mirror of
https://github.com/bytedance/deer-flow.git
synced 2026-04-25 11:18:22 +00:00
4ba3167f48
* feat: flush memory before summarization * fix: keep agent-scoped memory on summarization flush * fix: harden summarization hook plumbing * fix: address summarization review feedback * style: format memory middleware
99 lines
3.6 KiB
Python
99 lines
3.6 KiB
Python
"""Middleware for memory mechanism."""
|
|
|
|
import logging
|
|
from typing import override
|
|
|
|
from langchain.agents import AgentState
|
|
from langchain.agents.middleware import AgentMiddleware
|
|
from langgraph.config import get_config
|
|
from langgraph.runtime import Runtime
|
|
|
|
from deerflow.agents.memory.message_processing import detect_correction, detect_reinforcement, filter_messages_for_memory
|
|
from deerflow.agents.memory.queue import get_memory_queue
|
|
from deerflow.config.memory_config import get_memory_config
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
class MemoryMiddlewareState(AgentState):
|
|
"""Compatible with the `ThreadState` schema."""
|
|
|
|
pass
|
|
|
|
|
|
class MemoryMiddleware(AgentMiddleware[MemoryMiddlewareState]):
|
|
"""Middleware that queues conversation for memory update after agent execution.
|
|
|
|
This middleware:
|
|
1. After each agent execution, queues the conversation for memory update
|
|
2. Only includes user inputs and final assistant responses (ignores tool calls)
|
|
3. The queue uses debouncing to batch multiple updates together
|
|
4. Memory is updated asynchronously via LLM summarization
|
|
"""
|
|
|
|
state_schema = MemoryMiddlewareState
|
|
|
|
def __init__(self, agent_name: str | None = None):
|
|
"""Initialize the MemoryMiddleware.
|
|
|
|
Args:
|
|
agent_name: If provided, memory is stored per-agent. If None, uses global memory.
|
|
"""
|
|
super().__init__()
|
|
self._agent_name = agent_name
|
|
|
|
@override
|
|
def after_agent(self, state: MemoryMiddlewareState, runtime: Runtime) -> dict | None:
|
|
"""Queue conversation for memory update after agent completes.
|
|
|
|
Args:
|
|
state: The current agent state.
|
|
runtime: The runtime context.
|
|
|
|
Returns:
|
|
None (no state changes needed from this middleware).
|
|
"""
|
|
config = get_memory_config()
|
|
if not config.enabled:
|
|
return None
|
|
|
|
# Get thread ID from runtime context first, then fall back to LangGraph's configurable metadata
|
|
thread_id = runtime.context.get("thread_id") if runtime.context else None
|
|
if thread_id is None:
|
|
config_data = get_config()
|
|
thread_id = config_data.get("configurable", {}).get("thread_id")
|
|
if not thread_id:
|
|
logger.debug("No thread_id in context, skipping memory update")
|
|
return None
|
|
|
|
# Get messages from state
|
|
messages = state.get("messages", [])
|
|
if not messages:
|
|
logger.debug("No messages in state, skipping memory update")
|
|
return None
|
|
|
|
# Filter to only keep user inputs and final assistant responses
|
|
filtered_messages = filter_messages_for_memory(messages)
|
|
|
|
# Only queue if there's meaningful conversation
|
|
# At minimum need one user message and one assistant response
|
|
user_messages = [m for m in filtered_messages if getattr(m, "type", None) == "human"]
|
|
assistant_messages = [m for m in filtered_messages if getattr(m, "type", None) == "ai"]
|
|
|
|
if not user_messages or not assistant_messages:
|
|
return None
|
|
|
|
# Queue the filtered conversation for memory update
|
|
correction_detected = detect_correction(filtered_messages)
|
|
reinforcement_detected = not correction_detected and detect_reinforcement(filtered_messages)
|
|
queue = get_memory_queue()
|
|
queue.add(
|
|
thread_id=thread_id,
|
|
messages=filtered_messages,
|
|
agent_name=self._agent_name,
|
|
correction_detected=correction_detected,
|
|
reinforcement_detected=reinforcement_detected,
|
|
)
|
|
|
|
return None
|