fix(core): respect embedding opt-outs during reindex

Signed-off-by: phernandez <paul@basicmachines.co>

Author: phernandez

src/basic_memory/services/search_service.py

@@ -1,5 +1,6 @@
 """Service for search operations."""
 
+import asyncio
 import ast
 import re
 from datetime import datetime
@@ -462,8 +463,10 @@ async def sync_entity_vectors_batch(
                 and not self._entity_embeddings_enabled(entity)
             )
         ]
-        for entity_id in opted_out_ids:
-            await self._clear_entity_vectors(entity_id)
+        if opted_out_ids:
+            await asyncio.gather(
+                *(self._clear_entity_vectors(entity_id) for entity_id in opted_out_ids)
+            )
 
         eligible_entity_ids = [
             entity_id
@@ -502,7 +505,7 @@ async def reindex_vectors(self, progress_callback=None) -> dict:
         # that reference entity_ids no longer in the entity table
         await self._purge_stale_search_rows()
 
-        batch_result = await self.repository.sync_entity_vectors_batch(
+        batch_result = await self.sync_entity_vectors_batch(
             entity_ids,
             progress_callback=progress_callback,
         )

src/basic_memory/sync/sync_service.py

@@ -626,8 +626,6 @@ async def load(path: str) -> None:
                         created_at=metadata.created_at,
                         content=content,
                     )
-                except FileNotFoundError:
-                    await self.handle_delete(path)
                 except FileOperationError as exc:
                     # Trigger: FileService wraps binary read failures in FileOperationError.
                     # Why: the service contract should stay consistent for direct callers.

tests/services/test_semantic_search.py

@@ -217,3 +217,40 @@ async def test_embed_opt_out_note_still_participates_in_fts(
     )
 
     assert any(result.entity_id == entity.id for result in results)
+
+
+@pytest.mark.asyncio
+async def test_reindex_vectors_respects_embed_opt_out(search_service, monkeypatch):
+    """Full vector reindex should route through the service-level opt-out filter."""
+    monkeypatch.setattr(
+        search_service.entity_repository,
+        "find_all",
+        AsyncMock(
+            return_value=[
+                SimpleNamespace(id=41, entity_metadata={"embed": False}),
+                SimpleNamespace(id=42, entity_metadata={}),
+            ]
+        ),
+    )
+    purge_stale_rows = AsyncMock()
+    sync_batch = AsyncMock(
+        return_value=VectorSyncBatchResult(
+            entities_total=2,
+            entities_synced=1,
+            entities_failed=0,
+            entities_skipped=1,
+        )
+    )
+    monkeypatch.setattr(search_service, "_purge_stale_search_rows", purge_stale_rows)
+    monkeypatch.setattr(search_service, "sync_entity_vectors_batch", sync_batch)
+
+    stats = await search_service.reindex_vectors()
+
+    purge_stale_rows.assert_awaited_once()
+    sync_batch.assert_awaited_once_with([41, 42], progress_callback=None)
+    assert stats == {
+        "total_entities": 2,
+        "embedded": 1,
+        "skipped": 1,
+        "errors": 0,
+    }