branavanselvasingham/ai-presence-kit

AI Presence Kit is an experiment in making AI chat interfaces feel alive while they are working, instead of showing a frozen screen or a spinner while waiting for a response. It takes runtime signals such as the user typing, the system reading, waiting, thinking, streaming a reply, speaking, being interrupted, or hitting an error, and turns each of those moments into a visible interface state. The project includes a simple SVG face as a working example. The face is not meant to guess how a user feels. Instead, it uses small movements such as gaze direction, blinking, eyebrow position, mouth shape, posture, and motion to show that the AI is present and paying attention before, during, and after it produces a response. The core idea is kept separate from the face itself, so other kinds of visual indicators, not just a face, could use the same underlying state information. The project is organized into a few planned packages: a core state engine that tracks the presence signals, a set of React hooks and components for using it in web apps, the reference face itself, and optional adapters that connect it to common AI tools. The repository includes demo pages that compare a plain loading indicator against the presence system, along with technical markers in the page that record exactly how quickly the presence state appears compared to when the AI's actual reply text shows up. The author frames this as filling a gap between static avatar libraries, which just give an AI a fixed look, and full chat interface frameworks, which handle messages and threads but not this kind of moment-to-moment attentiveness. The project is still early stage and is actively looking for collaborators interested in interaction design, rendering, or building non-face versions of the same presence idea. The full README is longer than what was shown.