Fix: bulk reprocess latest version for root documents

An interrupted rebuild left an empty index stamped as current, so the next
start reported it as up to date. Mark the rebuild as in progress and only
clear the marker once it completes.

.github/workflows/ci-backend.yml

@@ -80,7 +80,7 @@ jobs:
     needs: changes
     if: needs.changes.outputs.backend_changed == 'true'
     name: "Python ${{ matrix.python-version }}"
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: read
     strategy:
@@ -114,7 +114,7 @@ jobs:
           packages: unpaper tesseract-ocr imagemagick ghostscript poppler-utils
       - name: Configure ImageMagick
         run: |
-          sudo cp docker/rootfs/etc/ImageMagick-6/paperless-policy.xml /etc/ImageMagick-6/policy.xml
+          sudo cp docker/rootfs/etc/ImageMagick-6/paperless-policy.xml /etc/ImageMagick-7/policy.xml
       - name: Install Python dependencies
         env:
           PYTHON_VERSION: ${{ steps.setup-python.outputs.python-version }}
@@ -158,7 +158,7 @@ jobs:
     needs: changes
     if: needs.changes.outputs.backend_changed == 'true'
     name: Check project typing
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: read
     env:

.github/workflows/ci-docker.yml

@@ -24,10 +24,10 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - runner: ubuntu-24.04
+          - runner: ubuntu-26.04
             arch: amd64
             platform: linux/amd64
-          - runner: ubuntu-24.04-arm
+          - runner: ubuntu-26.04-arm
             arch: arm64
             platform: linux/arm64
     runs-on: ${{ matrix.runner }}
@@ -163,7 +163,7 @@ jobs:
           archive: false
   merge-and-push:
     name: Merge and Push Manifest
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     needs: build-arch
     if: needs.build-arch.outputs.should-push == 'true'
     environment: image-publishing

.github/workflows/ci-docs.yml

@@ -65,7 +65,7 @@ jobs:
     needs: changes
     if: needs.changes.outputs.docs_changed == 'true'
     name: Build Documentation
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     steps:
       - uses: actions/configure-pages@45bfe0192ca1faeb007ade9deae92b16b8254a0d # v6.0.0
       - name: Checkout
@@ -102,7 +102,7 @@ jobs:
     name: Deploy Documentation
     needs: [changes, build]
     if: github.event_name == 'push' && github.ref == 'refs/heads/main' && needs.changes.outputs.docs_changed == 'true'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       pages: write
       id-token: write

.github/workflows/ci-frontend.yml

@@ -72,7 +72,7 @@ jobs:
     needs: changes
     if: needs.changes.outputs.frontend_changed == 'true'
     name: Install Dependencies
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: read
     steps:
@@ -104,7 +104,7 @@ jobs:
     name: Lint
     needs: [changes, install-dependencies]
     if: needs.changes.outputs.frontend_changed == 'true'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: read
     steps:
@@ -137,7 +137,7 @@ jobs:
     name: "Unit Tests (${{ matrix.shard-index }}/${{ matrix.shard-count }})"
     needs: [changes, install-dependencies]
     if: needs.changes.outputs.frontend_changed == 'true'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: read
     strategy:
@@ -188,10 +188,10 @@ jobs:
     name: E2E Tests
     needs: [changes, install-dependencies]
     if: needs.changes.outputs.frontend_changed == 'true'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: read
-    container: mcr.microsoft.com/playwright:v1.62.1-noble
+    container: mcr.microsoft.com/playwright:v1.62.1-resolute
     env:
       PLAYWRIGHT_BROWSERS_PATH: /ms-playwright
       PLAYWRIGHT_SKIP_BROWSER_DOWNLOAD: 1
@@ -246,7 +246,7 @@ jobs:
     name: Frontend Build
     needs: [changes, unit-tests, e2e-tests]
     if: needs.changes.outputs.frontend_changed == 'true'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: read
     steps:

.github/workflows/ci-release.yml

@@ -14,7 +14,7 @@ permissions: {}
 jobs:
   wait-for-docker:
     name: Wait for Docker Build
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       checks: read
       statuses: read
@@ -30,7 +30,7 @@ jobs:
   build-release:
     name: Build Release
     needs: wait-for-docker
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: read
     steps:
@@ -73,7 +73,7 @@ jobs:
         timeout-minutes: 12
         uses: $/.github/actions/apt-install
         with:
-          packages: gettext liblept5
+          packages: gettext libleptonica6
       # ---- Build Documentation ----
       - name: Build documentation
         env:
@@ -145,7 +145,7 @@ jobs:
   publish-release:
     name: Publish Release
     needs: build-release
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: write
       pull-requests: write
@@ -197,7 +197,7 @@ jobs:
     name: Append Changelog
     needs: publish-release
     if: needs.publish-release.outputs.prerelease == 'false'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: write
       pull-requests: write

.github/workflows/ci-static-analysis.yml

@@ -15,7 +15,7 @@ permissions:
 jobs:
   zizmor:
     name: Run zizmor
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       contents: read
       actions: read
@@ -29,7 +29,7 @@ jobs:
         uses: zizmorcore/zizmor-action@cc914d7f3750a2d13d75c7f184a1060aa0e9d482 # v0.6.4
   semgrep:
     name: Semgrep CE
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     container:
       image: semgrep/semgrep:1.155.0@sha256:cc869c685dcc0fe497c86258da9f205397d8108e56d21a86082ea4886e52784d
     if: github.actor != 'dependabot[bot]'

.github/workflows/cleanup-tags.yml

@@ -17,7 +17,7 @@ jobs:
   cleanup-images:
     name: Cleanup Image Tags for ${{ matrix.primary-name }}
     if: github.repository_owner == 'paperless-ngx'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     environment: registry-maintenance
     strategy:
       fail-fast: false
@@ -42,7 +42,7 @@ jobs:
   cleanup-untagged-images:
     name: Cleanup Untagged Images Tags for ${{ matrix.primary-name }}
     if: github.repository_owner == 'paperless-ngx'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     needs:
       - cleanup-images
     environment: registry-maintenance

.github/workflows/codeql-analysis.yml

@@ -21,7 +21,7 @@ on:
 jobs:
   analyze:
     name: Analyze
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       actions: read
       contents: read

.github/workflows/crowdin.yml

@@ -13,7 +13,7 @@ jobs:
   synchronize-with-crowdin:
     name: Crowdin Sync
     if: github.repository_owner == 'paperless-ngx'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     environment: translation-sync
     steps:
       - name: Checkout

.github/workflows/issue-bot.yml

@@ -7,7 +7,7 @@ jobs:
     # Note: peakoss/anti-slop does not support the `issues` event yet (all of its
     # issue inputs are still commented out upstream), so the checks that the PR Bot
     # workflow gets from the action are implemented manually here.
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-slim
     permissions:
       issues: write
     steps:

.github/workflows/pr-bot.yml

@@ -4,7 +4,7 @@ on:
     types: [opened]
 jobs:
   Anti-slop:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-slim
     permissions:
       contents: read
       issues: read
@@ -24,7 +24,7 @@ jobs:
             ASLOP-PR-VERIFY
   pr-bot:
     name: Automated PR Bot
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-slim
     # Runs after Anti-slop so the welcome comment can see whether the PR was closed
     # instead of racing it. Still runs if that job fails, so labeling is not lost.
     needs: Anti-slop

.github/workflows/project-actions.yml

@@ -12,7 +12,7 @@ permissions:
 jobs:
   pr_opened_or_reopened:
     name: pr_opened_or_reopened
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-slim
     permissions:
       # write permission is required for autolabeler
       pull-requests: write

.github/workflows/repo-maintenance.yml

@@ -9,7 +9,7 @@ jobs:
   stale:
     name: 'Stale'
     if: github.repository_owner == 'paperless-ngx'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       issues: write
       pull-requests: write
@@ -34,7 +34,7 @@ jobs:
   lock-threads:
     name: 'Lock Old Threads'
     if: github.repository_owner == 'paperless-ngx'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-26.04
     permissions:
       issues: write
       pull-requests: write
@@ -58,7 +58,7 @@ jobs:
   close-answered-discussions:
     name: 'Close Answered Discussions'
     if: github.repository_owner == 'paperless-ngx'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-slim
     permissions:
       discussions: write
     steps:
@@ -117,7 +117,7 @@ jobs:
   close-outdated-discussions:
     name: 'Close Outdated Discussions'
     if: github.repository_owner == 'paperless-ngx'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-slim
     permissions:
       discussions: write
     steps:
@@ -211,7 +211,7 @@ jobs:
   close-unsupported-feature-requests:
     name: 'Close Unsupported Feature Requests'
     if: github.repository_owner == 'paperless-ngx'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-slim
     permissions:
       discussions: write
     steps:

.github/workflows/translate-strings.yml

@@ -8,7 +8,7 @@ env:
 jobs:
   generate-translate-strings:
     name: Generate Translation Strings
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-26.04
     environment: translation-sync
     permissions:
       contents: write

docs/faq.md

@@ -76,9 +76,7 @@ is not supported by any of the available parsers.
 
 **A:** Not by default. As of v3, a file whose contents match an existing document is still
 consumed, and the duplicate is flagged in the UI — open the document and check the
-**Duplicates** tab to review documents that share the same content, or filter the document
-list by **Duplicates** to find all of them (see
-[Duplicate documents](usage.md#duplicate-documents)). If you prefer the old
+**Duplicates** tab to review documents that share the same content. If you prefer the old
 behavior of rejecting duplicates during consumption, set
 [`PAPERLESS_CONSUMER_DELETE_DUPLICATES`](configuration.md#PAPERLESS_CONSUMER_DELETE_DUPLICATES)
 to `true`.

docs/usage.md

@@ -299,18 +299,19 @@ for details.
 ### Duplicate documents
 
 By default, Paperless-ngx **does not reject duplicates**. If you consume a file whose
-contents match an existing document (same original or archive checksum), the new copy is
-still consumed and a warning is logged.
+contents exactly match an existing document (same checksum), the new copy is still
+consumed and a warning is logged. The task entry for the upload also flags that a
+duplicate was detected and links to the existing document(s).
 
-When a document has duplicates, a **Duplicates** tab appears on its detail page, listing
-the other documents you can view that share the same content (including any in the trash).
-To find all documents with duplicates, choose **Duplicates** in the document list's text
-filter dropdown, or use `has_duplicates=true` in the REST API.
+To review duplicates, open a document and switch to the **Duplicates** tab on the
+document detail page. It lists other documents that share the same content, including any
+that are in the trash (shown with a badge), and links to each so you can decide which to
+keep.
 
 If you would rather reject duplicates at consumption time (the pre-v3 behavior), set
 [`PAPERLESS_CONSUMER_DELETE_DUPLICATES`](configuration.md#PAPERLESS_CONSUMER_DELETE_DUPLICATES)
 to `true`. The duplicate file is then deleted instead of consumed, and the task fails with
-a "Document already exists" message linking to the existing document.
+a "document already exists" message.
 
 ## Document Suggestions
 

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "paperless-ngx"
-version = "3.3.0"
+version = "3.2.1"
 description = """\
   A community-supported supercharged document management system: scan, index and archive all your physical documents\
   """

src-ui/package.json

@@ -1,6 +1,6 @@
 {
   "name": "paperless-ngx-ui",
-  "version": "3.3.0",
+  "version": "3.2.1",
   "scripts": {
     "preinstall": "npx only-allow pnpm",
     "ng": "ng",

src-ui/src/environments/environment.prod.ts

@@ -8,7 +8,7 @@ export const environment = {
   apiVersion: '10', // match src/paperless/settings.py
   appTitle: DEFAULT_APP_TITLE,
   tag: 'prod',
-  version: '3.3.0',
+  version: '3.2.1',
   webSocketHost: window.location.host,
   webSocketProtocol: window.location.protocol == 'https:' ? 'wss:' : 'ws:',
   webSocketBaseUrl: base_url.pathname + 'ws/',

src/documents/bulk_edit.py

@@ -408,10 +408,16 @@ def reprocess(doc_ids: list[int], *, remote_ocr: bool = False) -> Literal["OK"]:
 
     Consumption workflows do not run here, so ``remote_ocr`` is how the user
     asks for the remote engine when it is not configured to handle everything.
+
+    A root document with versions reprocesses its latest version, which is the
+    file whose content, archive and thumbnail are shown for it.
     """
-    for document_id in doc_ids:
+    for doc in Document.objects.select_related("root_document").filter(
+        id__in=doc_ids,
+    ):
+        pair = _resolve_root_and_source_doc(doc)
         update_document_content_maybe_archive_file.apply_async(
-            kwargs={"document_id": document_id, "remote_ocr": remote_ocr},
+            kwargs={"document_id": pair.source_doc.id, "remote_ocr": remote_ocr},
             headers={"trigger_source": PaperlessTask.TriggerSource.MANUAL},
         )
 

src/documents/search/_backend.py

@@ -31,6 +31,7 @@ from documents.search._query import parse_user_query
 from documents.search._schema import _write_sentinels
 from documents.search._schema import build_schema
 from documents.search._schema import open_or_rebuild_index
+from documents.search._schema import rebuild_in_progress
 from documents.search._schema import wipe_index
 from documents.search._tokenizer import ascii_fold
 from documents.search._tokenizer import autocomplete_tokens
@@ -1110,39 +1111,44 @@ class TantivyBackend:
                 flushing a segment, deferring merge work; they do not avoid it.
         """
         wipe_index(self._path)
-        new_index = tantivy.Index(build_schema(), path=str(self._path))
-        _write_sentinels(self._path)
-        register_tokenizers(new_index, settings.SEARCH_LANGUAGE)
+        # The marker covers the window where the empty index is already stamped
+        # as current but not yet populated, so an interrupted rebuild is retried.
+        with rebuild_in_progress(self._path):
+            new_index = tantivy.Index(build_schema(), path=str(self._path))
+            _write_sentinels(self._path)
+            register_tokenizers(new_index, settings.SEARCH_LANGUAGE)
 
-        # Point instance at the new index so _build_tantivy_doc uses it
-        old_index, old_schema = self._raw_index, self._raw_schema
-        self._raw_index = new_index
-        self._raw_schema = new_index.schema
-        # Stream documents one-by-one (so the progress bar advances per
-        # document) while fetching viewer permissions one SQL query per chunk.
-        # The stream is Sized, so iter_wrapper can still discover the total.
-        documents_stream = _DocumentViewerStream(documents, chunk_size=1000)
-        try:
-            writer = new_index.writer(heap_size=writer_heap_bytes)
-            for document, (viewer_ids, viewer_group_ids) in iter_wrapper(
-                documents_stream,
-            ):
-                doc = self._build_tantivy_doc(
-                    document,
-                    viewer_ids=viewer_ids,
-                    viewer_group_ids=viewer_group_ids,
-                )
-                writer.add_document(doc)
-            writer.commit()
-            # Wait for background merge threads to finish so all segments are
-            # fully merged and persisted before the index is considered rebuilt.
-            writer.wait_merging_threads()
-            new_index.reload()
-        except BaseException:  # pragma: no cover
-            # Restore old index on failure so the backend remains usable
-            self._raw_index = old_index
-            self._raw_schema = old_schema
-            raise
+            # Point instance at the new index so _build_tantivy_doc uses it
+            old_index, old_schema = self._raw_index, self._raw_schema
+            self._raw_index = new_index
+            self._raw_schema = new_index.schema
+            # Stream documents one-by-one (so the progress bar advances per
+            # document) while fetching viewer permissions one SQL query per
+            # chunk. The stream is Sized, so iter_wrapper can still discover
+            # the total.
+            documents_stream = _DocumentViewerStream(documents, chunk_size=1000)
+            try:
+                writer = new_index.writer(heap_size=writer_heap_bytes)
+                for document, (viewer_ids, viewer_group_ids) in iter_wrapper(
+                    documents_stream,
+                ):
+                    doc = self._build_tantivy_doc(
+                        document,
+                        viewer_ids=viewer_ids,
+                        viewer_group_ids=viewer_group_ids,
+                    )
+                    writer.add_document(doc)
+                writer.commit()
+                # Wait for background merge threads to finish so all segments
+                # are fully merged and persisted before the index is considered
+                # rebuilt.
+                writer.wait_merging_threads()
+                new_index.reload()
+            except BaseException:  # pragma: no cover
+                # Restore old index on failure so the backend remains usable
+                self._raw_index = old_index
+                self._raw_schema = old_schema
+                raise
 
 
 def chunked(iterable, size):

src/documents/search/_schema.py

@@ -4,6 +4,7 @@ import hashlib
 import json
 import logging
 import shutil
+from contextlib import contextmanager
 from typing import TYPE_CHECKING
 from typing import Final
 from typing import NamedTuple
@@ -16,6 +17,7 @@ from whoosh_compat import FieldKind
 from documents.search._fields import PUBLIC_FIELDS
 
 if TYPE_CHECKING:
+    from collections.abc import Iterator
     from pathlib import Path
 
 logger = logging.getLogger("paperless.search")
@@ -28,6 +30,11 @@ logger = logging.getLogger("paperless.search")
 # v3 - barcodes JSON field for stored barcode contents
 SCHEMA_VERSION: Final[int] = 3
 
+# Present in the index directory from the moment a full rebuild starts until it
+# finishes. If a rebuild is interrupted it is left behind, so the half-built
+# index is not mistaken for a complete one.
+REBUILD_MARKER: Final[str] = ".rebuilding"
+
 
 class FieldDescriptor(NamedTuple):
     """One tantivy field, in declaration order.
@@ -255,9 +262,9 @@ def needs_rebuild(index_dir: Path) -> bool:
     """
     Check if the search index needs rebuilding.
 
-    Reads .index_settings.json to compare the stored schema version, search
-    language and schema fingerprint against the current configuration. Returns
-    True if the file is missing, unparsable, or any value mismatches.
+    True if a previous full rebuild never finished (the rebuild marker is still
+    present), or if the index's stamped settings no longer match the current
+    configuration. See _settings_mismatch().
 
     Args:
         index_dir: Path to the search index directory
@@ -265,6 +272,40 @@ def needs_rebuild(index_dir: Path) -> bool:
     Returns:
         True if the index needs rebuilding, False if it's up to date
     """
+    if (index_dir / REBUILD_MARKER).exists():
+        logger.warning("Previous search index rebuild did not finish - rebuilding.")
+        return True
+    return _settings_mismatch(index_dir)
+
+
+@contextmanager
+def rebuild_in_progress(index_dir: Path) -> Iterator[None]:
+    """
+    Flag the index as incomplete for the duration of a full rebuild.
+
+    The marker is cleared only if the block exits cleanly. There is deliberately
+    no try/finally: an exception must leave the marker behind so the next
+    needs_rebuild() check retries the rebuild.
+    """
+    marker = index_dir / REBUILD_MARKER
+    marker.touch()
+    yield
+    marker.unlink(missing_ok=True)
+
+
+def _settings_mismatch(index_dir: Path) -> bool:
+    """
+    Check the stamped settings against the current configuration.
+
+    Reads .index_settings.json to compare the stored schema version, search
+    language and schema fingerprint. Returns True if the file is missing,
+    unparsable, or any value mismatches.
+
+    This deliberately ignores the rebuild marker: open_or_rebuild_index() uses it
+    so that a process opening the index while another process is mid-rebuild
+    (or after one died) does not wipe the partial index out from under it.
+    Repopulating is the job of ``document_index reindex``.
+    """
     settings_file = index_dir / ".index_settings.json"
     if not settings_file.exists():
         return True
@@ -333,7 +374,7 @@ def open_or_rebuild_index(index_dir: Path | None = None) -> tantivy.Index:
         index_dir = cast("Path", settings.INDEX_DIR)
     if not index_dir.exists():
         return tantivy.Index(build_schema())
-    if needs_rebuild(index_dir):
+    if _settings_mismatch(index_dir):
         wipe_index(index_dir)
         idx = tantivy.Index(build_schema(), path=str(index_dir))
         _write_sentinels(index_dir)

src/documents/tasks.py

@@ -490,18 +490,23 @@ def update_document_content_maybe_archive_file(
                     shutil.move(thumbnail, document.thumbnail_path)
 
             document.refresh_from_db()
+            root_document = (
+                document.root_document if document.root_document_id else document
+            )
             logger.info(
-                f"Updating index for document {document_id} ({document.archive_checksum})",
+                f"Updating index for document {root_document.pk} ({document.archive_checksum})",
             )
             from documents.search import get_backend
 
-            get_backend().add_or_update(document)
+            get_backend().add_or_update(root_document)
 
             ai_config = AIConfig()
             if ai_config.llm_index_enabled:
-                llm_index_add_or_update_document(document)
+                llm_index_add_or_update_document(root_document)
 
             clear_document_caches(document.pk)
+            if root_document.pk != document.pk:
+                clear_document_caches(root_document.pk)
 
         except Exception:
             logger.exception(

src/documents/tests/search/test_backend.py

@@ -16,7 +16,10 @@ from documents.search._backend import TantivyBackend
 from documents.search._backend import WriteBatch
 from documents.search._backend import get_backend
 from documents.search._backend import reset_backend
+from documents.search._schema import REBUILD_MARKER
+from documents.search._schema import needs_rebuild
 from documents.signals.handlers import add_to_index
+from paperless_testing.dirs import PaperlessDirs
 from paperless_testing.factories import CorrespondentFactory
 from paperless_testing.factories import DocumentFactory
 from paperless_testing.factories import DocumentTypeFactory
@@ -823,6 +826,53 @@ class TestRebuild:
         backend.rebuild(Document.objects.all(), iter_wrapper=wrapper)
         assert 30 in seen
 
+    def test_successful_rebuild_leaves_index_up_to_date(
+        self,
+        backend: TantivyBackend,
+        paperless_dirs: PaperlessDirs,
+    ) -> None:
+        """
+        GIVEN:
+            - A backend and one document
+        WHEN:
+            - rebuild() completes
+        THEN:
+            - needs_rebuild() is False and no rebuild marker remains
+        """
+        DocumentFactory.create()
+
+        backend.rebuild(Document.objects.all())
+
+        assert needs_rebuild(paperless_dirs.index_dir) is False
+        assert not (paperless_dirs.index_dir / REBUILD_MARKER).exists()
+
+    def test_interrupted_rebuild_is_retried(
+        self,
+        backend: TantivyBackend,
+        paperless_dirs: PaperlessDirs,
+    ) -> None:
+        """
+        GIVEN:
+            - A rebuild that dies while indexing documents (e.g. the database
+              connection is lost)
+        WHEN:
+            - needs_rebuild() is checked afterwards
+        THEN:
+            - It is True, even though the empty index was already stamped with
+              current settings, so the next start rebuilds instead of reporting
+              the index as up to date
+        """
+        DocumentFactory.create()
+
+        def die(pairs):
+            raise RuntimeError("terminating connection due to administrator command")
+            yield  # pragma: no cover
+
+        with pytest.raises(RuntimeError):
+            backend.rebuild(Document.objects.all(), iter_wrapper=die)
+
+        assert needs_rebuild(paperless_dirs.index_dir) is True
+
     def test_includes_group_granted_viewers(self, backend: TantivyBackend) -> None:
         """Rebuild must index viewer ids for group-only grants, not just direct ones.
 

src/documents/tests/test_bulk_edit.py

@@ -2021,3 +2021,69 @@ class TestBulkEditReprocess(DirectoriesMixin, TestCase):
         self.assertEqual(mock_task.apply_async.call_count, 2)
         for call in mock_task.apply_async.call_args_list:
             self.assertTrue(call.kwargs["kwargs"]["remote_ocr"])
+
+    @mock.patch("documents.bulk_edit.update_document_content_maybe_archive_file")
+    def test_reprocess_root_uses_latest_version(self, mock_task: mock.Mock) -> None:
+        """
+        GIVEN:
+            - A root document with two versions
+        WHEN:
+            - reprocess is called with the root document
+        THEN:
+            - The latest version is reprocessed, not the root's original file
+        """
+        Document.objects.create(
+            title="test",
+            checksum="A-v1",
+            mime_type="application/pdf",
+            root_document=self.doc,
+            version_index=1,
+        )
+        latest = Document.objects.create(
+            title="test",
+            checksum="A-v2",
+            mime_type="application/pdf",
+            root_document=self.doc,
+            version_index=2,
+        )
+
+        bulk_edit.reprocess([self.doc.id])
+
+        mock_task.apply_async.assert_called_once()
+        self.assertEqual(
+            mock_task.apply_async.call_args.kwargs["kwargs"]["document_id"],
+            latest.id,
+        )
+
+    @mock.patch("documents.bulk_edit.update_document_content_maybe_archive_file")
+    def test_reprocess_explicit_version(self, mock_task: mock.Mock) -> None:
+        """
+        GIVEN:
+            - A root document with two versions
+        WHEN:
+            - reprocess is called with the older version
+        THEN:
+            - That version is reprocessed
+        """
+        older = Document.objects.create(
+            title="test",
+            checksum="A-v1",
+            mime_type="application/pdf",
+            root_document=self.doc,
+            version_index=1,
+        )
+        Document.objects.create(
+            title="test",
+            checksum="A-v2",
+            mime_type="application/pdf",
+            root_document=self.doc,
+            version_index=2,
+        )
+
+        bulk_edit.reprocess([older.id])
+
+        mock_task.apply_async.assert_called_once()
+        self.assertEqual(
+            mock_task.apply_async.call_args.kwargs["kwargs"]["document_id"],
+            older.id,
+        )

src/documents/tests/test_tasks.py

@@ -287,6 +287,61 @@ class TestUpdateContent(DirectoriesMixin, TestCase):
         tasks.update_document_content_maybe_archive_file(doc.pk)
         self.assertNotEqual(Document.objects.get(pk=doc.pk).content, "test")
 
+    @mock.patch("documents.tasks.clear_document_caches")
+    @mock.patch("documents.search.get_backend")
+    def test_update_content_version_indexes_root(
+        self,
+        mock_get_backend: mock.Mock,
+        mock_clear_caches: mock.Mock,
+    ) -> None:
+        """
+        GIVEN:
+            - A root document with a version
+        WHEN:
+            - Update content task is called for the version
+        THEN:
+            - The version's content is updated
+            - The root document is indexed rather than the version
+            - Caches are cleared for both
+        """
+        sample1 = self.dirs.scratch_dir / "sample.pdf"
+        shutil.copy(
+            Path(__file__).parent
+            / "samples"
+            / "documents"
+            / "originals"
+            / "0000001.pdf",
+            sample1,
+        )
+        root = Document.objects.create(
+            title="test",
+            content="root content",
+            checksum="root",
+            mime_type="application/pdf",
+        )
+        version = Document.objects.create(
+            title="test",
+            content="my document",
+            checksum="wow",
+            filename=sample1,
+            mime_type="application/pdf",
+            root_document=root,
+            version_index=1,
+        )
+
+        tasks.update_document_content_maybe_archive_file(version.pk)
+
+        self.assertNotEqual(
+            Document.objects.get(pk=version.pk).content,
+            "my document",
+        )
+        self.assertEqual(Document.objects.get(pk=root.pk).content, "root content")
+        indexed = mock_get_backend.return_value.add_or_update.call_args.args[0]
+        self.assertEqual(indexed.pk, root.pk)
+        mock_clear_caches.assert_has_calls(
+            [mock.call(version.pk), mock.call(root.pk)],
+        )
+
 
 class TestUpdateContentRemoteOCR(DirectoriesMixin, TestCase):
     """

src/locale/en_US/LC_MESSAGES/django.po

@@ -2,7 +2,7 @@ msgid ""
 msgstr ""
 "Project-Id-Version: paperless-ngx\n"
 "Report-Msgid-Bugs-To: \n"
-"POT-Creation-Date: 2026-10-05 16:26+0000\n"
+"POT-Creation-Date: 2026-10-06 15:12+0000\n"
 "PO-Revision-Date: 2022-02-17 04:17\n"
 "Last-Translator: \n"
 "Language-Team: English\n"
@@ -1941,25 +1941,8 @@ msgstr ""
 msgid "As a final step, please complete the following form:"
 msgstr ""
 
-#: documents/validators.py:24
-#, python-brace-format
-msgid "Unable to parse URI {value}, missing scheme"
-msgstr ""
-
-#: documents/validators.py:29
-#, python-brace-format
-msgid "Unable to parse URI {value}, missing net location or path"
-msgstr ""
-
 #: documents/validators.py:36
-msgid ""
-"URI scheme '{parts.scheme}' is not allowed. Allowed schemes: {', '."
-"join(allowed_schemes)}"
-msgstr ""
-
-#: documents/validators.py:45
-#, python-brace-format
-msgid "Unable to parse URI {value}"
+msgid ", "
 msgstr ""
 
 #: documents/views.py:336 documents/views.py:2729

src/paperless/tests/parsers/test_mail_parser_live.py

@@ -137,9 +137,20 @@ class TestNginxService:
     reason="No Gotenberg/Tika servers to test with",
 )
 class TestParserLive:
-    @staticmethod
-    def imagehash(file: Path, hash_size: int = 18) -> str:
-        return f"{average_hash(Image.open(file), hash_size)}"
+    # Rasterizer versions shift a few pixels, so compare perceptual hashes by
+    # Hamming distance (out of 18 * 18 = 324 bits) rather than for equality
+    MAX_HASH_DISTANCE = 8
+
+    @classmethod
+    def assert_thumbnails_similar(cls, generated: Path, expected: Path) -> None:
+        distance = average_hash(Image.open(generated), 18) - average_hash(
+            Image.open(expected),
+            18,
+        )
+        assert distance <= cls.MAX_HASH_DISTANCE, (
+            f"Thumbnail {generated} differs from {expected} by {distance} bits "
+            f"(max {cls.MAX_HASH_DISTANCE})"
+        )
 
     def test_get_thumbnail(
         self,
@@ -168,12 +179,7 @@ class TestParserLive:
         assert thumb.exists()
         assert thumb.is_file()
 
-        assert self.imagehash(thumb) == self.imagehash(
-            simple_txt_email_thumbnail_file,
-        ), (
-            f"Created thumbnail {thumb} differs from expected file "
-            f"{simple_txt_email_thumbnail_file}"
-        )
+        self.assert_thumbnails_similar(thumb, simple_txt_email_thumbnail_file)
 
     def test_tika_parse_successful(self, mail_parser: MailDocumentParser) -> None:
         """
@@ -255,7 +261,7 @@ class TestParserLive:
         THEN:
             - Gotenberg shall be called to generate the PDF
             - The archive PDF shall contain the expected content
-            - The generated thumbnail shall match the expected image hash
+            - The generated thumbnail shall be perceptually close to the expected image
         """
         util_call_with_backoff(mail_parser.parse, [html_email_file, "message/rfc822"])
 
@@ -272,14 +278,4 @@ class TestParserLive:
             html_email_file,
             "message/rfc822",
         )
-        generated_thumbnail_hash = self.imagehash(generated_thumbnail)
-
-        # The created PDF is not reproducible, but the converted image
-        # should always look the same
-        expected_hash = self.imagehash(html_email_thumbnail_file)
-
-        assert generated_thumbnail_hash == expected_hash, (
-            f"PDF thumbnail differs from expected. "
-            f"Generated: {generated_thumbnail}, "
-            f"Hash: {generated_thumbnail_hash} vs {expected_hash}"
-        )
+        self.assert_thumbnails_similar(generated_thumbnail, html_email_thumbnail_file)

src/paperless/version.py

@@ -1,6 +1,6 @@
 from typing import Final
 
-__version__: Final[tuple[int, int, int]] = (3, 3, 0)
+__version__: Final[tuple[int, int, int]] = (3, 2, 1)
 # Version string like X.Y.Z
 __full_version_str__: Final[str] = ".".join(map(str, __version__))
 # Version string like X.Y

uv.lock

@@ -2971,7 +2971,7 @@ wheels = [
 
 [[package]]
 name = "paperless-ngx"
-version = "3.3.0"
+version = "3.2.1"
 source = { virtual = "." }
 dependencies = [
     { name = "azure-ai-documentintelligence" },