memory-system

You can persist user preferences across sessions using a memory system that stores and recalls facts globally and per project. This maintains contextual continuity by applying learned insights and preferences to future conversations automatically.

Managing context scope when recalling long-term memory is best handled with controlled scope operations and memory pruning. This retrieves only relevant project facts and global preferences, keeping the active context focused and preventing information overload.

Event logging for an auditable memory works by recording fact storage and recall operations within a tiered memory architecture. This creates a traceable history of changes alongside an active_context.md file that guides ongoing session behavior.

Yes, you can use a tiered memory architecture to handle both global preferences and project-specific facts. It applies standardized fact schemas to manage different scopes, ensuring project data remains isolated from global user learnings.

You should use memory pruning operations in session management when accumulated facts become outdated or exceed relevant scope. Pruning removes irrelevant data from the active context, ensuring recall operations remain efficient and highly targeted.