diff --git a/backend/app/config.py b/backend/app/config.py
index 093d7e5..7f8d6a8 100644
--- a/backend/app/config.py
+++ b/backend/app/config.py
@@ -9,12 +9,24 @@ DEPRECATED_VISION_MODELS: dict[str, str] = {
     "google/gemini-2.0-flash-lite": "google/gemini-2.5-flash-lite",
 }
 
+DEPRECATED_EXTRACT_MODELS: dict[str, str] = {
+    "google/gemini-2.0-flash-001": "google/gemini-2.5-flash-lite",
+    "google/gemini-2.0-flash": "google/gemini-2.5-flash-lite",
+    "google/gemini-2.0-flash-lite-001": "google/gemini-2.5-flash-lite",
+    "google/gemini-2.0-flash-lite": "google/gemini-2.5-flash-lite",
+}
+
 
 def resolve_vision_model(model: str) -> str:
     stripped = model.strip()
     return DEPRECATED_VISION_MODELS.get(stripped, stripped)
 
 
+def resolve_extract_model(model: str) -> str:
+    stripped = model.strip()
+    return DEPRECATED_EXTRACT_MODELS.get(stripped, stripped)
+
+
 class Settings(BaseSettings):
     model_config = SettingsConfigDict(
         env_file=(".env", "../.env"),
diff --git a/backend/app/llm/client.py b/backend/app/llm/client.py
index 047d163..27d5dff 100644
--- a/backend/app/llm/client.py
+++ b/backend/app/llm/client.py
@@ -19,6 +19,9 @@ class LLMClient:
             base_url=settings.openrouter_base_url,
         )
 
+    async def aclose(self) -> None:
+        await self.client.close()
+
     def _runtime(self) -> tuple[str, str, str]:
         from app.db.base import SessionLocal
         from app.settings.service import SettingsService
diff --git a/backend/app/memory/extract.py b/backend/app/memory/extract.py
index f07a405..703d280 100644
--- a/backend/app/memory/extract.py
+++ b/backend/app/memory/extract.py
@@ -5,10 +5,11 @@ from typing import Any
 
 from sqlalchemy.orm import Session
 
-from app.config import get_settings
+from app.db.base import SessionLocal
 from app.llm.client import LLMClient
 from app.memory.service import MemoryService
 from app.projects.structuring import strip_markdown_json
+from app.settings.service import SettingsService
 
 logger = logging.getLogger(__name__)
 
@@ -62,28 +63,34 @@ 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
+    db = SessionLocal()
+    try:
+        extract_model = str(SettingsService(db).get_effective("memory_extract_model")).strip() or None
+    finally:
+        db.close()
 
     llm = LLMClient()
-    result = await llm.complete(
-        [
-            {"role": "system", "content": EXTRACTION_PROMPT},
-            {
-                "role": "user",
-                "content": (
-                    "\n".join(known)
-                    + "\n\n---\nДиалог:\nПользователь: "
-                    + user_text
-                    + "\nАссистент: "
-                    + assistant_text[:1500]
-                ),
-            },
-        ],
-        temperature=0.2,
-        model=extract_model,
-        for_extraction=True,
-    )
+    try:
+        result = await llm.complete(
+            [
+                {"role": "system", "content": EXTRACTION_PROMPT},
+                {
+                    "role": "user",
+                    "content": (
+                        "\n".join(known)
+                        + "\n\n---\nДиалог:\nПользователь: "
+                        + user_text
+                        + "\nАссистент: "
+                        + assistant_text[:1500]
+                    ),
+                },
+            ],
+            temperature=0.2,
+            model=extract_model,
+            for_extraction=True,
+        )
+    finally:
+        await llm.aclose()
     raw = strip_markdown_json(result.get("content") or "")
     if not raw:
         return {"facts": [], "profile": {}}
diff --git a/backend/app/rag/embeddings.py b/backend/app/rag/embeddings.py
index 2e057e8..f281f01 100644
--- a/backend/app/rag/embeddings.py
+++ b/backend/app/rag/embeddings.py
@@ -7,4 +7,7 @@ async def embed_texts(texts: list[str]) -> list[list[float]]:
     if not texts:
         return []
     client = LLMClient()
-    return await client.embed(texts)
+    try:
+        return await client.embed(texts)
+    finally:
+        await client.aclose()
diff --git a/backend/app/settings/service.py b/backend/app/settings/service.py
index b5ca24e..9505404 100644
--- a/backend/app/settings/service.py
+++ b/backend/app/settings/service.py
@@ -7,7 +7,7 @@ from typing import Any
 from sqlalchemy import select
 from sqlalchemy.orm import Session
 
-from app.config import Settings, get_settings, resolve_vision_model
+from app.config import Settings, get_settings, resolve_extract_model, resolve_vision_model
 from app.db.models import AssistantState
 
 SETTING_KEYS = (
@@ -69,6 +69,8 @@ class SettingsService:
                 return self._default_for(key)
         if key == "openrouter_vision_model":
             return resolve_vision_model(raw.strip())
+        if key == "memory_extract_model":
+            return resolve_extract_model(raw.strip())
         return raw
 
     def snapshot(self) -> dict[str, Any]: