Developer builds local Markdown memory layer to sync context across multiple LLMs

A developer created Nexus, a local LLM-agnostic memory system, to solve the recurring problem of losing project context when switching between AI models like Claude, ChatGPT, and Gemini. Nexus stores all knowledge as plain Markdown files, making it independent of any specific platform or AI tool. It uses a hybrid search approach combining local embeddings and keyword-based BM25 search, and runs fully offline after an initial model download. Rather than deleting outdated notes, the system archives and cross-links them, preserving the reasoning behind past decisions for future reference. Every change is logged as a git commit, giving users a readable history of their entire knowledge base.