in30mn1a/lumimuse

LumiMuse is a self-hosted AI companion application that lets you create customizable characters to chat with over time. Rather than a generic chatbot, each character you build has its own personality, backstory, opening greeting, and example dialogue. The system keeps a persistent memory that grows through your conversations, so the character gradually "remembers" facts about you and your relationship history across sessions. When you chat, the app can generate images of the character using external image-generation services you connect yourself, such as Stable Diffusion, NovelAI, or ComfyUI. It streams AI responses in real time, lets you edit or regenerate messages, and supports attaching images and text files as context. Memory extraction runs in the background according to rules you configure, for example after every few messages or when certain keywords appear, and those memories are injected back into future conversations so the character stays consistent. You would use LumiMuse if you want a private, locally stored AI companion experience where your data never leaves your own machine or server. All characters, chats, and memories are saved in a SQLite database file on your device, and the application connects only to AI API endpoints you supply. It is built with Next.js and React on TypeScript, uses Tailwind CSS for styling, and runs through Docker for server deployment. You need to bring your own OpenAI-compatible model API key to drive the conversations. The full README is longer than what was provided.