hermes-memory-landscape

To fix Hermes Agent memory bloat and uncontrolled injection size, use the compact worker to automatically truncate long facts and remove duplicates, while tracking memory health indicators like over-budget ratio.

Memory health checks in the Hermes Agent Memory System track quality metrics including duplicate rate, noise ratio, and over-budget ratio to monitor memory health and identify areas needing optimization.

Yes, you can customize recall weights for individual agents using the per-agent profile weighting feature, which adjusts BM25, FTS5, and HRR recall components to suit specific agent requirements.

Yes, the hermes-memory-landscape skill requires access to the Hermes Agent Memory System Phase 2 to manage structured metadata schema, quality metrics, and memory health checks.

The best way to retune recall weights for BM25 and FTS5 components is using the recall weight retuning feature, which adjusts recall component weights to optimize memory retrieval quality.

A structured metadata schema is needed for Hermes Holographic memory to provide a structured format for storing metadata with rich information, addressing weak recall quality and memory bloat challenges.