Perf: select_related/prefetch_related for the scoped LLM index batch

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/indexing.py

@@ -318,8 +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():
@@ -369,9 +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

@@ -252,6 +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