Memory Flush

Automated memory management scans recent logs and promotes important entries into MEMORY.md. It reconciles conflicts, applies graded decay to stale details, and writes canonical facts with recall signals to maintain a coherent, up-to-date memory state.

Resolving contradictions in a memory store involves scanning candidate entries and keeping newer, grounded facts while removing outdated ones. This rule-based reconciliation ensures memory coherence by preserving only accurate, verified information across logs.

Graded decay applies to older, less referenced memory logs to preserve canonical facts. By scanning Aeon log data and decaying stale details, the pipeline maintains recall-readiness and links memory to related topics using keywords and timing signals.

Yes, you can maintain recall-readiness by writing canonical facts into MEMORY.md with attached recall signals. These signals use keywords and timing metadata to link memory entries to related topics, ensuring a coherent, up-to-date memory state between runs.

No dependencies are required to run a rule-based pipeline for memory curation. The process operates directly on Aeon log data across the memory store, memory/logs, and related topics to enforce scanning, conflict reconciliation, and decay without external tools.

A memory store accumulates contradictions and stale details between runs because logs continuously generate new entries without automatic curation. Without a pipeline to promote recent logs, reconcile conflicts, and decay outdated data, memory loses coherence and recall-readiness.