memory-consolidation

Consolidating episodic memories into semantic storage involves running review, promote, and archive phases that organize episodic data into long-term knowledge while archiving outdated information to prevent clutter and drift.

Automating memory maintenance across agent workspaces is achieved by executing periodic health checks and consolidation cycles that review, promote, archive, and decay episodic data based on configurable age thresholds and decay policies.

Memory decay and aging policies apply configurable thresholds and decay rates to balance recency against long-term retention, automatically archiving outdated episodic memories while promoting stable data into semantic storage.

You can configure custom thresholds, ages, and decay rates to define when episodic memories are promoted to semantic storage or archived, allowing you to balance recency and long-term retention within your agent workspaces.

Running a full memory health cycle from the CLI requires executing the memory-consolidation consolidate command, which triggers the complete review, promote, archive, and decay workflow to clean and optimize the memory store.

You should archive outdated memories during periodic health checks when episodic data exceeds configured age thresholds, preventing knowledge graph drift and maintaining coherent semantic storage across agent workspaces.