semantic-memory

Store learnings using semantic-memory_store() with information and metadata tags, then retrieve them later with semantic-memory_find() queries. Memories persist in a searchable index, letting you access past solutions and decisions without re-solving the same problems.

Store the rationale (WHY) alongside solutions, outcomes, and architectural decisions with relevant metadata tags. Include context like problem type, technology stack, and decision reasoning so future queries surface the most relevant learnings.

Use semantic-memory_validate() to reset the decay timer on memories you've confirmed are still correct. Validation prevents stale information from becoming unreliable, maintaining a high-quality searchable knowledge base.

Yes, semantic memory is designed to persist long-term learnings of solutions, architectural decisions, and recurring patterns specific to your projects, making them queryable before starting new work.

Without validation, memories decay and lose reliability over time. Decay-based validation is a core requirement, so explicitly refreshing important learnings keeps your searchable index trustworthy and actionable.