superlocalmemory 3.4.35__tar.gz → 3.4.36__tar.gz

{superlocalmemory-3.4.35/src/superlocalmemory.egg-info → superlocalmemory-3.4.36}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: superlocalmemory
-Version: 3.4.35
+Version: 3.4.36
 Summary: Information-geometric agent memory with mathematical guarantees
 Author-email: Varun Pratap Bhardwaj <admin@superlocalmemory.com>
 License: AGPL-3.0-or-later

{superlocalmemory-3.4.35 → superlocalmemory-3.4.36}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "superlocalmemory"
-version = "3.4.35"
+version = "3.4.36"
 description = "Information-geometric agent memory with mathematical guarantees"
 readme = "README.md"
 license = {text = "AGPL-3.0-or-later"}

{superlocalmemory-3.4.35 → superlocalmemory-3.4.36}/src/superlocalmemory/__init__.py

@@ -1,3 +1,3 @@
 """SuperLocalMemory — information-geometric agent memory."""
 
-__version__ = "3.4.35"
+__version__ = "3.4.36"

{superlocalmemory-3.4.35 → superlocalmemory-3.4.36}/src/superlocalmemory/hooks/auto_recall_hook.py

@@ -76,6 +76,30 @@ def _get_queue_db_path():
     return slm_dir / "recall_queue.db"
 
 
+def _try_socket_first(prompt: str, session_id: str) -> dict | None:
+    """Try the persistent hook daemon socket. Returns full envelope or None.
+
+    The socket path returns an already-formatted Claude Code envelope.
+    If this returns a non-None dict, the caller writes it to stdout directly
+    (skip _do_recall + _format_envelope). Returns None on any failure,
+    triggering the subprocess fallback.
+    """
+    try:
+        from superlocalmemory.hooks.hook_daemon import try_socket_recall
+        response = try_socket_recall(
+            prompt=prompt,
+            session_id=session_id,
+            timeout=_get_mode_timeout(_detect_mode()),
+        )
+        if response is None or not isinstance(response, dict):
+            return None
+        if not response:
+            return {}
+        return response
+    except Exception:
+        return None
+
+
 def _do_recall(query: str, limit: int = _DEFAULT_LIMIT, session_id: str = "") -> list[dict] | None:
     """Enqueue recall to queue, poll for result. Returns list of dicts or None."""
     try:
@@ -192,6 +216,14 @@ def main() -> int:
         sys.stdout.write("{}")
         return 0
 
+    try:
+        socket_result = _try_socket_first(prompt, session_id)
+        if socket_result is not None:
+            sys.stdout.write(json.dumps(socket_result) if socket_result else "{}")
+            return 0
+    except Exception:
+        pass
+
     try:
         results = _do_recall(prompt, limit=_DEFAULT_LIMIT, session_id=session_id)
     except Exception:

superlocalmemory-3.4.36/src/superlocalmemory/hooks/hook_daemon.py

@@ -0,0 +1,276 @@
+# Copyright (c) 2026 Varun Pratap Bhardwaj / Qualixar
+# Licensed under AGPL-3.0-or-later - see LICENSE file
+# Part of SuperLocalMemory V3 | https://qualixar.com | https://varunpratap.com
+
+"""Persistent hook daemon — Unix socket server for sub-200ms recall.
+
+Eliminates Python subprocess startup (~300-500ms) by keeping a long-lived
+process that Claude Code hooks talk to via Unix domain socket.
+
+Protocol (newline-delimited JSON):
+  Client → {"prompt": "...", "session_id": "..."}\n
+  Server → {"hookSpecificOutput": {...}}\n  (or {}\n for ack/empty)
+
+MEMORY SAFETY: This module NEVER imports MemoryEngine. All recall goes
+through recall_queue.db → QueueConsumer → pool.recall(). The hook daemon
+stays at ~15-20MB RSS.
+
+Lifecycle: started by unified_daemon.py alongside QueueConsumer. If it
+crashes, auto_recall_hook.py falls back to subprocess (v3.4.35 path).
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import os
+import socket
+import threading
+import time
+from pathlib import Path
+
+logger = logging.getLogger(__name__)
+
+_DEFAULT_SOCK_NAME = "hook_daemon.sock"
+
+
+def _default_sock_path() -> Path:
+    return Path.home() / ".superlocalmemory" / _DEFAULT_SOCK_NAME
+
+
+def _default_queue_db_path() -> Path:
+    return Path.home() / ".superlocalmemory" / "recall_queue.db"
+
+
+class HookDaemon:
+    """Unix socket server for persistent auto-recall.
+
+    Accepts newline-delimited JSON requests, runs the same logic as
+    auto_recall_hook.main() but without Python startup cost.
+    """
+
+    def __init__(
+        self,
+        sock_path: Path | None = None,
+        queue_db_path: Path | None = None,
+    ) -> None:
+        self._sock_path = sock_path or _default_sock_path()
+        self._queue_db_path = queue_db_path or _default_queue_db_path()
+        self._running = False
+        self._stop_event = threading.Event()
+        self._thread: threading.Thread | None = None
+        self._server_sock: socket.socket | None = None
+        self._queue = None
+
+    @property
+    def running(self) -> bool:
+        return self._running
+
+    def start(self) -> None:
+        if self._running:
+            return
+        if self._sock_path.exists():
+            self._sock_path.unlink()
+
+        from superlocalmemory.core.recall_queue import RecallQueue
+        self._queue = RecallQueue(self._queue_db_path)
+
+        self._server_sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
+        self._server_sock.bind(str(self._sock_path))
+        self._server_sock.listen(8)
+        self._server_sock.settimeout(1.0)
+
+        self._stop_event.clear()
+        self._running = True
+        self._thread = threading.Thread(
+            target=self._accept_loop,
+            daemon=True,
+            name="slm-hook-daemon",
+        )
+        self._thread.start()
+        logger.info("HookDaemon started on %s", self._sock_path)
+
+    def stop(self) -> None:
+        if not self._running:
+            return
+        self._stop_event.set()
+        self._running = False
+        if self._server_sock is not None:
+            try:
+                self._server_sock.close()
+            except Exception:
+                pass
+            self._server_sock = None
+        if self._thread is not None:
+            self._thread.join(timeout=3.0)
+            self._thread = None
+        if self._sock_path.exists():
+            try:
+                self._sock_path.unlink()
+            except Exception:
+                pass
+        if self._queue is not None:
+            try:
+                self._queue.close()
+            except Exception:
+                pass
+            self._queue = None
+        logger.info("HookDaemon stopped")
+
+    def _accept_loop(self) -> None:
+        while not self._stop_event.is_set():
+            try:
+                client, _ = self._server_sock.accept()
+            except socket.timeout:
+                continue
+            except OSError:
+                if self._stop_event.is_set():
+                    break
+                continue
+            threading.Thread(
+                target=self._handle_client,
+                args=(client,),
+                daemon=True,
+                name="slm-hook-client",
+            ).start()
+
+    def _handle_client(self, client: socket.socket) -> None:
+        try:
+            client.settimeout(30.0)
+            data = b""
+            while b"\n" not in data:
+                chunk = client.recv(4096)
+                if not chunk:
+                    return
+                data += chunk
+
+            line = data.decode("utf-8").strip()
+            if not line:
+                client.sendall(b"{}\n")
+                return
+
+            try:
+                payload = json.loads(line)
+            except Exception:
+                client.sendall(b"{}\n")
+                return
+
+            response = self._process_request(payload)
+            client.sendall((json.dumps(response) + "\n").encode("utf-8"))
+        except Exception:
+            try:
+                client.sendall(b"{}\n")
+            except Exception:
+                pass
+        finally:
+            try:
+                client.close()
+            except Exception:
+                pass
+
+    def _process_request(self, payload: dict) -> dict:
+        from superlocalmemory.hooks.auto_recall_hook import (
+            _is_ack, _get_mode_timeout, _detect_mode, _format_envelope,
+            _DEFAULT_LIMIT,
+        )
+        from superlocalmemory.core.recall_queue import QueueTimeoutError
+
+        prompt = payload.get("prompt", "")
+        session_id = payload.get("session_id", "")
+
+        if not prompt or not isinstance(prompt, str):
+            return {}
+
+        if _is_ack(prompt):
+            return {}
+
+        try:
+            mode = _detect_mode()
+            timeout = _get_mode_timeout(mode)
+            stall_timeout = max(timeout - 5.0, 5.0)
+
+            request_id = self._queue.enqueue(
+                query=prompt,
+                limit_n=_DEFAULT_LIMIT,
+                mode=mode,
+                agent_id="hook_daemon",
+                session_id=session_id,
+                priority="high",
+                stall_timeout_s=stall_timeout,
+            )
+
+            result = self._queue.poll_result(request_id, timeout_s=timeout)
+
+            if isinstance(result, dict) and result.get("ok") is not False:
+                results = result.get("results", [])
+                if results:
+                    return _format_envelope(results)
+            return {}
+        except (QueueTimeoutError, Exception):
+            return {}
+
+
+def try_socket_recall(
+    sock_path: Path | None = None,
+    prompt: str = "",
+    session_id: str = "",
+    timeout: float = 15.0,
+) -> dict | None:
+    """Try to get recall result via the persistent hook daemon socket.
+
+    Returns the hook envelope dict on success, or None if the daemon
+    is unavailable (triggers subprocess fallback in auto_recall_hook).
+    """
+    path = sock_path or _default_sock_path()
+    if not path.exists():
+        return None
+
+    try:
+        client = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
+        client.settimeout(timeout)
+        client.connect(str(path))
+
+        request = json.dumps({"prompt": prompt, "session_id": session_id}) + "\n"
+        client.sendall(request.encode("utf-8"))
+
+        data = b""
+        while b"\n" not in data:
+            chunk = client.recv(8192)
+            if not chunk:
+                break
+            data += chunk
+
+        client.close()
+
+        if not data.strip():
+            return None
+
+        response = json.loads(data.decode("utf-8").strip())
+        return response if isinstance(response, dict) else None
+    except Exception:
+        return None
+
+
+def ensure_hook_daemon(
+    sock_path: Path | None = None,
+    queue_db_path: Path | None = None,
+) -> HookDaemon | None:
+    """Start hook daemon if not already running. Returns daemon or None."""
+    path = sock_path or _default_sock_path()
+
+    if path.exists():
+        try:
+            test = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
+            test.settimeout(1.0)
+            test.connect(str(path))
+            test.close()
+            return None
+        except Exception:
+            pass
+
+    daemon = HookDaemon(
+        sock_path=path,
+        queue_db_path=queue_db_path or _default_queue_db_path(),
+    )
+    daemon.start()
+    return daemon

{superlocalmemory-3.4.35 → superlocalmemory-3.4.36}/src/superlocalmemory/server/unified_daemon.py

@@ -438,6 +438,17 @@ async def lifespan(application: FastAPI):
             application.state.queue_consumer = _queue_consumer
             application.state.recall_queue = _recall_queue
             logger.info("QueueConsumer started (recall_queue.db)")
+
+            # v3.4.36: Start persistent hook daemon (Unix socket server).
+            # Eliminates Python subprocess startup for each recall hook call.
+            try:
+                from superlocalmemory.hooks.hook_daemon import HookDaemon
+                _hook_daemon = HookDaemon(queue_db_path=_queue_db)
+                _hook_daemon.start()
+                application.state.hook_daemon = _hook_daemon
+            except Exception as _hd_exc:
+                logger.warning("HookDaemon start failed (non-fatal): %s", _hd_exc)
+                application.state.hook_daemon = None
         except Exception as _qc_exc:
             logger.warning("QueueConsumer start failed (non-fatal): %s", _qc_exc)
             application.state.queue_consumer = None
@@ -592,6 +603,14 @@ async def lifespan(application: FastAPI):
         except Exception as exc:  # pragma: no cover — defensive
             logger.warning("bandit_tasks cancel failed: %s", exc)
 
+    # v3.4.36: Stop HookDaemon (Unix socket server).
+    _hd = getattr(application.state, "hook_daemon", None)
+    if _hd is not None:
+        try:
+            _hd.stop()
+        except Exception as exc:  # pragma: no cover — defensive
+            logger.warning("hook_daemon stop failed: %s", exc)
+
     # v3.4.26: Stop QueueConsumer (recall_queue.db drainer).
     _qc = getattr(application.state, "queue_consumer", None)
     if _qc is not None:

{superlocalmemory-3.4.35 → superlocalmemory-3.4.36/src/superlocalmemory.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: superlocalmemory
-Version: 3.4.35
+Version: 3.4.36
 Summary: Information-geometric agent memory with mathematical guarantees
 Author-email: Varun Pratap Bhardwaj <admin@superlocalmemory.com>
 License: AGPL-3.0-or-later

{superlocalmemory-3.4.35 → superlocalmemory-3.4.36}/src/superlocalmemory.egg-info/SOURCES.txt

@@ -158,6 +158,7 @@ src/superlocalmemory/hooks/context_payload.py
 src/superlocalmemory/hooks/copilot_adapter.py
 src/superlocalmemory/hooks/cross_platform_connector.py
 src/superlocalmemory/hooks/cursor_adapter.py
+src/superlocalmemory/hooks/hook_daemon.py
 src/superlocalmemory/hooks/hook_handlers.py
 src/superlocalmemory/hooks/ide_connector.py
 src/superlocalmemory/hooks/post_tool_async_hook.py