# Aelvoxim

A self-hosted cognitive engine that plugs into any language model to give it persistent, tiered memory, a background self-improvement loop, and a runtime self-monitoring layer that watches for and corrects degraded output.

Aelvoxim is positioned as a layer that sits between an application and a language model rather than a model itself, adding four tiers of memory, working, episodic, semantic, and procedural, plus a knowledge graph so the underlying model remembers a user across sessions instead of starting from zero each conversation. A background learner loop runs continuously to improve behavior over time through curiosity-driven discovery, spaced repetition, and automatic tuning, independent of any single conversation.

A metacognition layer watches the model's own output for six kinds of degradation signal and triggers self-correction when it detects a problem, and safety is enforced through six layered, ethically gated protection rules. Because the memory and reasoning layer are described as framework-level rather than tied to one vendor, the underlying model can be swapped between GPT, Claude, DeepSeek, or a local model without losing accumulated memory, and the system is designed to run fully on CPU.

Website: https://github.com/macor24/aelvoxim