ANKR-EON: A Three-Tier Self-Evolving AI Memory Architecture with Continuous Learning for Multi-Tenant AI Systems

Read full text Summarise with VIRAT
Authors: Anil Kumar Sharma
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This document is submitted as a defensive publication. No patent rights are asserted. The submission date establishes prior art for each mechanism described. We describe ANKR-EON, a three-tier self-evolving AI memory architecture for multi-tenant AI systems, and its companion continuous learning daemon. The architecture comprises three distinct memory tiers — episodic (TimescaleDB time-indexed event store), semantic (pgvector embedding-based similarity search), and procedural (learned workflow and tool-use pattern store) — with a promotion pipeline that consolidates episodic events into semantic memories based on access frequency and importance scoring. The system implements automatic embedding generation on memory storage with an eight-provider fallback chain, dimension normalization to 384 components regardless of provider, and PostgreSQL-backed context compression with hash-keyed TTL caching achieving 70-80% token reduction. The companion continuous learning daemon monitors AI interaction streams as a background process, categorizes failure patterns across 15+ error taxonomy nodes, and generates improvement hypotheses without model retraining — adapting system behavior in real time. Multi-tenant isolation is maintained through entity-scoped memory partitions while cross-tenant behavioral patterns are shared in anonymized aggregate form. Observed improvement rates reach 15% per 100 production builds. The system represents a qualitatively different approach from simple key-value memory stores (as implemented by ChatGPT Memory), introducing temporal indexing, semantic promotion, evolution feedback loops, and autonomous improvement generation as first-class architectural primitives.

Fields of Study & Subjects

Artificial Intelligence Applications AI-based Problem Solving and Planning Advanced Database Systems and Queries

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
25 March 2026
Language
EN
Document type
Preprint