SaMu Chat

SaMu Chat is a lightweight Android client built to talk to any OpenAI compatible endpoint including your own SaMu Lab server or any cloud API you prefer. It ships as a pure native Kotlin and Jetpack Compose app with Room for local storage and OkHttp for streaming so it stays fast and small. You add one or more APIs by giving a label a base URL an API key an optional system prompt a model name a context window and a max reply length and the app saves them for reuse. Switching between APIs takes one tap from the drawer and each API keeps its own conversations and settings. Chat runs over server sent events so tokens stream into the bubble as the model produces them and a red stop button lets you cancel mid reply whenever you want. Every message can be long pressed to copy it to the clipboard or if it is your own message to edit and resend which removes everything after that point and regenerates from the edited text. While the model is thinking you see an animated thinking bubble with moving dots so the screen never feels frozen. A live context chip in the top bar shows estimated tokens used against the model context window and turns orange then red as you approach the limit. When the chat crosses ninety percent of the context window a dialog pops up offering to start a new conversation before replies start getting cut off. Conversations are stored locally in a Room database and listed in the drawer with automatic titles taken from your first message and each can be deleted with one tap. The input box sits above the keyboard thanks to ime padding so you always see what you are typing. A paperclip button on the left lets you attach any file type including video audio image PDF and zip and the attachment appears as a removable chip above the input while its filename is included with the message. Long pressing any message opens a small menu for copy and for user messages edit and resend. Everything is themed in a dark orange palette that matches SaMu Lab and the entire app runs without any analytics trackers or background services. SaMu Chat is not a model host it does not run inference itself and it needs an external OpenAI compatible server reachable over the network whether that is your own SaMu Lab on the same WiFi or a remote API on the internet. It does not render markdown yet does not send actual file bytes to the server only filenames and does not manage local model downloads. It is built for people who already have a server or an API key and want a clean fast private chat interface that does exactly what they need and nothing they do not.