What question did this study set out to answer?

The research aims to develop a memory architecture enabling AI systems to retain knowledge across multiple sessions.

April 1, 2026Open Access

ANKR SMRITI: Cross-Session Memory Architecture for AI Systems — RECALL Layer, Session Compression, and Epistemological Continuity

Key Points

The research aims to develop a memory architecture enabling AI systems to retain knowledge across multiple sessions.
Developed a three-tier memory store including session JSON and compressed data extracts.
Implemented a session compression pipeline to enhance data efficiency.
Introduced an epistemological continuity model to facilitate knowledge accumulation.
Demonstrated improved knowledge retention across sessions.
Reduced context window limitations for AI agents.
Enhanced overall performance in AI systems through better data management.

Abstract

SMRITI (Sanskrit: memory, recollection) is a cross-session memory architecture for AI systems. The protocol defines a three-tier store (session JSON, compressed SHASTRA/VIVEKA extracts, living index), a session compression pipeline, and an epistemological continuity model that allows AI agents to accumulate knowledge across sessions without context window limitations.

Read Full Paperexternally

Connected Papers

Building similarity graph...

Analyzing shared references across papers

Discussion

Institutions

Turbo Power Systems (United Kingdom)

References and Citations

Add This Paper to Your Research Feed

Any time a new paper drops it will be there.