MineEcho
A source-available, local-first Memory OS for personal AI assistants that remembers user preferences and past work, learns from imported knowledge through a linked wiki and graph, routes to skills and external AI apps, and compresses context to spend fewer tokens.
MineEcho is a framework for building a private, extensible personal AI assistant rather than another chat interface. Its memory layer stores raw interaction history plus daily, weekly, and monthly summaries and links memory to knowledge instead of treating each conversation as disposable; its knowledge system turns imported sources into raw files, wiki pages, chunk indexes, and graph nodes that stay inspectable rather than becoming opaque vector chunks; and its skill layer registers both local skills and external AI apps behind one routing, health-check, and workflow surface.
Its cost layer compresses long command logs, tool outputs, URL imports, and document extraction results with scenario-aware reducers before they enter memory or context, aiming to keep the assistant useful without letting cost or noise grow unbounded over a long-running relationship with a user.
Runtime data, provider keys, chat history, audio, and metrics are kept in local state and excluded from the release tree, in keeping with its local-first design.