memory-manager

Agent memory organization uses a three-tier layout for episodic, semantic, and procedural structures to prevent context loss. It monitors compression thresholds and preserves snapshots, ensuring historical recall and recovery before memory is overwritten or compressed.

Monitoring agent memory compression risk requires checking overall memory usage against safe, warning, and critical thresholds. This flags when context becomes fragile, allowing you to preserve important state and create backups before memory pressure rises and causes data loss.

Categorizing flat agent memory files involves migrating them into episodic, semantic, and procedural memory buckets. This process supports manual classification of accumulated notes, decisions, and workflows, enabling agents to find relevant history quickly and maintain organized knowledge structures.

Searching agent memory by type allows you to query specific episodic, semantic, or procedural categories. The system reports breakdowns, growth metrics, and recent activity, providing ongoing maintenance statistics to track memory distribution and usage patterns across the workspace.

Agent memory initialization requires file-based setup across local workspace directories. You need a local environment to create the three-tier episodic, semantic, and procedural memory layout, plus dedicated directories for snapshots and legacy storage to enable backup and recovery operations.

Preserving agent memory snapshots is necessary when compression monitoring flags warning or critical thresholds. This saves important state before context becomes fragile or memory pressure rises, creating a recovery trail for agents with accumulated notes, decisions, and workflows in context-heavy environments.