SLUELY: A REVIEW OF THE ALWAYS-ON AI ASSISTANT THAT NEVER LEAVES YOUR MACHINE

Sluely represents a local-first approach to always-available desktop assistance. Its central idea is to combine recent screen and audio context with local language-model inference and present help through a lightweight overlay. The described live workflow—screen sampling, on-device audio capture, a rolling context buffer, Ollama streaming, and overlay presentation—offers a foundation for in-context questions and meeting-oriented assistance. Its value depends not only on model quality but also on reliable context selection, low-latency processing, transparent capture controls, and responsible handling of sensitive information. Future work should prioritize task-mode switching, carefully bounded episodic memory, configurable proactive triggers, cross-platform permission handling, and systematic evaluation. Before making strong claims about accuracy, privacy, or efficiency, the project should publish reproducible measurements and a clear data-flow description covering local storage, network activity, retention, and deletion.

[1]  Ollama, “Ollama API Documentation.” Available: https://docs.ollama.com/api [2]  A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust Speech Recognition via Large-Scale Weak Supervision,” Proceedings of the 40th International Conference on Machine Learning, PMLR 202, pp. 28492–28518, 2023. Available: https://proceedings.mlr.press/v202/radford23a.html [3]  Apple Support, “Control access to screen and system audio recording on Mac.” Available: https://support.apple.com/guide/mac-help/control-acces s-screen-system-audio-recording-mchld6aa7d23/mac [4]  Microsoft Support, “Windows camera, microphone, and privacy.” Available: https://support.microsoft.com/en-us/windows/privacy/window s-camera-microphone-and-privacy