perf: background post-response memory consolidation for faster replies

nanobot/agent/loop.py

@@ -100,7 +100,7 @@ class AgentLoop:
         self._mcp_connected = False
         self._mcp_connecting = False
         self._active_tasks: dict[str, list[asyncio.Task]] = {}  # session_key -> tasks
-        self._pending_archives: list[asyncio.Task] = []
+        self._background_tasks: list[asyncio.Task] = []
         self._processing_lock = asyncio.Lock()
         self.memory_consolidator = MemoryConsolidator(
             workspace=workspace,
@@ -257,8 +257,6 @@ class AgentLoop:
         """Run the agent loop, dispatching messages as tasks to stay responsive to /stop."""
         self._running = True
         await self._connect_mcp()
-        # Start background consolidation task
-        await self.memory_consolidator.start_background_task()
         logger.info("Agent loop started")
 
         while self._running:
@@ -334,9 +332,9 @@ class AgentLoop:
 
     async def close_mcp(self) -> None:
         """Drain pending background archives, then close MCP connections."""
-        if self._pending_archives:
-            await asyncio.gather(*self._pending_archives, return_exceptions=True)
-            self._pending_archives.clear()
+        if self._background_tasks:
+            await asyncio.gather(*self._background_tasks, return_exceptions=True)
+            self._background_tasks.clear()
         if self._mcp_stack:
             try:
                 await self._mcp_stack.aclose()
@@ -344,11 +342,16 @@ class AgentLoop:
                 pass  # MCP SDK cancel scope cleanup is noisy but harmless
             self._mcp_stack = None
 
-    async def stop(self) -> None:
-        """Stop the agent loop and background tasks."""
+    def _schedule_background(self, coro) -> None:
+        """Schedule a coroutine as a tracked background task (drained on shutdown)."""
+        task = asyncio.create_task(coro)
+        self._background_tasks.append(task)
+        task.add_done_callback(self._background_tasks.remove)
+
+    def stop(self) -> None:
+        """Stop the agent loop."""
         self._running = False
-        await self.memory_consolidator.stop_background_task()
-        logger.info("Agent loop stopped")
+        logger.info("Agent loop stopping")
 
     async def _process_message(
         self,
@@ -364,8 +367,7 @@ class AgentLoop:
             logger.info("Processing system message from {}", msg.sender_id)
             key = f"{channel}:{chat_id}"
             session = self.sessions.get_or_create(key)
-            self.memory_consolidator.record_activity(key)
-            await self.memory_consolidator.maybe_consolidate_by_tokens_async(session)
+            await self.memory_consolidator.maybe_consolidate_by_tokens(session)
             self._set_tool_context(channel, chat_id, msg.metadata.get("message_id"))
             history = session.get_history(max_messages=0)
             messages = self.context.build_messages(
@@ -375,6 +377,7 @@ class AgentLoop:
             final_content, _, all_msgs = await self._run_agent_loop(messages)
             self._save_turn(session, all_msgs, 1 + len(history))
             self.sessions.save(session)
+            self._schedule_background(self.memory_consolidator.maybe_consolidate_by_tokens(session))
             return OutboundMessage(channel=channel, chat_id=chat_id,
                                   content=final_content or "Background task completed.")
 
@@ -383,7 +386,6 @@ class AgentLoop:
 
         key = session_key or msg.session_key
         session = self.sessions.get_or_create(key)
-        self.memory_consolidator.record_activity(key)
 
         # Slash commands
         cmd = msg.content.strip().lower()
@@ -394,11 +396,7 @@ class AgentLoop:
             self.sessions.invalidate(session.key)
 
             if snapshot:
-                task = asyncio.create_task(
-                    self.memory_consolidator.archive_messages(snapshot)
-                )
-                self._pending_archives.append(task)
-                task.add_done_callback(self._pending_archives.remove)
+                self._schedule_background(self.memory_consolidator.archive_messages(snapshot))
 
             return OutboundMessage(channel=msg.channel, chat_id=msg.chat_id,
                                   content="New session started.")
@@ -413,8 +411,7 @@ class AgentLoop:
             return OutboundMessage(
                 channel=msg.channel, chat_id=msg.chat_id, content="\n".join(lines),
             )
-        # Record activity and schedule background consolidation for non-slash commands
-        self.memory_consolidator.record_activity(key)
+        await self.memory_consolidator.maybe_consolidate_by_tokens(session)
 
         self._set_tool_context(msg.channel, msg.chat_id, msg.metadata.get("message_id"))
         if message_tool := self.tools.get("message"):
@@ -446,6 +443,7 @@ class AgentLoop:
 
         self._save_turn(session, all_msgs, 1 + len(history))
         self.sessions.save(session)
+        self._schedule_background(self.memory_consolidator.maybe_consolidate_by_tokens(session))
 
         if (mt := self.tools.get("message")) and isinstance(mt, MessageTool) and mt._sent_in_turn:
             return None

nanobot/agent/memory.py

@@ -220,14 +220,9 @@ class MemoryStore:
 
 
 class MemoryConsolidator:
-    """Owns consolidation policy, locking, and session offset updates.
-
-    Consolidation runs asynchronously in the background when sessions are idle,
-    so it doesn't block user interactions.
-    """
+    """Owns consolidation policy, locking, and session offset updates."""
 
     _MAX_CONSOLIDATION_ROUNDS = 5
-    _IDLE_CHECK_INTERVAL = 30  # seconds between idle checks
 
     def __init__(
         self,
@@ -247,57 +242,11 @@ class MemoryConsolidator:
         self._build_messages = build_messages
         self._get_tool_definitions = get_tool_definitions
         self._locks: weakref.WeakValueDictionary[str, asyncio.Lock] = weakref.WeakValueDictionary()
-        self._background_task: asyncio.Task[None] | None = None
-        self._stop_event = asyncio.Event()
-        self._session_last_activity: dict[str, float] = {}  # session_key -> last activity timestamp
 
     def get_lock(self, session_key: str) -> asyncio.Lock:
         """Return the shared consolidation lock for one session."""
         return self._locks.setdefault(session_key, asyncio.Lock())
 
-    def record_activity(self, session_key: str) -> None:
-        """Record that a session is active (for idle detection)."""
-        self._session_last_activity[session_key] = asyncio.get_event_loop().time()
-
-    async def start_background_task(self) -> None:
-        """Start the background task that checks for idle sessions and consolidates."""
-        if self._background_task is not None and not self._background_task.done():
-            return  # Already running
-        self._stop_event.clear()
-        self._background_task = asyncio.create_task(self._idle_consolidation_loop())
-
-    async def stop_background_task(self) -> None:
-        """Stop the background task."""
-        self._stop_event.set()
-        if self._background_task is not None and not self._background_task.done():
-            self._background_task.cancel()
-            try:
-                await self._background_task
-            except asyncio.CancelledError:
-                pass
-            self._background_task = None
-
-    async def _idle_consolidation_loop(self) -> None:
-        """Background loop that checks for idle sessions and triggers consolidation."""
-        while not self._stop_event.is_set():
-            try:
-                await asyncio.sleep(self._IDLE_CHECK_INTERVAL)
-                if self._stop_event.is_set():
-                    break
-
-                # Check all sessions for idleness
-                current_time = asyncio.get_event_loop().time()
-                for session in list(self.sessions.all()):
-                    last_active = self._session_last_activity.get(session.key, 0)
-                    if current_time - last_active > self._IDLE_CHECK_INTERVAL * 2:
-                        # Session is idle, trigger consolidation
-                        await self.maybe_consolidate_by_tokens_async(session)
-
-            except asyncio.CancelledError:
-                break
-            except Exception:
-                logger.exception("Error in background consolidation loop")
-
     async def consolidate_messages(self, messages: list[dict[str, object]]) -> bool:
         """Archive a selected message chunk into persistent memory."""
         return await self.store.consolidate(messages, self.provider, self.model)
@@ -350,26 +299,8 @@ class MemoryConsolidator:
                 return True
         return True
 
-    def maybe_consolidate_by_tokens(self, session: Session) -> None:
-        """Schedule token-based consolidation to run asynchronously in background.
-
-        This method is synchronous and just schedules the consolidation task.
-        The actual consolidation runs in the background when the session is idle.
-        """
-        if not session.messages or self.context_window_tokens <= 0:
-            return
-        # Schedule for background execution
-        asyncio.create_task(self._schedule_consolidation(session))
-
-    async def _schedule_consolidation(self, session: Session) -> None:
-        """Internal method to run consolidation asynchronously."""
-        await self.maybe_consolidate_by_tokens_async(session)
-
-    async def maybe_consolidate_by_tokens_async(self, session: Session) -> None:
-        """Async version: Loop and archive old messages until prompt fits within half the context window.
-
-        This is called from the background task when a session is idle.
-        """
+    async def maybe_consolidate_by_tokens(self, session: Session) -> None:
+        """Loop: archive old messages until prompt fits within half the context window."""
         if not session.messages or self.context_window_tokens <= 0:
             return
 
@@ -424,11 +355,3 @@ class MemoryConsolidator:
                 estimated, source = self.estimate_session_prompt_tokens(session)
                 if estimated <= 0:
                     return
-
-            logger.debug(
-                "Token consolidation complete for {}: {}/{} via {}",
-                session.key,
-                estimated,
-                self.context_window_tokens,
-                source,
-            )