neuro-simulator 0.1.2__tar.gz → 0.1.3__tar.gz

{neuro_simulator-0.1.2 → neuro_simulator-0.1.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: neuro_simulator
-Version: 0.1.2
+Version: 0.1.3
 Summary: Neuro Simulator Server
 Author-email: Moha-Master <hongkongreporter@outlook.com>
 License-Expression: MIT

{neuro_simulator-0.1.2 → neuro_simulator-0.1.3}/neuro_simulator/agent/core.py

@@ -149,12 +149,34 @@ Always think about whether you need to use tools before responding.
 
 IMPORTANT GUIDELINES:
 - Be creative and engaging in your responses
-- Avoid repeating the same phrases or ideas from your last response: "{last_response}" (if available)
 - Keep responses concise and conversational
 - Maintain your character's personality
+- Pay close attention to the conversation history in the context to understand the flow of the dialogue
+- Respond to the most recent user messages while considering the overall context of the conversation
 
-User messages:
+=== YOUR SPEAK HISTORY ===
 """
+
+        # Add assistant's recent responses to the prompt
+        # Get the recent context and extract speak results from llm_interaction entries
+        recent_context = await self.memory_manager.get_recent_context(100)  # Get more entries to filter
+        assistant_responses = []
+        
+        # Filter for llm_interaction entries with role: "assistant" and extract speak results
+        for entry in reversed(recent_context):  # Reverse to get newest first
+            if entry.get("type") == "llm_interaction" and entry.get("role") == "assistant":
+                tool_executions = entry.get("tool_executions", [])
+                for tool_execution in tool_executions:
+                    if tool_execution.get("name") == "speak":
+                        result = tool_execution.get("result")
+                        if result:
+                            assistant_responses.append(result)
+                    
+        # Add up to 64 most recent assistant responses
+        for i, response in enumerate(assistant_responses[:64]):
+            prompt += f"{i+1}. {response}\n"
+            
+        prompt += f"\nUser messages to respond to:\n"
         
         for msg in messages:
             prompt += f"{msg['username']}: {msg['text']}\n"
@@ -273,9 +295,10 @@ User messages:
             else:
                 agent_logger.warning(f"Failed to parse tool call from incomplete JSON block: {json_buffer}")
         
-        # If we have a final response, add it to context
-        if processing_result["final_response"]:
-            await self.memory_manager.add_context_entry("assistant", processing_result["final_response"])
+        # If we have a final response, we don't need to add it to context here
+        # because add_detailed_context_entry will add a short entry for us
+        # if processing_result["final_response"]:
+        #     await self.memory_manager.add_context_entry("assistant", processing_result["final_response"])
             
         # Update the detailed context entry with final LLM interaction details
         await self.memory_manager.add_detailed_context_entry(

{neuro_simulator-0.1.2 → neuro_simulator-0.1.3}/neuro_simulator/agent/memory/manager.py

@@ -143,10 +143,6 @@ class MemoryManager:
         }
         self.context_history.append(entry)
         
-        # Keep only last 20 context entries (10 rounds)
-        if len(self.context_history) > 20:
-            self.context_history = self.context_history[-20:]
-            
         await self._save_context()
         
     async def add_detailed_context_entry(self, input_messages: List[Dict[str, str]], 
@@ -185,18 +181,12 @@ class MemoryManager:
         }
         self.context_history.append(entry)
         
-        # Keep only last 20 context entries
-        if len(self.context_history) > 20:
-            self.context_history = self.context_history[-20:]
-            
         await self._save_context()
         return entry["id"]
         
-    async def get_recent_context(self, rounds: int = 5) -> List[Dict[str, Any]]:
-        """Get recent context (default: last 5 rounds, 10 entries)"""
-        # Each round consists of user message and assistant response
-        entries_needed = rounds * 2
-        return self.context_history[-entries_needed:] if self.context_history else []
+    async def get_recent_context(self, entries: int = 10) -> List[Dict[str, Any]]:
+        """Get recent context entries"""
+        return self.context_history[-entries:] if self.context_history else []
         
     async def get_detailed_context_history(self) -> List[Dict[str, Any]]:
         """Get the full detailed context history"""
@@ -206,9 +196,15 @@ class MemoryManager:
         """Get the last response from the agent"""
         for entry in reversed(self.context_history):
             if entry.get("role") == "assistant":
-                return entry.get("content")
+                content = entry.get("content")
+                # Skip placeholder responses
+                if content and content != "Processing started":
+                    return content
             elif entry.get("type") == "llm_interaction":
-                return entry.get("final_response")
+                final_response = entry.get("final_response")
+                # Skip placeholder responses
+                if final_response and final_response != "Processing started" and final_response != "Prompt sent to LLM" and final_response != "LLM response received":
+                    return final_response
         return None
         
     async def reset_context(self):
@@ -259,26 +255,6 @@ class MemoryManager:
                 for item in block["content"]:
                     context_parts.append(f"  - {item}")
                     
-        # Add context (recent conversation history)
-        context_parts.append("\n=== CONTEXT (Recent Conversation) ===")
-        recent_context = await self.get_recent_context(5)
-        for i, entry in enumerate(recent_context):
-            # Handle entries with and without 'role' field
-            if "role" in entry:
-                role_display = "User" if entry["role"] == "user" else "Assistant"
-                content = entry.get('content', entry.get('final_response', 'Unknown entry'))
-                context_parts.append(f"{i+1}. [{role_display}] {content}")
-            elif "type" in entry and entry["type"] == "llm_interaction":
-                # For detailed LLM interaction entries with role: assistant
-                if entry.get("role") == "assistant":
-                    context_parts.append(f"{i+1}. [Assistant] {entry.get('final_response', 'Processing step')}")
-                else:
-                    # For other llm_interaction entries without role
-                    context_parts.append(f"{i+1}. [System] {entry.get('final_response', 'Processing step')}")
-            else:
-                # Default fallback
-                context_parts.append(f"{i+1}. [System] {entry.get('content', 'Unknown entry')}")
-            
         # Add temp memory (only for temporary state, not dialog history)
         if self.temp_memory:
             context_parts.append("\n=== TEMP MEMORY (Processing State) ===")

{neuro_simulator-0.1.2 → neuro_simulator-0.1.3}/neuro_simulator.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: neuro_simulator
-Version: 0.1.2
+Version: 0.1.3
 Summary: Neuro Simulator Server
 Author-email: Moha-Master <hongkongreporter@outlook.com>
 License-Expression: MIT

{neuro_simulator-0.1.2 → neuro_simulator-0.1.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "neuro_simulator"
-version = "0.1.2"
+version = "0.1.3"
 description = "Neuro Simulator Server"
 readme = "README.md"
 authors = [{ name = "Moha-Master", email = "hongkongreporter@outlook.com" }]