fix: reload persisted sessions into memory at startup (#176)

After a server restart, session_exists() returns False for sessions
that are persisted on disk, causing 404 errors on the REST endpoints.

Instead of checking disk on every request (which opens an abuse
vector), this loads all persisted sessions into memory once during
application startup. session_exists() stays a fast memory-only check.

- Added get_persisted_session_ids() to sessionmanager.py
- Added reload_persisted_sessions() to memory.py
- Called reload at startup in main.py lifespan handler

Fixes #175

Co-authored-by: Bervianto Leo Pratama <bervianto.leo@gmail.com>

Author: sharma-sugurthi

chatbot-core/api/main.py

@@ -8,7 +8,7 @@
 from fastapi.middleware.cors import CORSMiddleware
 from api.routes import chatbot
 from api.config.loader import CONFIG
-from api.services.memory import cleanup_expired_sessions
+from api.services.memory import cleanup_expired_sessions, reload_persisted_sessions
 from utils import LoggerFactory
 from pydantic import BaseModel
 
@@ -37,6 +37,9 @@ async def lifespan(app_instance: FastAPI):  # pylint: disable=unused-argument
     """
     Manages the application lifecycle, starting background tasks on startup.
     """
+    loaded = reload_persisted_sessions()
+    logger.info("Restored %s persisted session(s) from disk", loaded)
+
     # Startup: Create the cleanup task
     cleanup_task = asyncio.create_task(periodic_session_cleanup())
     logger.info("Application startup complete, background tasks initialized")

chatbot-core/api/services/memory.py

@@ -13,7 +13,8 @@
     delete_session_file,
     load_session,
     session_exists_in_json,
-    append_message
+    append_message,
+    get_persisted_session_ids
 )
 # sessionId --> {"memory": ConversationBufferMemory, "last_accessed": datetime}
 
@@ -139,6 +140,24 @@ def reset_sessions():
     with _lock:
         _sessions.clear()
 
+
+def reload_persisted_sessions() -> int:
+    """
+    Load all persisted sessions from disk into memory.
+    Called once at application startup so that session_exists()
+    can remain a fast, memory-only check.
+
+    Returns:
+        int: The number of sessions restored.
+    """
+    session_ids = get_persisted_session_ids()
+    loaded = 0
+    for session_id in session_ids:
+        if get_session(session_id) is not None:
+            loaded += 1
+    return loaded
+
+
 def get_last_accessed(session_id: str) -> Optional[datetime]:
     """
     Get the last accessed timestamp for a given session.

chatbot-core/api/services/sessionmanager.py

@@ -93,3 +93,12 @@ def delete_session_file(session_id: str) -> bool:
     Public function to delete a session's JSON file.
     """
     return _delete_session(session_id)
+
+
+def get_persisted_session_ids() -> set:
+    """Return all session IDs that have persisted JSON files on disk."""
+    return {
+        filename[:-5]
+        for filename in os.listdir(_SESSION_DIRECTORY)
+        if filename.endswith(".json")
+    }