update

metagpt/memory/role_zero_memory.py

@@ -57,11 +57,11 @@ class RoleZeroLongTermMemory(BaseModel):
         if not query:
             return []
 
-        nodes: list[NodeWithScore] = self.rag_engine.retrieve(query)
+        nodes = self.rag_engine.retrieve(query)
+        items = self._get_items_from_nodes(nodes)
 
         memories = []
-        for node in nodes:
-            item: LongTermMemoryItem = node.metadata["obj"]
+        for item in items:
             memories.append(item.user_message)
             memories.append(item.ai_message)
 
@@ -78,3 +78,11 @@ class RoleZeroLongTermMemory(BaseModel):
             return
 
         self.rag_engine.add_objs([item])
+
+    def _get_items_from_nodes(self, nodes: list["NodeWithScore"]) -> list[LongTermMemoryItem]:
+        """Get items from nodes and arrange them in order of their `created_at`."""
+
+        items: list[LongTermMemoryItem] = [node.metadata["obj"] for node in nodes]
+        items.sort(key=lambda item: item.created_at)
+
+        return items

metagpt/roles/di/role_zero.py

@@ -285,12 +285,12 @@ class RoleZero(Role):
         self._add_memory(AIMessage(content=self.command_rsp))
         if not ok:
             error_msg = commands
-            self._add_memory(UserMessage(content=error_msg))
+            self._add_memory(UserMessage(content=error_msg, cause_by=RunCommand))
             return error_msg
         logger.info(f"Commands: \n{commands}")
         outputs = await self._run_commands(commands)
         logger.info(f"Commands outputs: \n{outputs}")
-        self._add_memory(UserMessage(content=outputs))
+        self._add_memory(UserMessage(content=outputs, cause_by=RunCommand))
 
         return AIMessage(
             content=f"I have finished the task, please mark my task as finished. Outputs: {outputs}",
@@ -605,7 +605,8 @@ class RoleZero(Role):
         if not self._should_use_longterm_memory(k=k, k_memories=memories):
             return memories
 
-        related_memories = self.longterm_memory.fetch(memories[-1].content)
+        query = self._build_longterm_memory_query(memories)
+        related_memories = self.longterm_memory.fetch(query)
         logger.info(f"Fetched {len(related_memories)} long-term memories.")
 
         # Keep user and AI messages are paired.
@@ -666,3 +667,11 @@ class RoleZero(Role):
 
     def _is_first_message_from_ai(self, memories: list[Message]) -> bool:
         return bool(memories and memories[0].is_ai_message())
+
+    def _build_longterm_memory_query(self, memories: list[Message]) -> str:
+        """Build the content used to query related long-term memory.
+
+        Default is to get the most recent user message, or an empty string if none is found.
+        """
+
+        return next((m.content for m in reversed(memories) if m.is_real_user_message()), "")

metagpt/schema.py

@@ -18,6 +18,7 @@ from __future__ import annotations
 import asyncio
 import json
 import os.path
+import time
 import uuid
 from abc import ABC
 from asyncio import Queue, QueueEmpty, wait_for
@@ -408,12 +409,15 @@ class Message(BaseModel):
         dynamic_class = create_model(class_name, **{key: (value.__class__, ...) for key, value in kvs.items()})
         return dynamic_class.model_validate(kvs)
 
-    def is_user_message(self):
+    def is_user_message(self) -> bool:
         return self.role == "user"
 
-    def is_ai_message(self):
+    def is_ai_message(self) -> bool:
         return self.role == "assistant"
 
+    def is_real_user_message(self) -> bool:
+        return self.is_user_message() and "UserRequirement" in self.cause_by
+
 
 class UserMessage(Message):
     """便于支持OpenAI的消息
@@ -966,6 +970,7 @@ class BaseEnum(Enum):
 class LongTermMemoryItem(BaseModel):
     user_message: Message
     ai_message: Message
+    created_at: Optional[float] = Field(default_factory=time.time)
 
     def rag_key(self) -> str:
         return self.user_message.content