OGMA – persistent memory and dual-brain AI, newcomer seeks pro feedback (github.com)

Yohan Brocard has announced OGMA, a novel personal conversational AI assistant designed with persistent hybrid memory and a dual AI architecture, set to revolutionize human-AI interactions. Unlike traditional chatbots, OGMA aims to offer a deeply personalized companionship experience by learning from and adapting to users over time. Its architecture emphasizes behavioral persistence, contextual memory retrieval, and an evolving personality, challenging conventional AI paradigms where every interaction starts fresh. This exploration is particularly significant for the AI/ML community as it addresses crucial questions about memory, ethics, and the nature of companionship in AI, indicating a shift towards more thoughtful, context-aware systems. OGMA consists of two distinct AI brains: a conversational main AI that provides empathetic dialogue and an analytical "Archivist" that enriches memory and analyzes user interactions. This setup enables OGMA to maintain a stable identity while flexibly applying ethical considerations. Key technical innovations include structured memory with SQLite and semantic similarity search using FAISS, allowing the AI to recall facts contextually. Brocard invites feedback from researchers in related fields, emphasizing the exploratory nature of OGMA as a platform to prototype and refine concepts of memory, identity, and decision-making autonomy in AI systems, reflecting an essential stride towards creating authentic, adaptable AI companions.