fixed reasoning

backend/app/memory/extract.py

@@ -5,6 +5,7 @@ from typing import Any
 
 from sqlalchemy.orm import Session
 
+from app.config import get_settings
 from app.llm.client import LLMClient
 from app.memory.service import MemoryService
 from app.projects.structuring import strip_markdown_json
@@ -61,6 +62,9 @@ async def _call_extractor(
         *[f"- {f.get('content')}" for f in facts[:30]],
     ]
 
+    settings = get_settings()
+    extract_model = settings.memory_extract_model.strip() or None
+
     llm = LLMClient()
     result = await llm.complete(
         [
@@ -72,11 +76,12 @@ async def _call_extractor(
                     + "\n\n---\nДиалог:\nПользователь: "
                     + user_text
                     + "\nАссистент: "
-                    + (assistant_text[:1500] if assistant_text else "(нет ответа)")
+                    + assistant_text[:1500]
                 ),
             },
         ],
         temperature=0.2,
+        model=extract_model,
     )
     raw = strip_markdown_json(result.get("content") or "")
     if not raw:
@@ -98,6 +103,9 @@ async def extract_after_turn(
     if not force and _should_skip_extraction(user_text):
         return {"ok": True, "skipped": "short_message", "saved": []}
 
+    if not (assistant_text or "").strip():
+        return {"ok": True, "skipped": "no_assistant_reply", "saved": []}
+
     memory = MemoryService(db)
     snapshot = memory.snapshot(session_id)