Add system messages for memories as a user prompt as well as a system…

… message, which makes the performance much better.

src/llm_interface.py

@@ -32,6 +32,8 @@ def generate_response(self, message_history, system_message=None):
             "stream": False
         }
 
+        print(data)
+
         response = self._post_request(data)
         if response:
             return response.json().get("message", {}).get("content", "Error: No content received.")
@@ -46,6 +48,8 @@ def generate_streaming_response(self, message_history):
             "stream": True
         }
 
+        print(data)
+
         full_response = []
         response = self._post_request(data)
 

src/memory_manager.py

@@ -59,10 +59,13 @@ def get_active_memories(self):
     def remember(self, text, category="general", priority=3):
         """Summarizes the text using the LLM and stores it as a memory."""
         # Prepare the message history for summarization
-        message_history = [{"role": "user", "content": text}]
+        messages = [
+                {"role": "user", "content": self.llm.settings.get("system_message_how_to_remember_information_in_prompt")},
+                {"role": "user", "content": text}
+        ]
         # Generate the summary or relevant info from the LLM
         summary = self.llm.generate_response(
-                message_history, 
+                messages, 
                 system_message=self.llm.settings.get("system_message_how_to_remember_information_in_prompt")
         )
 
@@ -73,8 +76,12 @@ def remember(self, text, category="general", priority=3):
     def summarize_and_save(self, conversation, conversation_file):
         """Summarize the conversation and save it to memories."""
         conversation_text = "\n".join([entry["content"] for entry in conversation])
+        messages = [
+                {"role": "user", "content": self.llm.settings.get("system_message_how_to_extract_relevant_info_for_memory")},
+                {"role": "user", "content": conversation_text}
+        ]
         summary = self.llm.generate_response(
-                [{"role": "user", "content": conversation_text}],
+                messages,
                 system_message=self.llm.settings.get("system_message_how_to_extract_relevant_info_for_memory")
         )