Memory Operations

Fresh

Hindsight provides four core memory operations plus Memory Banks as the organizational container.

Operation Summary

Operation	Purpose	Token Cost	Output
Retain	Store and process content	Input + extraction	Categorized memories
Recall	Search with TEMPR	Query + results	Ranked raw memories
Reflect	AI reasoning over memories	Query + context + generation	Synthesized answer with citations
Mental Models	Pre-computed reflections	Reflect cost on create/refresh; lightweight on read	Cached insight documents

Memory Types

Hindsight automatically categorizes retained content into three types:

Type	Description	Example
World Facts	Objective external information	"Project deadline is March 15th"
Experience	Events and interactions	"User signed up for premium last week"
Observations	Synthesized consolidated knowledge	"User comfortable with async Python"

TEMPR Retrieval

All search operations use TEMPR multi-strategy search:

Pages in This Section

• Memory Banks — Create and configure isolated memory containers
• Retain — Store and process content into memories
• Recall — Search memories with TEMPR retrieval
• Reflect — AI-powered reasoning with citations
• Mental Models — Pre-computed cached reflections