Persistent Identity in AI Agents: A Multi-Anchor Architecture for Resilient Memory and Continuity

TL;DR

This paper introduces soul.py, an architecture inspired by human memory systems, to enhance AI agent resilience by distributing identity across multiple memory components, preventing catastrophic forgetting.

Contribution

It proposes a multi-anchor architecture with a hybrid retrieval system and formalizes identity anchors to improve AI agent continuity and fault tolerance.

Findings

Implemented soul.py with separable identity components

Demonstrated resilience to partial memory failures

Achieved efficient retrieval with hybrid RAG+RLM system

Abstract

Modern AI agents suffer from a fundamental identity problem: when context windows overflow and conversation histories are summarized, agents experience catastrophic forgetting -- losing not just information, but continuity of self. This technical limitation reflects a deeper architectural flaw: AI agent identity is centralized in a single memory store, creating a single point of failure. Drawing on neurological case studies of human memory disorders, we observe that human identity survives damage because it is distributed across multiple systems: episodic memory, procedural memory, emotional continuity, and embodied knowledge. We present soul.py, an open-source architecture that implements persistent identity through separable components (identity files and memory logs), and propose extensions toward multi-anchor resilience. The framework introduces a hybrid RAG+RLM retrieval system that automatically routes queries to appropriate memory access patterns, achieving efficient retrieval without sacrificing comprehensiveness. We formalize the notion of identity anchors for AI systems and present a roadmap for building agents whose identity can survive partial memory failures. Code is available at github.com/menonpg/soul.py