50K-token context reloads, replaced by a 2K local memory engine

Large Language Models (LLMs) like ChatGPT work by processing a limited "context window" (e.g., 50,000 tokens, roughly ~35,000 words) of text at once. To remember a long conversation or document, many systems simply re-feed the entire history each time — which is slow and expensive. This project (iai-personal-memory-engine) replaces that brute-force approach with a **local memory engine** that selectively stores and retrieves only the relevant past information (using ~2,000 tokens), making response generation faster and cheaper while still maintaining continuity. It's an open-source prototype showing how to build persistent, long-term memory into AI assistants without relying on costly context-window reloads.