feat(agent): integrate batch file writing with context offload

reme_ai/agent/react/agentic_retrieve_op.py

@@ -105,8 +105,9 @@ class AgenticRetrieveOp(ReactAgentOp):
     async def before_chat(self, messages: List[Message]):
         """Run context offload to trim prior messages before invoking the agent."""
         from reme_ai.context.offload import ContextOffloadOp
+        from reme_ai.context.file_tool import BatchWriteFileOp
 
-        op = ContextOffloadOp()
+        op = ContextOffloadOp() >> BatchWriteFileOp()
         await op.async_call(**self.input_dict)
         messages = op.context.response.answer
         messages = [Message(**x) for x in messages]

reme_ai/main.py

@@ -6,7 +6,6 @@ which extends FlowLLM with specialized memory management capabilities including:
 - Task Memory: Store and retrieve task execution histories
 - Tool Memory: Track tool usage patterns and experiences
 - Personal Memory: Manage user preferences and personal information
-- SOP Memory: Standard Operating Procedures for recurring workflows
 """
 
 import asyncio