Compare commits

..
Author SHA1 Message Date
stumpylog 3f258db0e2 Chore: share the authorizing owner rule between the id sets and the trash filter
The trash list needs to show a version to whoever owns its root document, but
it deliberately shows only owned and unowned documents and ignores explicit
shares, so it cannot use permitted_object_ids. #14384 gave it its own
filter_queryset with a hand-built condition through the root, which describes
the same rule as the annotations in permitted_object_ids a second time.

Move those annotations into annotate_authorizing_fields. PermittedObjectsFilter
gains a parent_field attribute that its owner-only and granted paths both use,
and the trash filter shrinks to setting include_granted and parent_field.
2026-10-08 15:22:04 -07:00
stumpylog 642511c7b9 Fix: use a root document's effective content for the LLM index and prompts
The search index and the document views use the newest version's content as a
root document's content, but everything the LLM side reads used the root's own
content: the text embedded in the LLM index, the content in the classification
prompt, the context blocks from similar documents, and the text that is
searched for similar documents. After a document was replaced by a new
version the LLM kept answering from the old text.

Read get_effective_content() in those four places. The index update annotates
the content in the query like the search index does, and the similar-document
lookup annotates it for the documents it fetches, so no query is made per
document. Entries already in the LLM index keep the old text until the
document next changes or the index is rebuilt.
2026-10-08 15:22:03 -07:00
stumpylog 210d97a522 Fix: index document versions as their root document
The search index and the LLM index hold root documents only, a root being
indexed with its newest version's content. Every write path therefore had to
remember to hand them the root. Several did not: adding or deleting a note on
a version, restoring a trashed document together with its versions, and
reprocessing a version all wrote the version into the index under its own id,
where it could be returned as a separate search result. The paths that walk the
whole library did the same: the document_index reindex command put every
version in the search index, a full LLM index rebuild embedded every version,
and an incremental LLM update scoped to a version id indexed it under its own
id. Asking for documents like a version looked the version's own id up in the
index and silently found nothing.

WriteBatch.add_or_update, WriteBatch.add_or_update_ids and
llm_index_add_or_update_document now resolve a version to its root themselves.
The reindex command and update_llm_index only walk root documents, and an
incremental LLM update is scoped to the roots of the given ids through
versioning.root_document_ids, which add_or_update_ids shares. more_like_id
returns the root's id for a version. The reprocess task no longer picks the root
for the indexes and only still clears the caches of both documents.
2026-10-08 15:22:02 -07:00
stumpylog 8a96190359 Fix: authorize document versions by their root in the single-object permission check
has_perms_owner_aware judged a document by its own owner and grants, so each
endpoint that fetches a document itself had to remember to map a version to
its root document first, and one that forgot, like the more-like-this search
filter, authorized by a stale version owner.

The check now maps a Document to its root before looking at the owner and the
guardian grants, matching what permitted_document_ids does for id sets. The
eight call sites that mapped the document themselves pass it straight through.
The DRF object permission class needs no change because the document viewset
only ever serves root documents.
2026-10-08 15:21:45 -07:00
stumpylog ece8769f59 Fix: authorize document versions by their root and speed up permission id sets
permitted_document_ids judged a version by its own owner and grants, so a
version whose owner had drifted from its root's was visible to the wrong
people and hidden from the right ones. Callers patched this individually by
mapping each document to its root first. The query itself was also slow on
MariaDB: the guardian grants were a UNION cast to integers and tested with
IN inside an OR with the owner checks, which MariaDB cannot materialize, so it
re-scans the user's grants for every document. At 20k documents that took
seconds for a user with a couple of hundred grants.

permitted_object_ids now looks grants up as an EXISTS keyed on the row id cast
to a string, which uses guardian's unique index, and matches the user's groups
with an IN subquery. It takes an optional parent_field naming a self-referencing
foreign key whose target authorizes the row, and permitted_document_ids passes
root_document, so a version is visible exactly when its root is. The helper
that mapped documents to their roots at the call sites is no longer needed, so
the email, selection data, share link bundle, trash, bulk download and bulk
edit checks use the id set directly.
2026-10-08 15:21:28 -07:00
19 changed files with 907 additions and 119 deletions

No files matched your search

+18 -2
View File
@@ -50,6 +50,7 @@ from documents.models import ShareLink
from documents.models import ShareLinkBundle from documents.models import ShareLinkBundle
from documents.models import StoragePath from documents.models import StoragePath
from documents.models import Tag from documents.models import Tag
from documents.permissions import annotate_authorizing_fields
from documents.permissions import permitted_document_ids from documents.permissions import permitted_document_ids
from documents.permissions import permitted_object_ids from documents.permissions import permitted_object_ids
from documents.versioning import annotate_effective_content from documents.versioning import annotate_effective_content
@@ -1075,6 +1076,9 @@ class PermittedObjectsFilter(BaseFilterBackend):
include_granted: bool = True include_granted: bool = True
perm_codename: str | None = None perm_codename: str | None = None
# A self-referencing foreign key whose target authorizes a row, so a row is
# judged by its parent's owner and grants (``Document.root_document``).
parent_field: str | None = None
def filter_queryset(self, request, queryset, view): def filter_queryset(self, request, queryset, view):
# Before the superuser and owner-only paths, neither of which consults # Before the superuser and owner-only paths, neither of which consults
@@ -1086,11 +1090,23 @@ class PermittedObjectsFilter(BaseFilterBackend):
if request.user.is_superuser: if request.user.is_superuser:
return queryset return queryset
if not self.include_granted: if not self.include_granted:
return queryset.filter(Q(owner=request.user) | Q(owner__isnull=True)) owner_field = "owner"
if self.parent_field is not None:
queryset = annotate_authorizing_fields(queryset, self.parent_field)
owner_field = "authorizing_owner"
return queryset.filter(
Q(**{owner_field: request.user.pk})
| Q(**{f"{owner_field}__isnull": True}),
)
model = queryset.model model = queryset.model
perm = self.perm_codename or f"view_{model._meta.model_name}" perm = self.perm_codename or f"view_{model._meta.model_name}"
return queryset.filter( return queryset.filter(
id__in=permitted_object_ids(request.user, model, perm), id__in=permitted_object_ids(
request.user,
model,
perm,
parent_field=self.parent_field,
),
) )
@@ -67,12 +67,15 @@ class Command(PaperlessCommand):
if options.get("recreate"): if options.get("recreate"):
wipe_index(settings.INDEX_DIR) wipe_index(settings.INDEX_DIR)
documents = Document.objects.select_related( documents = (
Document.objects.filter(root_document__isnull=True)
.select_related(
"correspondent", "correspondent",
"document_type", "document_type",
"storage_path", "storage_path",
"owner", "owner",
).prefetch_related( )
.prefetch_related(
"tags", "tags",
"notes__user", "notes__user",
"custom_fields__field", "custom_fields__field",
@@ -80,6 +83,7 @@ class Command(PaperlessCommand):
"barcodes", "barcodes",
"versions__barcodes", "versions__barcodes",
) )
)
total = documents.count() total = documents.count()
rebuild_kwargs = {} rebuild_kwargs = {}
if options.get("heap_size_mb") is not None: if options.get("heap_size_mb") is not None:
+74 -31
View File
@@ -6,9 +6,13 @@ from django.contrib.auth.models import Permission
from django.contrib.auth.models import User from django.contrib.auth.models import User
from django.contrib.contenttypes.models import ContentType from django.contrib.contenttypes.models import ContentType
from django.db.models import Case from django.db.models import Case
from django.db.models import CharField
from django.db.models import Count from django.db.models import Count
from django.db.models import Exists
from django.db.models import F
from django.db.models import IntegerField from django.db.models import IntegerField
from django.db.models import Model from django.db.models import Model
from django.db.models import OuterRef
from django.db.models import Q from django.db.models import Q
from django.db.models import QuerySet from django.db.models import QuerySet
from django.db.models import Value from django.db.models import Value
@@ -26,6 +30,7 @@ from rest_framework.permissions import BasePermission
from rest_framework.permissions import DjangoObjectPermissions from rest_framework.permissions import DjangoObjectPermissions
from documents.models import Document from documents.models import Document
from documents.versioning import get_root_document
class PaperlessObjectPermissions(DjangoObjectPermissions): class PaperlessObjectPermissions(DjangoObjectPermissions):
@@ -349,6 +354,7 @@ def permitted_object_ids(
perm: str, perm: str,
*, *,
include_deleted: bool = False, include_deleted: bool = False,
parent_field: str | None = None,
) -> QuerySet[int]: ) -> QuerySet[int]:
""" """
Generic version of ``permitted_document_ids`` for any model with an Generic version of ``permitted_document_ids`` for any model with an
@@ -357,6 +363,20 @@ def permitted_object_ids(
soft-delete pattern (currently only ``Document``); for every other model soft-delete pattern (currently only ``Document``); for every other model
it is accepted but has no effect, since those models have no soft-delete it is accepted but has no effect, since those models have no soft-delete
concept. concept.
``parent_field`` names a self-referencing foreign key whose target
authorizes the row (``Document.root_document``). A row with a parent is
visible exactly when its parent is, judged by the parent's owner and
grants, so the row's own owner and grants are ignored.
Guardian stores ``object_pk`` as a string, so each grant is an ``EXISTS``
keyed on the row's id cast to a string, which can use guardian's
(user, permission, object_pk) unique index. Casting every ``object_pk`` to
an integer for an ``id IN (...)`` is not indexable, and MariaDB cannot
materialize it inside the owner ``OR``, so it re-scans the user's grants
for every row. The user's groups are matched with an ``IN`` subquery
rather than a join through the membership table, which SQLite plans badly
once the grant tables grow.
""" """
has_soft_delete = hasattr(model, "global_objects") has_soft_delete = hasattr(model, "global_objects")
manager = ( manager = (
@@ -364,8 +384,14 @@ def permitted_object_ids(
) )
base_qs = manager.all().only("id", "owner") base_qs = manager.all().only("id", "owner")
owner_field, key_field = "owner", "pk"
if parent_field is not None:
owner_field, key_field = "authorizing_owner", "authorizing_id"
base_qs = annotate_authorizing_fields(base_qs, parent_field)
unowned = Q(**{f"{owner_field}__isnull": True})
if user is None or not getattr(user, "is_authenticated", False): if user is None or not getattr(user, "is_authenticated", False):
return base_qs.filter(owner__isnull=True).values_list("id", flat=True) return base_qs.filter(unowned).values_list("id", flat=True)
# Deactivated users get nothing, deactivated superusers included, so this # Deactivated users get nothing, deactivated superusers included, so this
# has to come before the superuser shortcut. guardian's # has to come before the superuser shortcut. guardian's
@@ -389,26 +415,50 @@ def permitted_object_ids(
"permission__content_type": content_type, "permission__content_type": content_type,
} }
user_perm_ids = ( key_as_text = Cast(OuterRef(key_field), CharField(max_length=64))
UserObjectPermission.objects.filter(user=user, **perm_filter) granted_to_user = Exists(
.annotate(object_pk_int=Cast("object_pk", IntegerField())) UserObjectPermission.objects.filter(
.values_list("object_pk_int", flat=True) user=user,
object_pk=key_as_text,
**perm_filter,
),
) )
group_perm_ids = ( granted_to_group = Exists(
GroupObjectPermission.objects.filter(group__user=user, **perm_filter) GroupObjectPermission.objects.filter(
.annotate(object_pk_int=Cast("object_pk", IntegerField())) group_id__in=user.groups.values("id"),
.values_list("object_pk_int", flat=True) object_pk=key_as_text,
**perm_filter,
),
) )
permitted_ids = user_perm_ids.union(group_perm_ids)
return base_qs.filter( return base_qs.filter(
Q(owner=user) | Q(owner__isnull=True) | Q(id__in=permitted_ids), Q(**{owner_field: user.pk}) | unowned | granted_to_user | granted_to_group,
).values_list("id", flat=True) ).values_list("id", flat=True)
ModelT = TypeVar("ModelT", bound=Model) ModelT = TypeVar("ModelT", bound=Model)
def annotate_authorizing_fields(
queryset: QuerySet[ModelT],
parent_field: str,
) -> QuerySet[ModelT]:
"""
Annotate each row with ``authorizing_id`` and ``authorizing_owner``: the id
and owner of the row that authorizes it. A row with a parent (the
self-referencing foreign key ``parent_field``) is authorized by its parent,
any other row by itself.
"""
return queryset.annotate(
authorizing_id=Coalesce(f"{parent_field}_id", "id"),
authorizing_owner=Case(
When(**{f"{parent_field}_id__isnull": True}, then=F("owner_id")),
default=F(f"{parent_field}__owner_id"),
output_field=IntegerField(),
),
)
def user_is_unrestricted(user: User | None) -> bool: def user_is_unrestricted(user: User | None) -> bool:
""" """
True when ``user`` means "no restriction at all" (an absent user, or an True when ``user`` means "no restriction at all" (an absent user, or an
@@ -471,30 +521,16 @@ def permitted_document_ids(
``include_deleted=True`` for callers that need to check permission on ``include_deleted=True`` for callers that need to check permission on
soft-deleted documents (e.g. trash restore). This intentionally avoids soft-deleted documents (e.g. trash restore). This intentionally avoids
``get_objects_for_user`` to keep the subquery small and index-friendly. ``get_objects_for_user`` to keep the subquery small and index-friendly.
"""
return permitted_object_ids(user, Document, perm, include_deleted=include_deleted)
A version is authorized by its root document, so a version's own owner and
def documents_without_permitted_root( grants never matter.
documents: QuerySet[Document],
user: User | None,
*,
perm: str = "view_document",
include_deleted: bool = False,
) -> QuerySet[Document]:
""" """
The documents the user lacks ``perm`` on. Versions are authorized by their return permitted_object_ids(
root document, so a version's own owner is ignored. A single query, without
loading the documents or joining the root.
"""
return documents.annotate(
root_id=Coalesce("root_document_id", "id"),
).exclude(
root_id__in=permitted_document_ids(
user, user,
perm=perm, Document,
perm,
include_deleted=include_deleted, include_deleted=include_deleted,
), parent_field="root_document",
) )
@@ -654,7 +690,14 @@ def has_perms_owner_aware(user, perms, obj):
single-object check still has many production callers. Several callers single-object check still has many production callers. Several callers
remain across ``documents/``, ``paperless_mail/``, and ``paperless_ai/`` remain across ``documents/``, ``paperless_mail/``, and ``paperless_ai/``
-- grep for this function name before removing it. -- grep for this function name before removing it.
A document version is authorized by its root document, like in
``permitted_document_ids``, so a version's own owner and grants never
matter. Fetch the root with ``select_related("root_document__owner")`` to
avoid extra queries.
""" """
if isinstance(obj, Document):
obj = get_root_document(obj)
checker = ObjectPermissionChecker(user) checker = ObjectPermissionChecker(user)
return obj.owner is None or obj.owner == user or checker.has_perm(perms, obj) return obj.owner is None or obj.owner == user or checker.has_perm(perms, obj)
+9 -2
View File
@@ -284,9 +284,14 @@ class WriteBatch:
and adding the new version. This ensures stale document data (e.g., after and adding the new version. This ensures stale document data (e.g., after
permission changes) doesn't persist in the index. permission changes) doesn't persist in the index.
Only root documents are indexed, with their effective content, so a
version is indexed as its root document.
Args: Args:
document: Django Document instance to index document: Django Document instance to index
""" """
if document.root_document_id is not None:
document = document.root_document
self.remove(document.pk) self.remove(document.pk)
doc = self._backend._build_tantivy_doc(document) doc = self._backend._build_tantivy_doc(document)
self._writer.add_document(doc) self._writer.add_document(doc)
@@ -311,20 +316,22 @@ class WriteBatch:
An id with no matching document (e.g. deleted between the caller An id with no matching document (e.g. deleted between the caller
collecting ids and the batch running) is silently skipped, matching collecting ids and the batch running) is silently skipped, matching
``add_or_update()``'s existing single-document deferred-task behavior ``add_or_update()``'s existing single-document deferred-task behavior
rather than erroring or leaving a stale index entry. rather than erroring or leaving a stale index entry. The id of a
version stands for its root document.
Args: Args:
ids: Primary keys of Document instances to index ids: Primary keys of Document instances to index
""" """
from documents.models import Document from documents.models import Document
from documents.versioning import annotate_effective_content from documents.versioning import annotate_effective_content
from documents.versioning import root_document_ids
ids = list(ids) ids = list(ids)
if not ids: if not ids:
return return
queryset = annotate_effective_content( queryset = annotate_effective_content(
Document.objects.filter(pk__in=ids) Document.objects.filter(pk__in=root_document_ids(ids))
.select_related("correspondent", "document_type", "storage_path", "owner") .select_related("correspondent", "document_type", "storage_path", "owner")
.prefetch_related( .prefetch_related(
"tags", "tags",
+1 -2
View File
@@ -90,7 +90,6 @@ from documents.templating.utils import convert_format_str_to_template_format
from documents.templating.workflows import validate_workflow_template from documents.templating.workflows import validate_workflow_template
from documents.validators import uri_validator from documents.validators import uri_validator
from documents.validators import url_validator from documents.validators import url_validator
from documents.versioning import get_root_document
from documents.versioning import has_prefetched_effective_content from documents.versioning import has_prefetched_effective_content
from documents.versioning import sort_versions_newest_first from documents.versioning import sort_versions_newest_first
@@ -2895,7 +2894,7 @@ class ShareLinkSerializer(OwnedObjectSerializer):
and has_perms_owner_aware( and has_perms_owner_aware(
self.user, self.user,
"view_document", "view_document",
get_root_document(document), document,
) )
): ):
return document return document
+5 -8
View File
@@ -490,23 +490,20 @@ def update_document_content_maybe_archive_file(
shutil.move(thumbnail, document.thumbnail_path) shutil.move(thumbnail, document.thumbnail_path)
document.refresh_from_db() document.refresh_from_db()
root_document = (
document.root_document if document.root_document_id else document
)
logger.info( logger.info(
f"Updating index for document {root_document.pk} ({document.archive_checksum})", f"Updating index for document {document_id} ({document.archive_checksum})",
) )
from documents.search import get_backend from documents.search import get_backend
get_backend().add_or_update(root_document) get_backend().add_or_update(document)
ai_config = AIConfig() ai_config = AIConfig()
if ai_config.llm_index_enabled: if ai_config.llm_index_enabled:
llm_index_add_or_update_document(root_document) llm_index_add_or_update_document(document)
clear_document_caches(document.pk) clear_document_caches(document.pk)
if root_document.pk != document.pk: if document.root_document_id is not None:
clear_document_caches(root_document.pk) clear_document_caches(document.root_document_id)
except Exception: except Exception:
logger.exception( logger.exception(
@@ -293,6 +293,80 @@ class TestAddOrUpdateIds:
assert backend.search_ids("updated", user=None) == [doc.pk] assert backend.search_ids("updated", user=None) == [doc.pk]
class TestVersionsAreIndexedAsTheirRoot:
"""Only root documents are indexed, with their effective content, so
every write path that is handed a version indexes its root instead."""
@staticmethod
def _root_with_version() -> tuple[Document, Document]:
root = DocumentFactory(title="Statement", content="stale text")
version = DocumentFactory(
title="Statement",
content="latest text",
root_document=root,
version_index=1,
)
return root, version
def test_add_or_update_indexes_the_root_of_a_version(
self,
backend: TantivyBackend,
) -> None:
"""
GIVEN:
- A root document with a version
WHEN:
- The version is passed to add_or_update
THEN:
- The root is indexed with the version's text, and the version is not
"""
root, version = self._root_with_version()
backend.add_or_update(version)
assert backend.search_ids("latest", user=None) == [root.pk]
assert backend.search_ids("stale", user=None) == []
def test_add_or_update_ids_indexes_each_root_once(
self,
backend: TantivyBackend,
) -> None:
"""
GIVEN:
- A root document with a version
WHEN:
- Both ids are passed to add_or_update_ids
THEN:
- The root is indexed once and the version is not indexed
"""
root, version = self._root_with_version()
with backend.batch_update() as batch:
batch.add_or_update_ids([version.pk, root.pk])
assert backend.search_ids("Statement", user=None) == [root.pk]
assert backend.search_ids("latest", user=None) == [root.pk]
def test_add_or_update_ids_resolves_a_lone_version(
self,
backend: TantivyBackend,
) -> None:
"""
GIVEN:
- A root document with a version
WHEN:
- Only the version's id is passed to add_or_update_ids
THEN:
- The root is indexed
"""
root, version = self._root_with_version()
with backend.batch_update() as batch:
batch.add_or_update_ids([version.pk])
assert backend.search_ids("latest", user=None) == [root.pk]
class TestSearch: class TestSearch:
"""Test search query parsing and matching via search_ids.""" """Test search query parsing and matching via search_ids."""
+53
View File
@@ -1196,6 +1196,59 @@ class TestDocumentSearchApi(DirectoriesMixin, APITestCase):
self.assertIn(d3.id, result_ids) self.assertIn(d3.id, result_ids)
self.assertNotIn(d4.id, result_ids) self.assertNotIn(d4.id, result_ids)
def test_search_more_like_version_uses_its_root(self) -> None:
"""
GIVEN:
- A document similar in content to a root document, and one that is not
- A version of the root document, which is never indexed
WHEN:
- API request for more like the version
THEN:
- The documents similar to the root are returned, not the version
"""
indexed = {}
for name, title, content, day in (
("root", "bank statement 1", "things i paid for in august", (2019, 3, 4)),
(
"similar",
"bank statement 3",
"things i paid for in september",
(2020, 7, 9),
),
(
"other",
"Quarterly Report",
"quarterly revenue profit margin",
(2021, 11, 30),
),
):
with time_machine.travel(
timezone.make_aware(datetime.datetime(*day)),
tick=False,
):
indexed[name] = DocumentFactory(
title=title,
content=content,
created=datetime.date(*day),
added=timezone.make_aware(datetime.datetime(*day)),
)
version = DocumentFactory(
root_document=indexed["root"],
version_index=1,
content="things i paid for in august",
)
backend = get_backend()
for document in indexed.values():
backend.add_or_update(document)
response = self.client.get(f"/api/documents/?more_like_id={version.id}")
self.assertEqual(response.status_code, status.HTTP_200_OK)
result_ids = [r["id"] for r in response.data["results"]]
self.assertIn(indexed["similar"].id, result_ids)
self.assertNotIn(indexed["other"].id, result_ids)
self.assertNotIn(version.id, result_ids)
def test_more_like_requires_id_of_existing_document(self) -> None: def test_more_like_requires_id_of_existing_document(self) -> None:
""" """
GIVEN: GIVEN:
+22
View File
@@ -22,6 +22,7 @@ from documents.models import Document
from documents.tasks import update_document_content_maybe_archive_file from documents.tasks import update_document_content_maybe_archive_file
from paperless_testing.assertions import FileSystemAssertsMixin from paperless_testing.assertions import FileSystemAssertsMixin
from paperless_testing.dirs import DirectoriesMixin from paperless_testing.dirs import DirectoriesMixin
from paperless_testing.factories import DocumentFactory
sample_file: Path = Path(__file__).parent / "samples" / "simple.pdf" sample_file: Path = Path(__file__).parent / "samples" / "simple.pdf"
@@ -116,6 +117,27 @@ class TestMakeIndex:
call_command("document_index", "reindex", skip_checks=True) call_command("document_index", "reindex", skip_checks=True)
mock_get_backend.return_value.rebuild.assert_called_once() mock_get_backend.return_value.rebuild.assert_called_once()
def test_reindex_skips_versions(self, mocker: MockerFixture) -> None:
"""
GIVEN:
- A root document with a version
WHEN:
- The reindex command runs
THEN:
- Only the root document is handed to the rebuild, since a version
is indexed as its root
"""
root = DocumentFactory()
DocumentFactory(root_document=root, version_index=1)
mock_get_backend = mocker.patch(
"documents.management.commands.document_index.get_backend",
)
call_command("document_index", "reindex", skip_checks=True)
documents = mock_get_backend.return_value.rebuild.call_args.args[0]
assert list(documents.values_list("pk", flat=True)) == [root.pk]
def test_optimize(self) -> None: def test_optimize(self) -> None:
"""Optimize command must execute without error (Tantivy handles optimization automatically).""" """Optimize command must execute without error (Tantivy handles optimization automatically)."""
call_command("document_index", "optimize", skip_checks=True) call_command("document_index", "optimize", skip_checks=True)
@@ -18,6 +18,7 @@ from documents.models import Correspondent
from documents.models import DocumentType from documents.models import DocumentType
from documents.models import StoragePath from documents.models import StoragePath
from documents.models import Tag from documents.models import Tag
from documents.permissions import has_perms_owner_aware
from documents.permissions import permitted_document_ids from documents.permissions import permitted_document_ids
from documents.permissions import permitted_object_ids from documents.permissions import permitted_object_ids
from documents.permissions import restrict_queryset_to_visible from documents.permissions import restrict_queryset_to_visible
@@ -32,6 +33,8 @@ from paperless_testing.permissions import grant_global
from paperless_testing.permissions import grant_object from paperless_testing.permissions import grant_object
if TYPE_CHECKING: if TYPE_CHECKING:
from django.contrib.auth.models import User
from paperless_testing.dirs import PaperlessDirs from paperless_testing.dirs import PaperlessDirs
@@ -178,6 +181,382 @@ class TestPermittedDocumentIdsIncludeDeleted:
) )
@pytest.mark.django_db
class TestPermittedDocumentIdsVersions:
"""
A version is authorized by its root document: the version's own owner and
grants never matter.
"""
@pytest.mark.parametrize(
("root_owner", "version_owner", "expected_visible"),
[
pytest.param(
"other",
"nobody",
False,
id="unowned-version-of-private-root",
),
pytest.param("other", "user", False, id="own-version-of-private-root"),
pytest.param("user", "other", True, id="foreign-version-of-own-root"),
pytest.param("user", "nobody", True, id="unowned-version-of-own-root"),
pytest.param("nobody", "other", True, id="private-version-of-unowned-root"),
],
)
def test_version_follows_root_owner(
self,
root_owner: str,
version_owner: str,
*,
expected_visible: bool,
) -> None:
"""
GIVEN:
- A root document and a version with differing owners
WHEN:
- The permitted document ids are resolved for the user
THEN:
- The version is visible exactly when its root is
"""
user = UserFactory()
owners = {"user": user, "other": UserFactory(), "nobody": None}
root = DocumentFactory(owner=owners[root_owner])
version = DocumentFactory(root_document=root, owner=owners[version_owner])
visible = set(permitted_document_ids(user))
assert (version.pk in visible) is expected_visible
assert (root.pk in visible) is expected_visible
@staticmethod
def grantee(user: User, kind: str) -> User | Group:
"""The user itself, or a new group the user belongs to."""
if kind == "user":
return user
group = Group.objects.create(name="shared")
user.groups.add(group)
return group
@pytest.mark.parametrize(
"grantee_kind",
[pytest.param("user", id="user"), pytest.param("group", id="group")],
)
def test_grant_on_root_applies_to_version(self, grantee_kind: str) -> None:
"""
GIVEN:
- A private root document shared with a user or one of their groups
- A version of it owned by someone else
WHEN:
- The permitted document ids are resolved for the user
THEN:
- Both the root and the version are visible
- A user without the grant sees neither
"""
user = UserFactory()
stranger = UserFactory()
root = DocumentFactory(owner=UserFactory())
version = DocumentFactory(root_document=root, owner=UserFactory())
grant_object(self.grantee(user, grantee_kind), root, "view_document")
assert_visible_document_ids(
permitted_document_ids(user),
expected_visible=[root.pk, version.pk],
expected_hidden=[],
)
assert_visible_document_ids(
permitted_document_ids(stranger),
expected_visible=[],
expected_hidden=[root.pk, version.pk],
)
@pytest.mark.parametrize(
"grantee_kind",
[pytest.param("user", id="user"), pytest.param("group", id="group")],
)
def test_grant_on_version_is_ignored(self, grantee_kind: str) -> None:
"""
GIVEN:
- A private root document
- A version with an explicit grant for the user or one of their groups
WHEN:
- The permitted document ids are resolved for the user
THEN:
- Neither the root nor the version is visible
"""
user = UserFactory()
root = DocumentFactory(owner=UserFactory())
version = DocumentFactory(root_document=root, owner=UserFactory())
grant_object(self.grantee(user, grantee_kind), version, "view_document")
assert_visible_document_ids(
permitted_document_ids(user),
expected_visible=[],
expected_hidden=[root.pk, version.pk],
)
def test_grant_on_one_root_does_not_reach_another_roots_version(self) -> None:
"""
GIVEN:
- Two private roots, each with a version
- The user may view only the first root
WHEN:
- The permitted document ids are resolved for the user
THEN:
- Only the first root and its version are visible
"""
user = UserFactory()
first = DocumentFactory(owner=UserFactory())
first_version = DocumentFactory(root_document=first, owner=UserFactory())
second = DocumentFactory(owner=UserFactory())
second_version = DocumentFactory(root_document=second, owner=user)
grant_object(user, first, "view_document")
assert_visible_document_ids(
permitted_document_ids(user),
expected_visible=[first.pk, first_version.pk],
expected_hidden=[second.pk, second_version.pk],
)
def test_user_in_several_groups(self) -> None:
"""
GIVEN:
- A user in two groups
- Two private roots shared with one group each, and a third shared with nobody
- A version of each root
WHEN:
- The permitted document ids are resolved for the user
THEN:
- The two shared roots and their versions are visible
- The third root and its version are not
"""
user = UserFactory()
groups = [Group.objects.create(name=f"group{i}") for i in range(2)]
user.groups.add(*groups)
shared = [DocumentFactory(owner=UserFactory()) for _ in groups]
for root, group in zip(shared, groups, strict=True):
grant_object(group, root, "view_document")
unshared = DocumentFactory(owner=UserFactory())
shared_versions = [
DocumentFactory(root_document=root, owner=UserFactory()) for root in shared
]
unshared_version = DocumentFactory(root_document=unshared, owner=None)
assert_visible_document_ids(
permitted_document_ids(user),
expected_visible=[
*(root.pk for root in shared),
*(version.pk for version in shared_versions),
],
expected_hidden=[unshared.pk, unshared_version.pk],
)
def test_permission_is_resolved_through_the_root(self) -> None:
"""
GIVEN:
- A private root document where the user may view and change
WHEN:
- The permitted ids are resolved for view, change and delete
THEN:
- The version is visible for view and change only
"""
user = UserFactory()
root = DocumentFactory(owner=UserFactory())
version = DocumentFactory(root_document=root, owner=UserFactory())
grant_object(user, root, "view_document", "change_document")
assert version.pk in set(permitted_document_ids(user))
assert version.pk in set(permitted_document_ids(user, perm="change_document"))
assert version.pk in set(
permitted_document_ids(user, perm="documents.change_document"),
)
assert version.pk not in set(
permitted_document_ids(user, perm="delete_document"),
)
def test_anonymous_sees_versions_of_unowned_roots_only(self) -> None:
"""
GIVEN:
- A version owned by nobody under a private root
- A version owned by someone under an unowned root
WHEN:
- The permitted document ids are resolved for an anonymous user
THEN:
- Only the version of the unowned root is visible
"""
private_root = DocumentFactory(owner=UserFactory())
private_version = DocumentFactory(root_document=private_root, owner=None)
open_root = DocumentFactory(owner=None)
open_version = DocumentFactory(root_document=open_root, owner=UserFactory())
assert_visible_document_ids(
permitted_document_ids(AnonymousUser()),
expected_visible=[open_root.pk, open_version.pk],
expected_hidden=[private_root.pk, private_version.pk],
)
def test_deleted_versions_follow_their_deleted_root(self) -> None:
"""
GIVEN:
- A soft-deleted root document and its version, which deleting the
root soft-deletes too; the version is owned by someone else
WHEN:
- The permitted document ids are resolved with and without deleted
documents
THEN:
- Nothing is visible by default
- With deleted documents included, the version is visible to the
root's owner and not to the version's own owner
"""
owner = UserFactory()
version_owner = UserFactory()
root = DocumentFactory(owner=owner)
version = DocumentFactory(root_document=root, owner=version_owner)
root.delete()
assert not {root.pk, version.pk} & set(permitted_document_ids(owner))
assert_visible_document_ids(
permitted_document_ids(owner, include_deleted=True),
expected_visible=[root.pk, version.pk],
expected_hidden=[],
)
assert_visible_document_ids(
permitted_document_ids(version_owner, include_deleted=True),
expected_visible=[],
expected_hidden=[root.pk, version.pk],
)
@pytest.mark.parametrize(
"is_superuser",
[
pytest.param(False, id="regular-user"),
pytest.param(True, id="superuser"),
],
)
def test_inactive_user_sees_no_versions(self, *, is_superuser: bool) -> None:
"""
GIVEN:
- An inactive user, possibly a superuser, who owns a root and its version
WHEN:
- The permitted document ids are resolved for them
THEN:
- Nothing is visible
"""
user = UserFactory(is_active=False, is_superuser=is_superuser)
root = DocumentFactory(owner=user)
version = DocumentFactory(root_document=root, owner=user)
assert_visible_document_ids(
permitted_document_ids(user),
expected_visible=[],
expected_hidden=[root.pk, version.pk],
)
def test_superuser_sees_all_versions(self) -> None:
"""
GIVEN:
- A private root owned by someone else, with a version
WHEN:
- The permitted document ids are resolved for a superuser
THEN:
- Both the root and the version are visible
"""
superuser = UserFactory(superuser=True)
root = DocumentFactory(owner=UserFactory())
version = DocumentFactory(root_document=root, owner=UserFactory())
assert_visible_document_ids(
permitted_document_ids(superuser),
expected_visible=[root.pk, version.pk],
expected_hidden=[],
)
@pytest.mark.django_db
class TestHasPermsOwnerAwareVersions:
"""
The single-object check agrees with permitted_document_ids: a version is
authorized by its root document.
"""
@pytest.mark.parametrize(
("root_owner", "version_owner", "expected"),
[
pytest.param(
"other",
"nobody",
False,
id="unowned-version-of-private-root",
),
pytest.param("other", "user", False, id="own-version-of-private-root"),
pytest.param("user", "other", True, id="foreign-version-of-own-root"),
pytest.param("nobody", "other", True, id="private-version-of-unowned-root"),
],
)
def test_version_follows_root_owner(
self,
root_owner: str,
version_owner: str,
*,
expected: bool,
) -> None:
"""
GIVEN:
- A root document and a version with differing owners
WHEN:
- The single-object check runs for the version
THEN:
- The version is allowed exactly when its root is
"""
user = UserFactory()
owners = {"user": user, "other": UserFactory(), "nobody": None}
root = DocumentFactory(owner=owners[root_owner])
version = DocumentFactory(root_document=root, owner=owners[version_owner])
assert has_perms_owner_aware(user, "view_document", version) is expected
assert has_perms_owner_aware(user, "view_document", root) is expected
def test_grant_on_root_applies_and_grant_on_version_does_not(self) -> None:
"""
GIVEN:
- A private root with a version, and a second private root with a version
- The user may change only the first root, and was granted the second
root's version directly
WHEN:
- The single-object check runs for each version
THEN:
- Only the first root's version is allowed
"""
user = UserFactory()
shared_root = DocumentFactory(owner=UserFactory())
shared_version = DocumentFactory(root_document=shared_root, owner=UserFactory())
private_root = DocumentFactory(owner=UserFactory())
private_version = DocumentFactory(
root_document=private_root,
owner=UserFactory(),
)
grant_object(user, shared_root, "change_document")
grant_object(user, private_version, "change_document")
assert has_perms_owner_aware(user, "change_document", shared_version)
assert not has_perms_owner_aware(user, "change_document", private_version)
def test_other_models_use_their_own_owner(self) -> None:
"""
GIVEN:
- A tag owned by someone else, and one owned by the user
WHEN:
- The single-object check runs for each
THEN:
- Only the user's own tag is allowed without a grant
"""
user = UserFactory()
mine = TagFactory(owner=user)
theirs = TagFactory(owner=UserFactory())
assert has_perms_owner_aware(user, "view_tag", mine)
assert not has_perms_owner_aware(user, "view_tag", theirs)
@pytest.mark.django_db @pytest.mark.django_db
class TestAiChatAllDocumentsPermissionBoundary: class TestAiChatAllDocumentsPermissionBoundary:
""" """
+7 -9
View File
@@ -310,7 +310,7 @@ class TestUpdateContent(DirectoriesMixin, TestCase):
@mock.patch("documents.tasks.clear_document_caches") @mock.patch("documents.tasks.clear_document_caches")
@mock.patch("documents.search.get_backend") @mock.patch("documents.search.get_backend")
def test_update_content_version_indexes_root( def test_update_content_version_clears_caches_for_root(
self, self,
mock_get_backend: mock.Mock, mock_get_backend: mock.Mock,
mock_clear_caches: mock.Mock, mock_clear_caches: mock.Mock,
@@ -321,8 +321,8 @@ class TestUpdateContent(DirectoriesMixin, TestCase):
WHEN: WHEN:
- Update content task is called for the version - Update content task is called for the version
THEN: THEN:
- The version's content is updated - The version's content is updated, not the root's
- The root document is indexed rather than the version - The document is indexed
- Caches are cleared for both - Caches are cleared for both
""" """
root, version = self._create_root_with_version() root, version = self._create_root_with_version()
@@ -334,8 +334,7 @@ class TestUpdateContent(DirectoriesMixin, TestCase):
"my document", "my document",
) )
self.assertEqual(Document.objects.get(pk=root.pk).content, "root content") self.assertEqual(Document.objects.get(pk=root.pk).content, "root content")
indexed = mock_get_backend.return_value.add_or_update.call_args.args[0] mock_get_backend.return_value.add_or_update.assert_called_once()
self.assertEqual(indexed.pk, root.pk)
mock_clear_caches.assert_has_calls( mock_clear_caches.assert_has_calls(
[mock.call(version.pk), mock.call(root.pk)], [mock.call(version.pk), mock.call(root.pk)],
) )
@@ -343,7 +342,7 @@ class TestUpdateContent(DirectoriesMixin, TestCase):
@override_settings(AI_ENABLED=True, LLM_EMBEDDING_BACKEND="huggingface") @override_settings(AI_ENABLED=True, LLM_EMBEDDING_BACKEND="huggingface")
@mock.patch("documents.tasks.llm_index_add_or_update_document") @mock.patch("documents.tasks.llm_index_add_or_update_document")
@mock.patch("documents.search.get_backend") @mock.patch("documents.search.get_backend")
def test_update_content_version_updates_llm_index_for_root( def test_update_content_version_updates_llm_index(
self, self,
mock_get_backend: mock.Mock, mock_get_backend: mock.Mock,
mock_llm_index: mock.Mock, mock_llm_index: mock.Mock,
@@ -355,14 +354,13 @@ class TestUpdateContent(DirectoriesMixin, TestCase):
WHEN: WHEN:
- Update content task is called for the version - Update content task is called for the version
THEN: THEN:
- The LLM index is updated for the root document, not the version - The LLM index is updated
""" """
root, version = self._create_root_with_version() _, version = self._create_root_with_version()
tasks.update_document_content_maybe_archive_file(version.pk) tasks.update_document_content_maybe_archive_file(version.pk)
mock_llm_index.assert_called_once() mock_llm_index.assert_called_once()
self.assertEqual(mock_llm_index.call_args.args[0].pk, root.pk)
class TestUpdateContentRemoteOCR(DirectoriesMixin, TestCase): class TestUpdateContentRemoteOCR(DirectoriesMixin, TestCase):
+17
View File
@@ -17,9 +17,26 @@ from django.db.models.functions import RowNumber
from documents.models import Document from documents.models import Document
if TYPE_CHECKING: if TYPE_CHECKING:
from collections.abc import Iterable
from rest_framework.request import Request from rest_framework.request import Request
def root_document_ids(ids: Iterable[int]) -> QuerySet[int]:
"""
The ids of the root documents of the given documents: a root stands for
itself and a version for its root. Only the indexes' bookkeeping needs
this, since they hold root documents only.
"""
return (
Document.objects.filter(pk__in=ids)
.annotate(root_id=Coalesce("root_document_id", "id"))
.order_by()
.values_list("root_id", flat=True)
.distinct()
)
def versions_newest_first(documents: QuerySet[Document]) -> QuerySet[Document]: def versions_newest_first(documents: QuerySet[Document]) -> QuerySet[Document]:
""" """
Sorts versions so the newest one comes first using version_index and not on id, Sorts versions so the newest one comes first using version_index and not on id,
+27 -50
View File
@@ -174,7 +174,6 @@ from documents.permissions import TrashPermissions
from documents.permissions import ViewDocumentsPermissions from documents.permissions import ViewDocumentsPermissions
from documents.permissions import annotate_document_count_by_ids from documents.permissions import annotate_document_count_by_ids
from documents.permissions import annotate_document_count_for_related_queryset from documents.permissions import annotate_document_count_for_related_queryset
from documents.permissions import documents_without_permitted_root
from documents.permissions import get_document_count_filter_for_user from documents.permissions import get_document_count_filter_for_user
from documents.permissions import get_objects_for_user_owner_aware from documents.permissions import get_objects_for_user_owner_aware
from documents.permissions import has_global_statistics_permission from documents.permissions import has_global_statistics_permission
@@ -330,9 +329,10 @@ def _get_tantivy_query_and_mode(params):
def _get_more_like_id(query_params: dict[str, Any], user: User | None) -> int: def _get_more_like_id(query_params: dict[str, Any], user: User | None) -> int:
try: try:
more_like_doc_id = int(query_params["more_like_id"]) more_like_doc_id = int(query_params["more_like_id"])
more_like_doc = Document.objects.select_related("owner").get( more_like_doc = Document.objects.select_related(
pk=more_like_doc_id, "owner",
) "root_document__owner",
).get(pk=more_like_doc_id)
except (TypeError, ValueError, Document.DoesNotExist): except (TypeError, ValueError, Document.DoesNotExist):
raise PermissionDenied(_("Invalid more_like_id")) raise PermissionDenied(_("Invalid more_like_id"))
@@ -343,7 +343,8 @@ def _get_more_like_id(query_params: dict[str, Any], user: User | None) -> int:
): ):
raise PermissionDenied(_("Insufficient permissions.")) raise PermissionDenied(_("Insufficient permissions."))
return more_like_doc_id # Only root documents are indexed, a version stands for its root
return more_like_doc.root_document_id or more_like_doc.pk
class SearchParams(NamedTuple): class SearchParams(NamedTuple):
@@ -1560,7 +1561,7 @@ class DocumentViewSet(
if request.user is not None and not has_perms_owner_aware( if request.user is not None and not has_perms_owner_aware(
request.user, request.user,
"change_document", "change_document",
get_root_document(doc), doc,
): ):
return HttpResponseForbidden("Insufficient permissions") return HttpResponseForbidden("Insufficient permissions")
@@ -1623,7 +1624,7 @@ class DocumentViewSet(
if request.user is not None and not has_perms_owner_aware( if request.user is not None and not has_perms_owner_aware(
request.user, request.user,
"change_document", "change_document",
get_root_document(doc), doc,
): ):
return HttpResponseForbidden("Insufficient permissions") return HttpResponseForbidden("Insufficient permissions")
@@ -1876,7 +1877,7 @@ class DocumentViewSet(
if currentUser is not None and not has_perms_owner_aware( if currentUser is not None and not has_perms_owner_aware(
currentUser, currentUser,
"view_document", "view_document",
get_root_document(doc), doc,
): ):
return HttpResponseForbidden("Insufficient permissions to view notes") return HttpResponseForbidden("Insufficient permissions to view notes")
except Document.DoesNotExist: except Document.DoesNotExist:
@@ -1898,7 +1899,7 @@ class DocumentViewSet(
if currentUser is not None and not has_perms_owner_aware( if currentUser is not None and not has_perms_owner_aware(
currentUser, currentUser,
"change_document", "change_document",
get_root_document(doc), doc,
): ):
return HttpResponseForbidden( return HttpResponseForbidden(
"Insufficient permissions to create notes", "Insufficient permissions to create notes",
@@ -1941,7 +1942,7 @@ class DocumentViewSet(
if currentUser is not None and not has_perms_owner_aware( if currentUser is not None and not has_perms_owner_aware(
currentUser, currentUser,
"change_document", "change_document",
get_root_document(doc), doc,
): ):
return HttpResponseForbidden("Insufficient permissions to delete notes") return HttpResponseForbidden("Insufficient permissions to delete notes")
@@ -1991,7 +1992,7 @@ class DocumentViewSet(
if currentUser is not None and not has_perms_owner_aware( if currentUser is not None and not has_perms_owner_aware(
currentUser, currentUser,
"change_document", "change_document",
get_root_document(doc), doc,
): ):
return HttpResponseForbidden( return HttpResponseForbidden(
"Insufficient permissions to add share link", "Insufficient permissions to add share link",
@@ -2117,7 +2118,7 @@ class DocumentViewSet(
documents = Document.objects.filter(pk__in=document_ids) documents = Document.objects.filter(pk__in=document_ids)
if ( if (
request.user is not None request.user is not None
and documents_without_permitted_root(documents, request.user).exists() and documents.exclude(id__in=permitted_document_ids(request.user)).exists()
): ):
return HttpResponseForbidden("Insufficient permissions") return HttpResponseForbidden("Insufficient permissions")
@@ -2452,7 +2453,7 @@ class ChatStreamingView(GenericAPIView[Any]):
if not has_perms_owner_aware( if not has_perms_owner_aware(
request.user, request.user,
"view_document", "view_document",
get_root_document(document), document,
): ):
return HttpResponseForbidden("Insufficient permissions") return HttpResponseForbidden("Insufficient permissions")
@@ -3009,12 +3010,8 @@ class DocumentOperationPermissionMixin(PassUserMixin, DocumentSelectionMixin):
user.has_perm( user.has_perm(
"documents.change_document", "documents.change_document",
) )
and not Document.global_objects.filter( and not Document.global_objects.filter(pk__in=documents)
pk__in=[doc.pk for doc in root_docs], .exclude(pk__in=permitted_document_ids(user, perm="change_document"))
)
.exclude(
pk__in=permitted_document_ids(user, perm="change_document"),
)
.exists() .exists()
) )
@@ -3627,7 +3624,7 @@ class SelectionDataView(DocumentSelectionMixin, GenericAPIView[Any]):
documents = Document.objects.filter(pk__in=ids) documents = Document.objects.filter(pk__in=ids)
if ( if (
documents.count() != len(ids) documents.count() != len(ids)
or documents_without_permitted_root(documents, request.user).exists() or documents.exclude(id__in=permitted_document_ids(request.user)).exists()
): ):
return HttpResponseForbidden("Insufficient permissions") return HttpResponseForbidden("Insufficient permissions")
@@ -4124,21 +4121,16 @@ class BulkDownloadView(DocumentSelectionMixin, GenericAPIView[Any]):
validated_data=serializer.validated_data, validated_data=serializer.validated_data,
) )
documents = Document.objects.filter(pk__in=ids) documents = Document.objects.filter(pk__in=ids)
versioned_documents = []
compression = serializer.validated_data.get("compression") compression = serializer.validated_data.get("compression")
content = serializer.validated_data.get("content") content = serializer.validated_data.get("content")
follow_filename_format = serializer.validated_data.get("follow_formatting") follow_filename_format = serializer.validated_data.get("follow_formatting")
permitted_ids = set(permitted_document_ids(request.user)) if documents.exclude(id__in=permitted_document_ids(request.user)).exists():
for document in documents:
root_doc = get_root_document(document)
if root_doc.pk not in permitted_ids:
return HttpResponseForbidden("Insufficient permissions") return HttpResponseForbidden("Insufficient permissions")
versioned_documents.append( versioned_documents = [
get_latest_version_for_root( get_latest_version_for_root(get_root_document(document))
root_doc, for document in documents
), ]
)
if content == "both": if content == "both":
strategy_class = OriginalAndArchiveStrategy strategy_class = OriginalAndArchiveStrategy
@@ -4811,7 +4803,7 @@ class ShareLinkBundleViewSet(PassUserMixin, ModelViewSet[ShareLinkBundle]):
) )
denied_id = ( denied_id = (
documents_without_permitted_root(documents_qs, request.user) documents_qs.exclude(id__in=permitted_document_ids(request.user))
.order_by("pk") .order_by("pk")
.values_list("pk", flat=True) .values_list("pk", flat=True)
.first() .first()
@@ -5610,23 +5602,7 @@ class TrashView(ListModelMixin, PassUserMixin):
class _TrashPermittedObjectsFilter(PermittedObjectsFilter): class _TrashPermittedObjectsFilter(PermittedObjectsFilter):
include_granted = False include_granted = False
parent_field = "root_document"
def filter_queryset(self, request, queryset, view):
if request.user.is_superuser or not request.user.is_active:
return super().filter_queryset(request, queryset, view)
# A version belongs to whoever owns its root
def owned_or_unowned(prefix: str) -> Q:
return Q(**{f"{prefix}owner": request.user}) | Q(
**{f"{prefix}owner__isnull": True},
)
return queryset.filter(
(Q(root_document__isnull=True) & owned_or_unowned(""))
| (
Q(root_document__isnull=False) & owned_or_unowned("root_document__")
),
)
filter_backends = (_TrashPermittedObjectsFilter,) filter_backends = (_TrashPermittedObjectsFilter,)
pagination_class = StandardPagination pagination_class = StandardPagination
@@ -5657,11 +5633,12 @@ class TrashView(ListModelMixin, PassUserMixin):
if doc_ids is not None if doc_ids is not None
else self.filter_queryset(self.get_queryset()).all() else self.filter_queryset(self.get_queryset()).all()
) )
if documents_without_permitted_root( if docs.exclude(
docs, id__in=permitted_document_ids(
request.user, request.user,
perm="delete_document", perm="delete_document",
include_deleted=True, include_deleted=True,
),
).exists(): ).exists():
return HttpResponseForbidden("Insufficient permissions") return HttpResponseForbidden("Insufficient permissions")
action = serializer.validated_data.get("action") action = serializer.validated_data.get("action")
+6 -3
View File
@@ -7,6 +7,7 @@ from documents.models import Document
from documents.permissions import permitted_object_ids from documents.permissions import permitted_object_ids
from documents.permissions import restrict_queryset_to_visible from documents.permissions import restrict_queryset_to_visible
from documents.permissions import user_is_unrestricted from documents.permissions import user_is_unrestricted
from documents.versioning import annotate_effective_content
from paperless.config import AIConfig from paperless.config import AIConfig
from paperless_ai.base_model import ClassificationSuggestions from paperless_ai.base_model import ClassificationSuggestions
from paperless_ai.base_model import TaxonomyChoiceDict from paperless_ai.base_model import TaxonomyChoiceDict
@@ -113,7 +114,7 @@ def build_prompt_without_rag(
) -> str: ) -> str:
filename = document.filename or "" filename = document.filename or ""
content = truncate_content( content = truncate_content(
document.content[:4000] or "", (document.get_effective_content() or "")[:4000],
chunk_size=config.llm_embedding_chunk_size, chunk_size=config.llm_embedding_chunk_size,
context_size=config.llm_context_size, context_size=config.llm_context_size,
) )
@@ -227,7 +228,9 @@ def get_taxonomy_context(
# similar_documents is already ordered by descending weight; don't lose it. # similar_documents is already ordered by descending weight; don't lose it.
similar_document_ids = [s["document_id"] for s in similar_documents] similar_document_ids = [s["document_id"] for s in similar_documents]
similar_documents_by_id = Document.objects.in_bulk(similar_document_ids) similar_documents_by_id = annotate_effective_content(
Document.objects.all(),
).in_bulk(similar_document_ids)
similar_docs = [ similar_docs = [
similar_documents_by_id[document_id] similar_documents_by_id[document_id]
for document_id in similar_document_ids for document_id in similar_document_ids
@@ -235,7 +238,7 @@ def get_taxonomy_context(
][:max_docs] ][:max_docs]
context_blocks = [] context_blocks = []
for similar in similar_docs: for similar in similar_docs:
text = similar.content[:1000] or "" text = (similar.get_effective_content() or "")[:1000]
title = similar.title or similar.filename or "Untitled" title = similar.title or similar.filename or "Untitled"
context_blocks.append(f"TITLE: {title}\n{text}") context_blocks.append(f"TITLE: {title}\n{text}")
except Exception: except Exception:
+1 -1
View File
@@ -135,6 +135,6 @@ def build_llm_index_text(doc: Document) -> str:
lines.append(f"Custom Field - {instance.field.name}: {instance}") lines.append(f"Custom Field - {instance.field.name}: {instance}")
lines.append("\nContent:\n") lines.append("\nContent:\n")
lines.append(doc.content or "") lines.append(doc.get_effective_content() or "")
return _normalize_llm_index_text("\n".join(lines)) return _normalize_llm_index_text("\n".join(lines))
+15 -8
View File
@@ -17,6 +17,8 @@ from documents.models import PaperlessTask
from documents.utils import IterWrapper from documents.utils import IterWrapper
from documents.utils import QuerySetStream from documents.utils import QuerySetStream
from documents.utils import identity from documents.utils import identity
from documents.versioning import annotate_effective_content
from documents.versioning import root_document_ids
from paperless.config import AIConfig from paperless.config import AIConfig
from paperless_ai.db import db_connection_released from paperless_ai.db import db_connection_released
from paperless_ai.embedding import build_llm_index_text from paperless_ai.embedding import build_llm_index_text
@@ -443,11 +445,11 @@ def update_llm_index(
"Skipping LLM index update: migration check deferred; " "Skipping LLM index update: migration check deferred; "
"will retry next run." "will retry next run."
) )
documents = Document.objects.select_related( documents = annotate_effective_content(
"correspondent", Document.objects.filter(root_document__isnull=True)
"document_type", .select_related("correspondent", "document_type", "storage_path")
"storage_path", .prefetch_related("tags", "notes", "custom_fields__field"),
).prefetch_related("tags", "notes", "custom_fields__field") )
no_documents = not documents.exists() no_documents = not documents.exists()
# Fast exit before touching config: nothing to index and no existing index. # Fast exit before touching config: nothing to index and no existing index.
@@ -483,7 +485,7 @@ def update_llm_index(
msg = "LLM index rebuilt successfully." msg = "LLM index rebuilt successfully."
else: else:
scoped_documents = ( scoped_documents = (
documents.filter(id__in=document_ids) documents.filter(id__in=root_document_ids(document_ids))
if document_ids is not None if document_ids is not None
else documents else documents
) )
@@ -510,7 +512,12 @@ def update_llm_index(
def llm_index_add_or_update_document(document: Document): def llm_index_add_or_update_document(document: Document):
"""Add or atomically replace a document's chunks in the index.""" """
Add or atomically replace a document's chunks in the index. Only root
documents are indexed, so a version is indexed as its root document.
"""
if document.root_document_id is not None:
document = document.root_document
config = AIConfig() config = AIConfig()
new_nodes = build_document_node( new_nodes = build_document_node(
document, document,
@@ -688,7 +695,7 @@ def retrieve_similar_nodes(
) )
query_text = truncate_embedding_query( query_text = truncate_embedding_query(
(document.title or "") + "\n" + (document.content or ""), (document.title or "") + "\n" + (document.get_effective_content() or ""),
chunk_size=config.llm_embedding_chunk_size, chunk_size=config.llm_embedding_chunk_size,
) )
# Hold the shared read lock for the whole retrieval so the connection is # Hold the shared read lock for the whole retrieval so the connection is
@@ -55,6 +55,7 @@ def mock_document():
doc.storage_path = None doc.storage_path = None
doc.archive_serial_number = "12345" doc.archive_serial_number = "12345"
doc.content = "This is the document content." doc.content = "This is the document content."
doc.get_effective_content.return_value = "This is the document content."
cf1 = MagicMock(__str__=lambda x: "Value1") cf1 = MagicMock(__str__=lambda x: "Value1")
cf1.field = MagicMock() cf1.field = MagicMock()
@@ -434,6 +435,55 @@ def test_get_taxonomy_context_preserves_similarity_order_and_distinct_documents(
) )
@pytest.mark.django_db
class TestClassifierEffectiveContent:
"""A root document's text for the LLM is its newest version's content."""
@staticmethod
def _root_with_version() -> Document:
root = DocumentFactory(title="Statement", content="stale text")
DocumentFactory(root_document=root, version_index=1, content="latest text")
return root
def test_prompt_uses_the_newest_versions_content(self) -> None:
"""
GIVEN:
- A root document with a version
WHEN:
- The classification prompt is built for the root
THEN:
- It contains the newest version's content
"""
prompt = build_prompt_without_rag(self._root_with_version(), AIConfig())
assert "latest text" in prompt
assert "stale text" not in prompt
@override_settings(LLM_EMBEDDING_BACKEND="huggingface")
def test_similar_document_context_uses_the_newest_versions_content(self) -> None:
"""
GIVEN:
- A similar root document with a version
WHEN:
- The similar-document context is built
THEN:
- It contains the newest version's content
"""
similar = self._root_with_version()
document = DocumentFactory(content="Some content")
fake_nodes = [
SimpleNamespace(metadata={"document_id": str(similar.pk)}, score=0.9),
]
with patch(
"paperless_ai.ai_classifier.retrieve_similar_nodes",
return_value=fake_nodes,
):
_candidates, context = get_taxonomy_context(document, user=None)
assert context == "TITLE: Statement\nlatest text"
@pytest.mark.django_db @pytest.mark.django_db
@override_settings(LLM_EMBEDDING_BACKEND="huggingface") @override_settings(LLM_EMBEDDING_BACKEND="huggingface")
def test_get_taxonomy_context_no_similar_docs(): def test_get_taxonomy_context_no_similar_docs():
+103
View File
@@ -388,6 +388,108 @@ def test_update_llm_index_partial_update(
assert after[str(doc2.pk)] == before[str(doc2.pk)] assert after[str(doc2.pk)] == before[str(doc2.pk)]
@pytest.mark.django_db
class TestLlmIndexVersions:
"""The LLM index holds root documents only: a version is indexed as its root."""
def test_add_or_update_document_indexes_a_version_as_its_root(
self,
temp_llm_index_dir: Path,
mock_embed_model: FakeEmbedding,
) -> None:
"""
GIVEN:
- A root document with a version
WHEN:
- The version is passed to llm_index_add_or_update_document
THEN:
- Only the root document is in the index
"""
root = DocumentFactory(content="root content")
version = DocumentFactory(root_document=root, version_index=1)
indexing.llm_index_add_or_update_document(version)
with indexing.get_vector_store() as store:
indexed = store.get_modified_times()
assert set(indexed) == {str(root.pk)}
def test_rebuild_skips_versions(
self,
temp_llm_index_dir: Path,
mock_embed_model: FakeEmbedding,
) -> None:
"""
GIVEN:
- A root document with a version
WHEN:
- The LLM index is rebuilt
THEN:
- Only the root document is in the index
"""
root = DocumentFactory()
DocumentFactory(root_document=root, version_index=1)
indexing.update_llm_index(rebuild=True)
with indexing.get_vector_store() as store:
indexed = store.get_modified_times()
assert set(indexed) == {str(root.pk)}
def test_rebuild_indexes_the_newest_versions_content(
self,
temp_llm_index_dir: Path,
mock_embed_model: FakeEmbedding,
mocker: pytest_mock.MockerFixture,
) -> None:
"""
GIVEN:
- A root document with a version
WHEN:
- The LLM index is rebuilt
THEN:
- The root's text for the index is the version's content, answered
from the query rather than a query per document
"""
root = DocumentFactory(content="stale text")
DocumentFactory(root_document=root, version_index=1, content="latest text")
spy = mocker.spy(indexing, "build_document_node")
indexing.update_llm_index(rebuild=True)
indexed = spy.call_args.args[0]
assert indexed.pk == root.pk
assert indexed.effective_content == "latest text"
def test_incremental_update_by_version_id_refreshes_the_root(
self,
temp_llm_index_dir: Path,
mock_embed_model: FakeEmbedding,
) -> None:
"""
GIVEN:
- An indexed root document with a version whose root was modified since
WHEN:
- An incremental update is scoped to the version's id
THEN:
- The root's entry is refreshed and no entry exists for the version
"""
root = DocumentFactory()
version = DocumentFactory(root_document=root, version_index=1)
indexing.update_llm_index(rebuild=True)
Document.objects.filter(pk=root.pk).update(modified=timezone.now())
root.refresh_from_db()
indexing.update_llm_index(document_ids=[version.pk])
with indexing.get_vector_store() as store:
indexed = store.get_modified_times()
assert indexed == {str(root.pk): root.modified.isoformat()}
@pytest.mark.django_db @pytest.mark.django_db
def test_add_or_update_document_updates_existing_entry( def test_add_or_update_document_updates_existing_entry(
temp_llm_index_dir: Path, temp_llm_index_dir: Path,
@@ -637,6 +739,7 @@ class TestLlmIndexAddOrUpdateDocumentEmptyContent:
doc = MagicMock(spec=Document) doc = MagicMock(spec=Document)
doc.id = 42 doc.id = 42
doc.root_document_id = None
# Must not raise # Must not raise
indexing.llm_index_add_or_update_document(doc) indexing.llm_index_add_or_update_document(doc)
+40 -1
View File
@@ -12,6 +12,7 @@ from paperless_ai.embedding import _normalize_llm_index_text
from paperless_ai.embedding import build_llm_index_text from paperless_ai.embedding import build_llm_index_text
from paperless_ai.embedding import get_configured_model_name from paperless_ai.embedding import get_configured_model_name
from paperless_ai.embedding import get_embedding_model from paperless_ai.embedding import get_embedding_model
from paperless_testing.factories import DocumentFactory
@pytest.fixture @pytest.fixture
@@ -46,6 +47,7 @@ def mock_document():
doc.correspondent.name = "Test Correspondent" doc.correspondent.name = "Test Correspondent"
doc.archive_serial_number = "12345" doc.archive_serial_number = "12345"
doc.content = "This is the document content." doc.content = "This is the document content."
doc.get_effective_content.return_value = "This is the document content."
cf1 = MagicMock(__str__=lambda x: "Value1") cf1 = MagicMock(__str__=lambda x: "Value1")
cf1.field = MagicMock() cf1.field = MagicMock()
@@ -280,7 +282,7 @@ def test_build_llm_index_text(mock_document):
def test_build_llm_index_text_normalizes_ocr_punctuation_runs(mock_document): def test_build_llm_index_text_normalizes_ocr_punctuation_runs(mock_document):
mock_document.content = ( mock_document.get_effective_content.return_value = (
"Introduction ................................................ 7\n" "Introduction ................................................ 7\n"
"Hardware Limitation ________________________________________ 9\n" "Hardware Limitation ________________________________________ 9\n"
"Keep short punctuation like INV-100 and ellipses..." "Keep short punctuation like INV-100 and ellipses..."
@@ -294,6 +296,43 @@ def test_build_llm_index_text_normalizes_ocr_punctuation_runs(mock_document):
assert "ellipses..." in result assert "ellipses..." in result
@pytest.mark.django_db
class TestBuildLlmIndexTextVersions:
"""A root document is indexed with its effective content, like in the search index."""
def test_root_uses_the_newest_versions_content(self) -> None:
"""
GIVEN:
- A root document with two versions
WHEN:
- The LLM index text is built for the root
THEN:
- It contains the newest version's content and not the others'
"""
root = DocumentFactory(content="stale text")
DocumentFactory(root_document=root, version_index=1, content="older text")
DocumentFactory(root_document=root, version_index=2, content="latest text")
text = build_llm_index_text(root)
assert "latest text" in text
assert "stale text" not in text
assert "older text" not in text
def test_root_without_versions_uses_its_own_content(self) -> None:
"""
GIVEN:
- A root document without versions
WHEN:
- The LLM index text is built for it
THEN:
- It contains the document's own content
"""
root = DocumentFactory(content="own text")
assert "own text" in build_llm_index_text(root)
def test_normalize_llm_index_text_collapses_ocr_leaders_without_joining_lines(): def test_normalize_llm_index_text_collapses_ocr_leaders_without_joining_lines():
assert _normalize_llm_index_text("A........B\nC____D----E") == "A B\nC D E" assert _normalize_llm_index_text("A........B\nC____D----E") == "A B\nC D E"