refactor(core): extract SearchReader and SemanticSearch from the search repository

CHANGELOG.md

@@ -131,6 +131,18 @@
   layer with defaults, and the filter helpers both backends share move from the base into
   `search_filters`. No query behavior changes.
 
+- **#1558**: Retrieval leaves `SearchRepositoryBase`. `SearchReader` runs one prepared
+  query over one `ProjectScope` in whichever mode it asks for, and `SemanticSearch` owns
+  vector and hybrid retrieval (adapter lookup, manifest hydration, the structured filter
+  pass, score fusion, reranking, pagination) over a `VectorRetrieval` that is present or
+  absent rather than probed with `hasattr`. The repository keeps what only it knows
+  (whether semantic search is enabled and its vector tables exist) and builds a reader
+  per call from its current state. Hydrated chunks are a typed `HydratedChunk`, which
+  retires the `best_distance` compatibility branch and the per-backend
+  `_distance_to_similarity` hooks the adapters had already replaced. Test doubles
+  construct the pipeline directly instead of subclassing the repository. No query
+  behavior changes.
+
 
 ## v0.23.2 (2026-08-25)
 

src/basic_memory/repository/postgres_search_repository.py

@@ -490,15 +490,6 @@ async def _delete_stale_chunks(
             expected_deletions=expected_deletions,
         )
 
-    @override
-    def _distance_to_similarity(self, distance: float) -> float:
-        """Convert pgvector cosine distance to cosine similarity.
-
-        pgvector's <=> operator returns cosine distance in [0, 2],
-        where cos_distance = 1 - cos_similarity.
-        """
-        return max(0.0, 1.0 - distance)
-
     @override
     def _timestamp_now_expr(self) -> str:
         return "NOW()"

src/basic_memory/repository/search_query.py

@@ -32,6 +32,28 @@ class PreparedSearchQuery:
     retrieval_mode: SearchRetrievalMode = SearchRetrievalMode.FTS
     min_similarity: float | None = None
 
+    @property
+    def has_filters(self) -> bool:
+        """Whether any predicate beyond the text itself narrows the result set.
+
+        Vector retrieval cannot evaluate these itself; when any is present it asks
+        the full-text pass which of its candidates the filters admit.
+        """
+        return any(
+            (
+                self.permalink,
+                self.permalink_match,
+                self.title,
+                self.note_types,
+                self.after_date,
+                self.search_item_types,
+                self.categories,
+                self.metadata_filters,
+                self.file_path_prefix,
+                self.temporal,
+            )
+        )
+
 
 # Interrogative/function words contribute lexical noise when a strict
 # full-text query is relaxed: "when OR did OR a" matches loud wrong documents