cognitive-memory

Long-term memory management for multi-agent systems routes triggered content into structured stores like core, episodic, semantic, and procedural. It updates graph indices and decay scores to maintain relevance across interactions.

Decay-based relevance scoring tracks memory usage over time to dynamically adjust importance. It updates a decay-scores.json file during remember or reflect actions, ensuring outdated information is deprioritized.

Yes, Python 3 is required to initialize the memory workspace and load the cognitive-memory templates. This setup bootstraps the multi-store architecture and prepares the system for agent collaboration.

Securing shared memory access uses gated write controls and mandatory backups to prevent data loss or leakage. It allows multiple agents to read shared stores while restricting write permissions to authorized triggers.

Remember and forget triggers classify memory content and update the knowledge graph index along with MEMORY.md. This ensures the graph accurately reflects current relationships and removes forgotten associations.

Sleep-time reflection consolidates memories by evaluating decay scores and guiding philosophical evolution of stored content. It updates audit trails to maintain an auditable workflow of structural changes.