diff --git a/core/agent/__pycache__/agent.cpython-310.pyc b/core/agent/__pycache__/agent.cpython-310.pyc
index e006a52..7206d64 100644
Binary files a/core/agent/__pycache__/agent.cpython-310.pyc and b/core/agent/__pycache__/agent.cpython-310.pyc differ
diff --git a/core/agent/agent.py b/core/agent/agent.py
index c519050..4367d0e 100644
--- a/core/agent/agent.py
+++ b/core/agent/agent.py
@@ -97,17 +97,15 @@ User`s name: {self.user_name}
 
         summary = self.llm.generate_stream(messages)
 
-        # сохраняем summary
         self.summary = summary
 
-        # 🔥 очищаем историю (оставляем последние 2-4 сообщения)
         try:
             self.chat_history = self.chat_history[-2:]
         except IndexError:
             pass
 
     def save_memory(self):
-        # Используем абсолютный путь от корня проекта
+        # абсолютный путь от корня проекта
         project_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
         memory_dir = os.path.join(project_root, "data", "memory")
         os.makedirs(memory_dir, exist_ok=True)
@@ -123,7 +121,7 @@ User`s name: {self.user_name}
             json.dump(data, f, ensure_ascii=False, indent=2)
 
     def load_memory(self):
-        # Используем абсолютный путь от корня проекта
+        # абсолютный путь от корня проекта
         project_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
         path = os.path.join(project_root, "data", "memory", f"{self.character.id}.json")
 
@@ -188,6 +186,7 @@ User`s name: {self.user_name}
     
     def stream_responce(self, user_input, temperature=None, max_tokens=None): #В ollamaapi нужно будет создать новый метод чисто на стриминг, пока что закомментил строки с цельным ответом
         messages = self.build_messages(user_input)
+        full_response = ""
         if temperature is not None:
             self.character.temperature = temperature
         if max_tokens is not None:
@@ -199,6 +198,12 @@ User`s name: {self.user_name}
             max_tokens=self.character.max_tokens
         )
 
+        for token in response:
+
+            full_response += token
+
+            yield token
+
         # сохраняем историю
         self.chat_history.append({
             "role": "user",
@@ -207,11 +212,13 @@ User`s name: {self.user_name}
 
         self.chat_history.append({
             "role": "assistant",
-            "content": response
+            "content": full_response
         })
 
         if len(self.chat_history) > 20:
             self.summarize_history()
+        
+        self.save_memory()
 
         return response
     
diff --git a/core/flaskui/__pycache__/flask.cpython-310.pyc b/core/flaskui/__pycache__/flask.cpython-310.pyc
index dafa56c..ae67c2f 100644
Binary files a/core/flaskui/__pycache__/flask.cpython-310.pyc and b/core/flaskui/__pycache__/flask.cpython-310.pyc differ
diff --git a/core/flaskui/flask.py b/core/flaskui/flask.py
index b9434b4..f1a6abf 100644
--- a/core/flaskui/flask.py
+++ b/core/flaskui/flask.py
@@ -49,9 +49,10 @@ def stream():
     def generate():
         for content in agent.stream_responce(last_prompt):
             yield f"data: {content}\n\n"
-    agent.save_memory()
+
     return Response(generate(), mimetype="text/event-stream")
 
+
 #список персонажей
 @ui.route("/characters", methods=["GET"])
 def get_characters():