Skip to main content
AI-AGENT-MEMORY5 MIN READ

Separate Episodic, Semantic, and Preference Memory

Classify agent memories as episode, semantic fact, or preference before deciding how to store and retrieve them.

A memory label is a guardrail against overgeneralization. In agent memory, a single sentence can accidentally carry too much authority. "Ava asked for a shorter demo" might describe one event, a general account rule, or Ava's personal communication preference. Those are different memory types, and each should behave differently. Episode An episode is tied to a time, place, project, or interaction. Store it with context and expiry. It helps the agent explain continuity: "last call, this happened." Semantic fact A semantic fact claims general truth. Store it with source, owner, and version. It helps the agent apply a rule across…

Read the full lesson

Sign up free — one personalized lesson every day, matched to your role and goals.

Already have an account? Sign in

← Back to library
Contact us