Superseded by ADR-075 (status corrected 2026-08-29 — see Status correction at end of file)
MVP features require user preferences and session context persistence. The current system has in-memory session management with TTL cleanup, but lacks persistent storage for user preferences and long-term context retention across conversations and system restarts.
Implement hierarchical preference system using JSON storage with existing database patterns. Leverage existing JSON context fields in Intent, Workflow, and FeedbackDB models to store user preferences and session context without requiring schema migrations.
# UserPreferenceManager (400+ lines)
class UserPreferenceManager:
async def get_preference(self, key: str, user_id: str = None,
session_id: str = None, default: Any = None) -> Any
async def set_preference(self, key: str, value: Any, user_id: str = None,
session_id: str = None) -> bool
async def merge_preferences(self, user_id: str, session_id: str) -> Dict[str, Any]
# SessionPersistenceManager (500+ lines)
class SessionPersistenceManager:
async def persist_session(self, session: ConversationSession) -> bool
async def restore_session(self, session_id: str) -> ConversationSession
async def cleanup_expired_sessions(self) -> int
# PreferenceAPI (600+ lines)
class PreferenceAPI:
async def get_user_preferences(self, user_id: str) -> Dict[str, Any]
async def update_user_preferences(self, user_id: str, preferences: Dict[str, Any]) -> bool
async def get_session_context(self, session_id: str) -> Dict[str, Any]
Date: August 20, 2025 Author: Cursor Agent Reviewers: Chief Architect, Lead Developer —
Status corrected → Superseded by ADR-075. The JSON-field preference persistence this ADR established was functionally replaced by ADR-075’s owner-scoped personalization store; no supersession was recorded at the time — this note closes that gap.