Perf: select_related/prefetch_related for the scoped LLM index batch

src-ui/src/app/components/document-list/filter-editor/filter-editor.component.html

@@ -25,7 +25,7 @@
         <input #textFilterInput class="form-control form-control-sm" type="text"
         [disabled]="textFilterModifierIsNull"
         [(ngModel)]="textFilter"
-        (keydown)="textFilterKeydown($event)"
+        (keyup)="textFilterKeyup($event)"
         [ngbTypeahead]="searchAutoComplete"
         (selectItem)="itemSelected($event)"
         [readonly]="textFilterTarget === 'fulltext-morelike'">

src-ui/src/app/components/document-list/filter-editor/filter-editor.component.spec.ts

@@ -2180,17 +2180,17 @@ describe('FilterEditorComponent', () => {
   it('should support Enter / Esc key on text field', () => {
     component.textFilterInput.nativeElement.value = 'foo'
     component.textFilterInput.nativeElement.dispatchEvent(
-      new KeyboardEvent('keydown', { key: 'Enter' })
+      new KeyboardEvent('keyup', { key: 'Enter' })
     )
     expect(component.textFilter).toEqual('foo')
     component.textFilterInput.nativeElement.value = 'foo bar'
     component.textFilterInput.nativeElement.dispatchEvent(
-      new KeyboardEvent('keydown', { key: 'Escape' })
+      new KeyboardEvent('keyup', { key: 'Escape' })
     )
     expect(component.textFilter).toEqual('')
     const blurSpy = jest.spyOn(component.textFilterInput.nativeElement, 'blur')
     component.textFilterInput.nativeElement.dispatchEvent(
-      new KeyboardEvent('keydown', { key: 'Escape' })
+      new KeyboardEvent('keyup', { key: 'Escape' })
     )
     expect(blurSpy).toHaveBeenCalled()
   })

src-ui/src/app/components/document-list/filter-editor/filter-editor.component.ts

@@ -1312,7 +1312,7 @@ export class FilterEditorComponent
     }
   }
 
-  textFilterKeydown(event: KeyboardEvent) {
+  textFilterKeyup(event: KeyboardEvent) {
     if (event.key == 'Enter') {
       const filterString = (
         this.textFilterInput.nativeElement as HTMLInputElement

src/documents/tasks.py

@@ -331,6 +331,7 @@ def bulk_update_documents(document_ids) -> None:
     if ai_config.llm_index_enabled:
         update_llm_index(
             rebuild=False,
+            document_ids=document_ids,
         )
 
 

src/documents/tests/test_tasks.py

@@ -401,6 +401,10 @@ class TestAIIndex(DirectoriesMixin, TestCase):
                 "documents.tasks.update_llm_index",
             ) as update_llm_index,
         ):
-            tasks.bulk_update_documents([doc.pk for doc in docs])
+            doc_ids = [doc.pk for doc in docs]
+            tasks.bulk_update_documents(doc_ids)
             self.assertEqual(update_document_in_llm_index.apply_async.call_count, 0)
-            update_llm_index.assert_called_once()
+            update_llm_index.assert_called_once_with(
+                rebuild=False,
+                document_ids=doc_ids,
+            )

src/paperless_ai/base_model.py

@@ -2,8 +2,6 @@ from pydantic import BaseModel
 
 
 class DocumentClassifierSchema(BaseModel):
-    """Schema for document classification suggestions."""
-
     title: str
     tags: list[str]
     correspondents: list[str]

src/paperless_ai/indexing.py

@@ -5,7 +5,6 @@ from datetime import timedelta
 from typing import TYPE_CHECKING
 
 from django.conf import settings
-from django.core.exceptions import ObjectDoesNotExist
 from django.utils import timezone
 from filelock import FileLock
 from filelock import ReadWriteLock
@@ -168,19 +167,6 @@ def write_store(embed_model_name: str | None = None):
         yield store
 
 
-def _safe_related_name(document: Document, field: str) -> str | None:
-    """
-    Returns the ``name`` of a related object (correspondent, document_type,
-    storage_path), or None if the FK is unset or points at a row that has
-    since been deleted (e.g. concurrently with this call).
-    """
-    try:
-        related = getattr(document, field)
-    except ObjectDoesNotExist:
-        return None
-    return related.name if related else None
-
-
 def build_document_node(
     document: Document,
     *,
@@ -194,10 +180,14 @@ def build_document_node(
         "document_id": str(document.id),
         "title": document.title,
         "tags": [t.name for t in document.tags.all()],
-        "correspondent": _safe_related_name(document, "correspondent"),
-        "document_type": _safe_related_name(document, "document_type"),
+        "correspondent": document.correspondent.name
+        if document.correspondent
+        else None,
+        "document_type": document.document_type.name
+        if document.document_type
+        else None,
         "filename": document.filename,
-        "storage_path": _safe_related_name(document, "storage_path"),
+        "storage_path": document.storage_path.name if document.storage_path else None,
         "archive_serial_number": document.archive_serial_number,
         "created": document.created.isoformat() if document.created else None,
         "added": document.added.isoformat() if document.added else None,
@@ -328,8 +318,16 @@ def update_llm_index(
     *,
     iter_wrapper: IterWrapper[Document] = identity,
     rebuild=False,
+    document_ids: Iterable[int] | None = None,
 ) -> str:
-    """Rebuild or incrementally update the LLM index."""
+    """Rebuild or incrementally update the LLM index.
+
+    ``document_ids``, when given, scopes an incremental update to just those
+    documents instead of scanning the whole library -- callers that already
+    know which documents changed (e.g. a bulk edit) should pass this to avoid
+    an O(library size) scan per call. Ignored whenever a rebuild actually
+    happens, since a rebuild always covers the whole library regardless.
+    """
     with write_store() as store:
         try:
             with _exclude_readers():
@@ -379,9 +377,16 @@ def update_llm_index(
                 store.add(nodes)
             msg = "LLM index rebuilt successfully."
         else:
+            scoped_documents = (
+                Document.objects.filter(id__in=document_ids)
+                .select_related("correspondent", "document_type", "storage_path")
+                .prefetch_related("tags")
+                if document_ids is not None
+                else documents
+            )
             existing = store.get_modified_times()
             changed = 0
-            for document in iter_wrapper(documents):
+            for document in iter_wrapper(scoped_documents):
                 doc_id = str(document.id)
                 if existing.get(doc_id) == document.modified.isoformat():
                     continue

src/paperless_ai/tests/test_ai_indexing.py

@@ -8,9 +8,7 @@ from django.test import override_settings
 from django.utils import timezone
 from llama_index.core.schema import MetadataMode
 
-from documents.models import Correspondent
 from documents.models import Document
-from documents.models import DocumentType
 from documents.models import PaperlessTask
 from documents.signals import document_consumption_finished
 from documents.signals import document_updated
@@ -97,38 +95,6 @@ def test_build_document_node_structured_fields_in_metadata(
         assert "modified" in node.metadata
 
 
-@pytest.mark.django_db
-def test_build_document_node_survives_concurrently_deleted_correspondent(
-    real_document: Document,
-) -> None:
-    """Regression test for #13314.
-
-    If a document's correspondent (or document type) is deleted after the
-    in-memory Document instance was loaded but before build_document_node
-    resolves the relation, accessing the FK must not raise -- it should
-    behave like an unset FK and produce None in the metadata instead of
-    aborting the whole indexing pass.
-    """
-    correspondent = Correspondent.objects.create(name="Stale Correspondent")
-    document_type = DocumentType.objects.create(name="Stale Type")
-    real_document.correspondent = correspondent
-    real_document.document_type = document_type
-    real_document.save()
-
-    # Re-fetch to get an instance whose correspondent/document_type relations
-    # are unresolved (not yet cached), mirroring a task that loaded the
-    # document before the concurrent deletion below.
-    stale_document = Document.objects.get(pk=real_document.pk)
-
-    correspondent.delete()
-    document_type.delete()
-
-    nodes = indexing.build_document_node(stale_document)
-    assert len(nodes) > 0
-    assert nodes[0].metadata["correspondent"] is None
-    assert nodes[0].metadata["document_type"] is None
-
-
 @pytest.mark.django_db
 def test_build_document_node_excludes_document_id_from_llm_context(
     real_document: Document,
@@ -286,6 +252,22 @@ def test_update_llm_index_partial_update(
         assert store.table_exists(), (
             "Expected the vector store table to exist after incremental update"
         )
+        before = store.get_modified_times()
+
+    # A further edit, scoped via document_ids to just doc3 -- doc2 must be
+    # left exactly as it was, proving document_ids restricts the scan
+    # instead of falling back to the whole library.
+    doc3.modified = timezone.now()
+    doc3.save()
+
+    result = indexing.update_llm_index(rebuild=False, document_ids=[doc3.pk])
+    assert result == "LLM index updated successfully."
+
+    with indexing.get_vector_store() as store:
+        after = store.get_modified_times()
+
+    assert after[str(doc3.pk)] == doc3.modified.isoformat()
+    assert after[str(doc2.pk)] == before[str(doc2.pk)]
 
 
 @pytest.mark.django_db