memory-management

To persist session context across compactions, you save your current session data with metadata and tags. This ensures long-running interactive tasks retain their context even after restarts or context window resets.

You can retrieve specific memories using semantic queries by filtering saved metadata. This allows you to fetch a memory by ID for precise context reconstruction or recall relevant session data dynamically.

Yes, claude-mem works for saving structured memory with metadata in iterative workflows. It applies to long-running tasks by storing session data persistently, satisfying requirements for structured saving and metadata filtering.

The best way to keep session data alive during long-running tasks is to store it persistently with metadata. This approach survives compactions and restarts, allowing you to later recall relevant context by query or ID.

Session context disappears after compaction because the active memory is cleared to free up space. Using a persistent storage approach to save context before compactions ensures your session data survives the restart.