+    mock_retrieve = mocker.patch(
+        "paperless_ai.ai_classifier.retrieve_similar_nodes",
+        return_value=[],
+    )
+    mock_fulltext = mocker.patch(
+        "paperless_ai.ai_classifier._fulltext_similar_documents",
+    )
+
+    get_taxonomy_context(document, user=None)
+
+    mock_retrieve.assert_called_once()
+    mock_fulltext.assert_not_called()
+
+
 class TestGetTaxonomyContextVisibility:
     """get_taxonomy_context must not materialize every visible document id
     for a user who can already see the whole library: a superuser (like no
@@ -459,6 +528,7 @@ class TestGetTaxonomyContextVisibility:
     """
 
     @pytest.mark.django_db
+    @override_settings(LLM_EMBEDDING_BACKEND="huggingface")
     def test_skips_permission_lookup_for_superuser(
         self,
         mocker: pytest_mock.MockerFixture,
@@ -477,17 +547,18 @@ class TestGetTaxonomyContextVisibility:
             "paperless_ai.ai_classifier.retrieve_similar_nodes",
             return_value=[],
         )
-        mock_get_objects = mocker.patch(
+        mock_permitted = mocker.patch(
-            "paperless_ai.ai_classifier.get_objects_for_user_owner_aware",
+            "paperless_ai.ai_classifier.permitted_object_ids",
         )
         user = UserFactory.create(is_superuser=True)
 
         get_taxonomy_context(document, user)
 
-        mock_get_objects.assert_not_called()
+        mock_permitted.assert_not_called()
         assert mock_retrieve.call_args.kwargs["document_ids"] is None
 
     @pytest.mark.django_db
+    @override_settings(LLM_EMBEDDING_BACKEND="huggingface")
     def test_skips_permission_lookup_when_no_user(
         self,
         mocker: pytest_mock.MockerFixture,
@@ -506,16 +577,17 @@ class TestGetTaxonomyContextVisibility:
             "paperless_ai.ai_classifier.retrieve_similar_nodes",
             return_value=[],
         )
-        mock_get_objects = mocker.patch(
+        mock_permitted = mocker.patch(
-            "paperless_ai.ai_classifier.get_objects_for_user_owner_aware",
+            "paperless_ai.ai_classifier.permitted_object_ids",
         )
 
         get_taxonomy_context(document, None)
 
-        mock_get_objects.assert_not_called()
+        mock_permitted.assert_not_called()
         assert mock_retrieve.call_args.kwargs["document_ids"] is None
 
     @pytest.mark.django_db
+    @override_settings(LLM_EMBEDDING_BACKEND="huggingface")
     def test_restricts_to_visible_documents_for_non_superuser(
         self,
         mocker: pytest_mock.MockerFixture,
@@ -526,7 +598,7 @@ class TestGetTaxonomyContextVisibility:
         WHEN:
             - get_taxonomy_context() is called
         THEN:
-            - The user's visible document ids are looked up and passed to
+            - The user's permitted document ids are looked up and passed to
               retrieve_similar_nodes() as a restriction
         """
         document = DocumentFactory.create(content="Some content")
@@ -534,21 +606,232 @@ class TestGetTaxonomyContextVisibility:
             "paperless_ai.ai_classifier.retrieve_similar_nodes",
             return_value=[],
         )
-        mock_queryset = mocker.MagicMock()
+        mock_permitted = mocker.patch(
-        mock_queryset.values_list.return_value = [1, 2, 3]
+            "paperless_ai.ai_classifier.permitted_object_ids",
-        mock_get_objects = mocker.patch(
+            return_value=[1, 2, 3],
-            "paperless_ai.ai_classifier.get_objects_for_user_owner_aware",
-            return_value=mock_queryset,
         )
         user = UserFactory.create(is_superuser=False)
 
         get_taxonomy_context(document, user)
 
-        mock_get_objects.assert_called_once_with(user, "view_document", Document)
+        mock_permitted.assert_called_once_with(user, Document, "view_document")
         assert mock_retrieve.call_args.kwargs["document_ids"] == [1, 2, 3]
 
 
 @pytest.mark.django_db
+class TestFulltextSimilarDocuments:
+    """_fulltext_similar_documents is the no-embedding-backend fallback: it
+    asks the Tantivy full-text index for "More Like This" neighbours instead
+    of the vector store, and synthesizes a rank-based weight since Tantivy's
+    more_like_this_ids returns only an ordered id list, no scores.
+    """
+
+    @pytest.fixture
+    def fulltext_backend(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> Generator[TantivyBackend, None, None]:
+        """An in-memory Tantivy backend, wired up as the module-level
+        singleton _fulltext_similar_documents resolves via get_backend()."""
+        backend = TantivyBackend(path=None)
+        backend.open()
+        mocker.patch("documents.search.get_backend", return_value=backend)
+        try:
+            yield backend
+        finally:
+            backend.close()
+
+    def test_ranks_by_rank_based_weight_descending(
+        self,
+        fulltext_backend: TantivyBackend,
+    ) -> None:
+        """
+        GIVEN:
+            - A source document and two similar documents indexed in Tantivy
+        WHEN:
+            - _fulltext_similar_documents() is called
+        THEN:
+            - Each result's weight reflects its rank (first result weighted
+              higher than the second), not a raw similarity score
+        """
+        source = DocumentFactory.create(content="quarterly financial report details")
+        first = DocumentFactory.create(content="quarterly financial report details")
+        second = DocumentFactory.create(content="financial report")
+        for doc in (source, first, second):
+            fulltext_backend.add_or_update(doc)
+
+        result = _fulltext_similar_documents(source, user=None, top_k=5)
+
+        assert len(result) == 2
+        weight_by_id = {s["document_id"]: s["weight"] for s in result}
+        assert weight_by_id[first.pk] > weight_by_id[second.pk]
+
+    def test_excludes_source_document(
+        self,
+        fulltext_backend: TantivyBackend,
+    ) -> None:
+        """
+        GIVEN:
+            - A source document indexed in Tantivy with no other documents
+        WHEN:
+            - _fulltext_similar_documents() is called
+        THEN:
+            - An empty list is returned - the source document is never its
+              own similar document
+        """
+        source = DocumentFactory.create(content="unique unrelated content")
+        fulltext_backend.add_or_update(source)
+
+        result = _fulltext_similar_documents(source, user=None, top_k=5)
+
+        assert result == []
+
+    def test_empty_index_returns_empty_list(
+        self,
+        fulltext_backend: TantivyBackend,
+    ) -> None:
+        """
+        GIVEN:
+            - A document that has never been indexed (fresh/empty Tantivy index)
+        WHEN:
+            - _fulltext_similar_documents() is called
+        THEN:
+            - An empty list is returned rather than raising
+        """
+        source = DocumentFactory.create(content="never indexed")
+
+        result = _fulltext_similar_documents(source, user=None, top_k=5)
+
+        assert result == []
+
+    def test_respects_top_k_limit(
+        self,
+        fulltext_backend: TantivyBackend,
+    ) -> None:
+        """
+        GIVEN:
+            - A source document and four similar documents indexed
+        WHEN:
+            - _fulltext_similar_documents() is called with top_k=2
+        THEN:
+            - At most 2 results are returned
+        """
+        source = DocumentFactory.create(content="shared overlapping keyword text")
+        fulltext_backend.add_or_update(source)
+        for _ in range(4):
+            fulltext_backend.add_or_update(
+                DocumentFactory.create(content="shared overlapping keyword text"),
+            )
+
+        result = _fulltext_similar_documents(source, user=None, top_k=2)
+
+        assert len(result) == 2
+
+    def test_result_shape_is_similar_document(
+        self,
+        fulltext_backend: TantivyBackend,
+    ) -> None:
+        """
+        GIVEN:
+            - A source document and one similar document indexed
+        WHEN:
+            - _fulltext_similar_documents() is called
+        THEN:
+            - Each result is a SimilarDocument (document_id + weight only)
+        """
+        source = DocumentFactory.create(content="shared content phrase")
+        other = DocumentFactory.create(content="shared content phrase")
+        fulltext_backend.add_or_update(source)
+        fulltext_backend.add_or_update(other)
+
+        result = _fulltext_similar_documents(source, user=None, top_k=5)
+
+        # rank 0 (the only/best result) with top_k=5 -> weight = top_k - rank = 5.0,
+        # per the "first result gets top_k, the last gets 1" formula.
+        assert result == [SimilarDocument(document_id=other.pk, weight=5.0)]
+
+    def test_superuser_sees_other_users_documents(
+        self,
+        fulltext_backend: TantivyBackend,
+    ) -> None:
+        """
+        GIVEN:
+            - A source document owned by one user and a similar document
+              owned by a different user, with no sharing between them
+        WHEN:
+            - _fulltext_similar_documents() is called with a superuser
+        THEN:
+            - The other user's document is still returned as a similar
+              document - a superuser must not be narrowed by the backend's
+              owner-based permission filter
+        """
+        owner = UserFactory.create()
+        other_owner = UserFactory.create()
+        superuser = UserFactory.create(is_superuser=True)
+        source = DocumentFactory.create(
+            content="shared content phrase",
+            owner=owner,
+        )
+        other = DocumentFactory.create(
+            content="shared content phrase",
+            owner=other_owner,
+        )
+        fulltext_backend.add_or_update(source)
+        fulltext_backend.add_or_update(other)
+
+        result = _fulltext_similar_documents(source, user=superuser, top_k=5)
+
+        assert [s["document_id"] for s in result] == [other.pk]
+
+    def test_excludes_stale_permitted_document_for_regular_user(
+        self,
+        fulltext_backend: TantivyBackend,
+    ) -> None:
+        """
+        GIVEN:
+            - A regular (non-superuser) user
+            - A similar document the user is permitted to view, and another
+              similar document indexed while the user still had view
+              permission but which has since had that permission revoked in
+              the database, i.e. the Tantivy index has stale permission data
+        WHEN:
+            - _fulltext_similar_documents() is called with that user
+        THEN:
+            - Only the still-permitted document is returned - the DB
+              re-check via restrict_queryset_to_visible() must catch the
+              document Tantivy's stale index still thinks is visible
+        """
+        owner = UserFactory.create()
+        viewer = UserFactory.create(is_superuser=False)
+        source = DocumentFactory.create(
+            content="shared content phrase",
+            owner=owner,
+        )
+        permitted = DocumentFactory.create(
+            content="shared content phrase",
+            owner=owner,
+        )
+        now_private = DocumentFactory.create(
+            content="shared content phrase",
+            owner=owner,
+        )
+        assign_perm("view_document", viewer, permitted)
+        assign_perm("view_document", viewer, now_private)
+        fulltext_backend.add_or_update(source)
+        fulltext_backend.add_or_update(permitted)
+        fulltext_backend.add_or_update(now_private)
+
+        # Revoke access after indexing, without reindexing: the index still
+        # carries viewer as a permitted viewer for `now_private`.
+        remove_perm("view_document", viewer, now_private)
+
+        result = _fulltext_similar_documents(source, user=viewer, top_k=5)
+
+        assert [s["document_id"] for s in result] == [permitted.pk]
+
+
+@pytest.mark.django_db
+@override_settings(LLM_EMBEDDING_BACKEND="huggingface")
 @patch("paperless_ai.ai_classifier.retrieve_similar_nodes")
 def test_get_taxonomy_context_retrieval_failure_degrades_to_no_hints(mock_retrieve):
     """
@@ -575,6 +858,7 @@ def test_get_taxonomy_context_retrieval_failure_degrades_to_no_hints(mock_retrie
 
 
 @pytest.mark.django_db
+@override_settings(LLM_EMBEDDING_BACKEND="huggingface")
 @patch("paperless_ai.ai_classifier.build_taxonomy_candidates")
 @patch("paperless_ai.ai_classifier.retrieve_similar_nodes")
 def test_get_taxonomy_context_candidate_building_failure_degrades_to_no_hints(

src/paperless_ai/tests/test_ai_indexing.py

@@ -1188,9 +1188,7 @@ class TestRetrieveSimilarNodesAgainstRealIndex:
 
         nodes = indexing.retrieve_similar_nodes(a, document_ids=[b.id])
 
-        assert all(
+        assert all(int(node.metadata["document_id"]) == b.id for node in nodes)
-            document_id == b.id for document_id in indexing._node_document_ids(nodes)
-        )
 
     def test_excludes_self(
         self,
@@ -1212,7 +1210,7 @@ class TestRetrieveSimilarNodesAgainstRealIndex:
 
         nodes = indexing.retrieve_similar_nodes(a, top_k=5)
 
-        assert set(indexing._node_document_ids(nodes)) == {b.id}
+        assert {int(node.metadata["document_id"]) for node in nodes} == {b.id}
 
     def test_excludes_self_with_multiple_chunks(
         self,
@@ -1235,4 +1233,4 @@ class TestRetrieveSimilarNodesAgainstRealIndex:
 
         nodes = indexing.retrieve_similar_nodes(a, top_k=3)
 
-        assert set(indexing._node_document_ids(nodes)) == {b.id}
+        assert {int(node.metadata["document_id"]) for node in nodes} == {b.id}

src/paperless_ai/tests/test_client.py

@@ -4,6 +4,7 @@ from unittest.mock import MagicMock
 from unittest.mock import patch
 
 import httpx
+import ollama
 import openai
 import pytest
 from llama_index.core.llms.llm import ToolSelection
@@ -11,6 +12,7 @@ from llama_index.core.llms.llm import ToolSelection
 from paperless_ai.client import LLM_SYSTEM_PROMPT
 from paperless_ai.client import PLACEHOLDER_API_KEY
 from paperless_ai.client import AIClient
+from paperless_ai.exceptions import LLMProviderError
 from paperless_ai.exceptions import LLMTimeoutError
 
 
@@ -214,6 +216,52 @@ def test_run_llm_query_openai_timeout_raises_local_error(
         client.run_llm_query("test_prompt")
 
 
+def test_run_llm_query_openai_status_error_raises_provider_error(
+    mock_ai_config,
+    mock_openai_llm,
+):
+    mock_ai_config.llm_backend = "openai-like"
+    mock_ai_config.llm_model = "test_model"
+    mock_ai_config.llm_endpoint = "http://test-url"
+
+    request = httpx.Request("POST", "http://test-url/v1/chat/completions")
+    body = {"error": {"message": "Thinking mode does not support this tool_choice"}}
+    mock_openai_llm.return_value.chat_with_tools.side_effect = openai.BadRequestError(
+        "Error code: 400",
+        response=httpx.Response(400, request=request, json=body),
+        body=body,
+    )
+
+    client = AIClient()
+
+    with pytest.raises(LLMProviderError) as exc_info:
+        client.run_llm_query("test_prompt")
+    assert str(exc_info.value) == ""
+    assert isinstance(exc_info.value.__cause__, openai.BadRequestError)
+
+
+def test_run_llm_query_ollama_response_error_raises_provider_error(
+    mock_ai_config,
+    mock_ollama_llm,
+):
+    mock_ai_config.llm_backend = "ollama"
+    mock_ai_config.llm_model = "test_model"
+    mock_ai_config.llm_endpoint = "http://test-url"
+
+    response_error = ollama.ResponseError(
+        "confidential provider response",
+        status_code=400,
+    )
+    mock_ollama_llm.return_value.chat.side_effect = response_error
+
+    client = AIClient()
+
+    with pytest.raises(LLMProviderError) as exc_info:
+        client.run_llm_query("test_prompt")
+    assert str(exc_info.value) == ""
+    assert exc_info.value.__cause__ is response_error
+
+
 def test_run_llm_query_httpx_timeout_raises_local_error(
     mock_ai_config,
     mock_ollama_llm,

src/paperless_ai/tests/test_taxonomy.py

@@ -1,5 +1,4 @@
 import json
-from types import SimpleNamespace
 
 import pytest
 import pytest_mock
@@ -10,14 +9,14 @@ from documents.tests.factories import DocumentTypeFactory
 from documents.tests.factories import StoragePathFactory
 from documents.tests.factories import TagFactory
 from documents.tests.factories import UserFactory
+from paperless_ai.taxonomy import SimilarDocument
 from paperless_ai.taxonomy import TaxonomyCandidates
 from paperless_ai.taxonomy import build_taxonomy_candidates
 from paperless_ai.taxonomy import format_taxonomy_for_prompt
 
 
-def make_node(document_id: int, score: float) -> SimpleNamespace:
+def make_similar(document_id: int, weight: float) -> SimilarDocument:
-    """A stand-in for NodeWithScore: only ``.metadata``/``.score`` are read."""
+    return SimilarDocument(document_id=document_id, weight=weight)
-    return SimpleNamespace(metadata={"document_id": str(document_id)}, score=score)
 
 
 @pytest.mark.django_db
@@ -53,9 +52,9 @@ class TestBuildTaxonomyCandidates:
         doc_a.tags.add(tag)
         doc_b = DocumentFactory.create()
         doc_b.tags.add(tag)
-        nodes = [make_node(doc_a.pk, 0.9), make_node(doc_b.pk, 0.4)]
+        similar_documents = [make_similar(doc_a.pk, 0.9), make_similar(doc_b.pk, 0.4)]
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert len(result["tags"]) == 1
         assert result["tags"][0]["id"] == tag.pk
@@ -80,9 +79,9 @@ class TestBuildTaxonomyCandidates:
         document.tags.add(tag)
         tag.name = "New Name"
         tag.save()
-        nodes = [make_node(document.pk, 0.5)]
+        similar_documents = [make_similar(document.pk, 0.5)]
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert result["tags"][0]["name"] == "New Name"
 
@@ -102,9 +101,9 @@ class TestBuildTaxonomyCandidates:
         document = DocumentFactory.create()
         document.tags.add(tag)
         tag.delete()
-        nodes = [make_node(document.pk, 0.5)]
+        similar_documents = [make_similar(document.pk, 0.5)]
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert result["tags"] == []
 
@@ -123,9 +122,12 @@ class TestBuildTaxonomyCandidates:
         strong_doc.tags.add(strong_tag)
         weak_doc = DocumentFactory.create()
         weak_doc.tags.add(weak_tag)
-        nodes = [make_node(strong_doc.pk, 0.9), make_node(weak_doc.pk, 0.1)]
+        similar_documents = [
+            make_similar(strong_doc.pk, 0.9),
+            make_similar(weak_doc.pk, 0.1),
+        ]
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert [c["name"] for c in result["tags"]] == ["Strong", "Weak"]
 
@@ -141,9 +143,9 @@ class TestBuildTaxonomyCandidates:
         document = DocumentFactory.create()
         for i in range(15):
             document.tags.add(TagFactory.create(name=f"Tag{i}"))
-        nodes = [make_node(document.pk, 0.5)]
+        similar_documents = [make_similar(document.pk, 0.5)]
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert len(result["tags"]) == 10
 
@@ -157,12 +159,12 @@ class TestBuildTaxonomyCandidates:
             - Only 5 correspondents are returned
         """
         correspondents = CorrespondentFactory.create_batch(7)
-        nodes = [
+        similar_documents = [
-            make_node(DocumentFactory.create(correspondent=c).pk, 0.5)
+            make_similar(DocumentFactory.create(correspondent=c).pk, 0.5)
             for c in correspondents
         ]
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert len(result["correspondents"]) == 5
 
@@ -177,9 +179,9 @@ class TestBuildTaxonomyCandidates:
         """
         document_type = DocumentTypeFactory.create(name="Invoice")
         document = DocumentFactory.create(document_type=document_type)
-        nodes = [make_node(document.pk, 0.5)]
+        similar_documents = [make_similar(document.pk, 0.5)]
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert len(result["document_types"]) == 1
         assert result["document_types"][0]["id"] == document_type.pk
@@ -195,12 +197,12 @@ class TestBuildTaxonomyCandidates:
             - Only 5 document_types are returned
         """
         document_types = DocumentTypeFactory.create_batch(7)
-        nodes = [
+        similar_documents = [
-            make_node(DocumentFactory.create(document_type=dt).pk, 0.5)
+            make_similar(DocumentFactory.create(document_type=dt).pk, 0.5)
             for dt in document_types
         ]
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert len(result["document_types"]) == 5
 
@@ -215,9 +217,9 @@ class TestBuildTaxonomyCandidates:
         """
         storage_path = StoragePathFactory.create(name="Invoices")
         document = DocumentFactory.create(storage_path=storage_path)
-        nodes = [make_node(document.pk, 0.5)]
+        similar_documents = [make_similar(document.pk, 0.5)]
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert len(result["storage_paths"]) == 1
         assert result["storage_paths"][0]["id"] == storage_path.pk
@@ -233,12 +235,12 @@ class TestBuildTaxonomyCandidates:
             - Only 5 storage_paths are returned
         """
         storage_paths = StoragePathFactory.create_batch(7)
-        nodes = [
+        similar_documents = [
-            make_node(DocumentFactory.create(storage_path=sp).pk, 0.5)
+            make_similar(DocumentFactory.create(storage_path=sp).pk, 0.5)
             for sp in storage_paths
         ]
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert len(result["storage_paths"]) == 5
 
@@ -258,14 +260,14 @@ class TestBuildTaxonomyCandidates:
         tag = TagFactory.create(name="Restricted")
         document = DocumentFactory.create()
         document.tags.add(tag)
-        nodes = [make_node(document.pk, 0.5)]
+        similar_documents = [make_similar(document.pk, 0.5)]
         user = UserFactory.create()
         mocker.patch(
             "documents.permissions.permitted_object_ids",
             return_value=[],  # user cannot see this tag
         )
 
-        result = build_taxonomy_candidates(nodes, user=user)
+        result = build_taxonomy_candidates(similar_documents, user=user)
 
         assert result["tags"] == []
 
@@ -295,10 +297,10 @@ class TestBuildTaxonomyCandidates:
         tag.save()
         document = DocumentFactory.create()
         document.tags.add(tag)
-        nodes = [make_node(document.pk, 0.5)]
+        similar_documents = [make_similar(document.pk, 0.5)]
         spy = mocker.patch("documents.permissions.permitted_object_ids")
 
-        result = build_taxonomy_candidates(nodes, user=None)
+        result = build_taxonomy_candidates(similar_documents, user=None)
 
         assert result["tags"][0]["name"] == "Owned"
         spy.assert_not_called()

src/paperless_mail/tasks.py

@@ -1,7 +1,9 @@
 import logging
 
+from celery import Task
 from celery import shared_task
 
+from documents.models import PaperlessTask
 from paperless_mail.mail import MailAccountHandler
 from paperless_mail.mail import MailError
 from paperless_mail.models import MailAccount
@@ -10,8 +12,26 @@ from paperless_mail.models import MailRule
 logger = logging.getLogger("paperless.mail.tasks")
 
 
-@shared_task
+@shared_task(bind=True)
-def process_mail_accounts(account_ids: list[int] | None = None) -> str:
+def process_mail_accounts(self: Task, account_ids: list[int] | None = None) -> str:
+    # A scheduled check can still be running (or queued) when the next one
+    # ProcessedMail dedup only records a message once its
+    # handling has finished, so an overlapping run can still pick up the same
+    # not-yet-recorded message. Skip outright rather than race it.
+    other_mail_fetch_running = (
+        PaperlessTask.objects.filter(
+            task_type=PaperlessTask.TaskType.MAIL_FETCH,
+            status__in=[PaperlessTask.Status.PENDING, PaperlessTask.Status.STARTED],
+        )
+        .exclude(task_id=self.request.id)
+        .exists()
+    )
+    if other_mail_fetch_running:
+        logger.info(
+            "Mail account processing is already running; skipping this run.",
+        )
+        return "Skipped: mail account processing already in progress."
+
     total_new_documents = 0
     accounts = (
         MailAccount.objects.filter(pk__in=account_ids)

src/paperless_mail/tests/test_process_mail_accounts_overlap.py

@@ -0,0 +1,134 @@
+from typing import Final
+
+import pytest
+import pytest_mock
+
+from documents.models import PaperlessTask
+from documents.tests.factories import PaperlessTaskFactory
+from paperless_mail import tasks
+from paperless_mail.tests.factories import MailAccountFactory
+from paperless_mail.tests.factories import MailRuleFactory
+
+NO_DOCUMENTS_ADDED: Final = "No new documents were added."
+SKIPPED: Final = "Skipped: mail account processing already in progress."
+
+
+@pytest.mark.django_db
+@pytest.mark.usefixtures("account_with_rule")
+class TestProcessMailAccountsOverlap:
+    @pytest.fixture
+    def account_with_rule(self) -> None:
+        """An enabled mail account with a single enabled rule."""
+        account = MailAccountFactory.create()
+        MailRuleFactory.create(account=account, enabled=True)
+
+    @pytest.mark.parametrize(
+        ("status", "expected_result", "expected_call_count"),
+        [
+            pytest.param(
+                PaperlessTask.Status.PENDING,
+                SKIPPED,
+                0,
+                id="pending-task-blocks",
+            ),
+            pytest.param(
+                PaperlessTask.Status.STARTED,
+                SKIPPED,
+                0,
+                id="started-task-blocks",
+            ),
+            pytest.param(
+                PaperlessTask.Status.SUCCESS,
+                NO_DOCUMENTS_ADDED,
+                1,
+                id="finished-task-does-not-block",
+            ),
+        ],
+    )
+    def test_skips_only_while_another_mail_fetch_task_runs(
+        self,
+        mocker: pytest_mock.MockerFixture,
+        status: PaperlessTask.Status,
+        expected_result: str,
+        expected_call_count: int,
+    ) -> None:
+        """
+        GIVEN:
+            - An enabled mail account with a rule
+            - Another mail fetch task row in the given status
+        WHEN:
+            - Mail accounts are processed
+        THEN:
+            - Processing is skipped only if that other task is pending or running
+        """
+        PaperlessTaskFactory.create(
+            task_type=PaperlessTask.TaskType.MAIL_FETCH,
+            trigger_source=PaperlessTask.TriggerSource.SCHEDULED,
+            status=status,
+        )
+
+        mocked_handle = mocker.patch.object(
+            tasks.MailAccountHandler,
+            "handle_mail_account",
+            return_value=0,
+        )
+
+        result = tasks.process_mail_accounts()
+
+        assert mocked_handle.call_count == expected_call_count
+        assert result == expected_result
+
+    def test_runs_when_no_other_mail_fetch_task_exists(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        """
+        GIVEN:
+            - An enabled mail account with a rule
+            - No other mail fetch task rows
+        WHEN:
+            - Mail accounts are processed
+        THEN:
+            - The account is handled
+        """
+        mocked_handle = mocker.patch.object(
+            tasks.MailAccountHandler,
+            "handle_mail_account",
+            return_value=0,
+        )
+
+        result = tasks.process_mail_accounts()
+
+        mocked_handle.assert_called_once()
+        assert result == NO_DOCUMENTS_ADDED
+
+    def test_does_not_skip_due_to_its_own_task_row(
+        self,
+        mocker: pytest_mock.MockerFixture,
+    ) -> None:
+        """
+        GIVEN:
+            - An enabled mail account with a rule
+            - A running mail fetch task row belonging to this very task
+        WHEN:
+            - Mail accounts are processed under that task id
+        THEN:
+            - The task does not skip itself and handles the account
+        """
+        PaperlessTaskFactory.create(
+            task_id="self-task-id",
+            task_type=PaperlessTask.TaskType.MAIL_FETCH,
+            trigger_source=PaperlessTask.TriggerSource.SCHEDULED,
+            status=PaperlessTask.Status.STARTED,
+        )
+
+        mocked_handle = mocker.patch.object(
+            tasks.MailAccountHandler,
+            "handle_mail_account",
+            return_value=0,
+        )
+
+        result = tasks.process_mail_accounts.apply(task_id="self-task-id").result
+
+        mocked_handle.assert_called_once()
+        assert result == NO_DOCUMENTS_ADDED