wtf is rigel?

RIGEL is a local-first AI workspace that runs on a person's own computer instead of relying on a cloud AI service. The project's stated problem with most AI chatbots is that they forget everything between conversations, so a user has to re-explain their project and re-upload the same documents every time. RIGEL is built around keeping a persistent memory of a user's projects, conversations, preferences, and important facts across sessions, so that context builds up over time instead of resetting. The system combines several pieces into one workspace: a memory engine that remembers past interactions, the ability to run open-source language models directly on the user's own hardware rather than a hosted service, and document handling that lets a person upload PDFs, codebases, and research papers so the AI can pull in relevant context automatically instead of it being copied and pasted into every prompt. The backend is written in Rust and the interface is a React and TypeScript client. By default the app runs offline with no telemetry, no tracking, and no required account. The easiest way to try it is a portable release distributed as a ZIP file rather than something that needs to be compiled or installed with package managers. A Full version comes with a small pre-packaged language model already included so the interface works right after unzipping, while a Lite version leaves out the model to keep the download small, requiring the user to pick and download a model from within the app afterward. Both versions launch with a script for Windows, macOS, or Linux that starts the local system and opens the interface in a web browser. Because it is portable, the whole setup can also be carried on a USB drive and run on a different computer, including ones with no internet access. This is an early release labeled MK-I, described by the author as the foundation of a larger long-term plan, with future memory, document, and workflow features planned for later versions.