app/core/retriever.py aktualisiert

app/core/retriever.py

@@ -1,7 +1,7 @@
 from __future__ import annotations
 
 import time
-from typing import Any, Dict, List, Tuple, Iterable
+from typing import Any, Dict, List, Tuple
 
 from app.config import get_settings
 from app.models.dto import QueryRequest, QueryResponse, QueryHit
@@ -141,7 +141,8 @@ def _build_hits_from_semantic(
         edge_bonus = 0.0
         cent_bonus = 0.0
         if subgraph is not None:
-            node_key = payload.get("chunk_id") or payload.get("note_id")
+            # WICHTIG: Knoten im Graphen sind note_ids, nicht chunk_ids.
+            node_key = payload.get("note_id")
             if node_key:
                 try:
                     edge_bonus = float(subgraph.edge_bonus(node_key))
@@ -216,10 +217,12 @@ def hybrid_retrieve(req: QueryRequest) -> QueryResponse:
     depth, edge_types = _extract_expand_options(req)
     subgraph = None
     if depth > 0:
-        # Seeds: stabile IDs aus dem Payload (chunk_id bevorzugt, sonst note_id)
+        # Seeds: stabile IDs aus dem Payload
+        # WICHTIG: Wir verwenden note_id als Knoten-ID, da Edges zwischen Notes
+        # modelliert sind (source_id/target_id = note_id).
         seed_ids: List[str] = []
         for _, _score, payload in hits:
-            key = payload.get("chunk_id") or payload.get("note_id")
+            key = payload.get("note_id")
             if key and key not in seed_ids:
                 seed_ids.append(key)