refactor(memory): update personal memory flows and adjust related components

- Move retrieve_personal_memory and summary_personal_memory flows to new positions in default.yaml
- Update info_filter_op to handle trajectories instead of messages
- Adjust use_personal_memory_demo to use trajectories in API requests

cookbook/simple_demo/use_personal_memory_demo.py

@@ -1,5 +1,6 @@
 import asyncio
 import json
+
 import aiohttp
 
 # API base URL
@@ -45,7 +46,9 @@ async def main():
         async with session.post(
             f"{base_url}/summary_personal_memory",
             json={
-                "messages": messages,
+                "trajectories": [
+                    {"messages": messages, "score": 1.0}
+                ],
                 "workspace_id": workspace_id,
             },
             headers={"Content-Type": "application/json"}

reme_ai/config/default.yaml

@@ -34,6 +34,24 @@ flow:
         description: "A list of conversation trajectory information, including message content and score. This field does not need to be filled in, the system will complete it automatically."
         required: false
 
+  retrieve_personal_memory:
+    flow_content: set_query_op >> (extract_time_op | (retrieve_memory_op >> semantic_rank_op)) >> fuse_rerank_op
+    description: "Retrieves the most relevant personal memories from historical data based on the query to enhance response quality"
+    input_schema:
+      query:
+        type: "str"
+        description: "user query"
+        required: true
+
+  summary_personal_memory:
+    flow_content: info_filter_op >> (get_observation_op | get_observation_with_time_op | load_today_memory_op) >> contra_repeat_op >> update_vector_store_op
+    description: "Consolidates user observations and memories by filtering information and removing redundancies for efficient storage"
+    input_schema:
+      trajectories:
+        type: "list"
+        description: "A list of conversation trajectory information, including message content and score. This field does not need to be filled in, the system will complete it automatically."
+        required: false
+
   retrieve_task_memory_simple:
     flow_content: build_query_op >> recall_vector_store_op >> merge_memory_op
     description: "Retrieves the most relevant top-k memory experiences from historical data based on the current query with simplified processing"
@@ -62,24 +80,6 @@ flow:
         required: true
         enum: [ copy, delete, delete_ids, dump, load ]
 
-  retrieve_personal_memory:
-    flow_content: set_query_op >> (extract_time_op | (retrieve_memory_op >> semantic_rank_op)) >> fuse_rerank_op
-    description: "Retrieves the most relevant personal memories from historical data based on the query to enhance response quality"
-    input_schema:
-      query:
-        type: "str"
-        description: "user query"
-        required: true
-
-  summary_personal_memory:
-    flow_content: info_filter_op >> (get_observation_op | get_observation_with_time_op | load_today_memory_op) >> contra_repeat_op >> update_vector_store_op
-    description: "Consolidates user observations and memories by filtering information and removing redundancies for efficient storage"
-    input_schema:
-      messages:
-        type: "list"
-        description: "A list of conversation messages information. This field does not need to be filled in, the system will complete it automatically."
-        required: false
-
   record_task_memory:
     flow_content: update_memory_freq_op >> update_memory_utility_op >> update_vector_store_op
     description: "Update the freq & utility attributes of retrieved task memories"

reme_ai/summary/personal/info_filter_op.py

@@ -2,7 +2,7 @@ import re
 from typing import List
 
 from flowllm import C, BaseLLMOp
-from flowllm.schema.message import Message
+from flowllm.schema.message import Message, Trajectory
 from loguru import logger
 
 from reme_ai.schema.memory import PersonalMemory
@@ -19,7 +19,12 @@ class InfoFilterOp(BaseLLMOp):
     def execute(self):
         """Filter messages based on information content scores"""
         # Get messages from context - guaranteed to exist by flow input
-        self.context.messages = [Message(**x) if isinstance(x, dict) else x for x in self.context.messages]
+        trajectories: list = self.context.trajectories
+        trajectories: List[Trajectory] = [Trajectory(**x) if isinstance(x, dict) else x for x in trajectories]
+
+        self.context.messages = []
+        for trajectory in trajectories:
+            self.context.messages.extend(trajectory.messages)
         messages: List[Message] = self.context.messages
         if not messages:
             logger.warning("No messages found in context")