Hearthlight

Hearthlight is an open-source framework for building camera-driven automation pipelines that run entirely on-device. Point one or more cameras at a scene, define what to watch for in plain text, and Hearthlight converts the video feed into structured events that fire downstream workflows.

It runs on CPU, Apple Silicon via MLX, and CUDA GPUs, so it works on everything from a Mac mini to a dedicated Linux workstation. Because all processing is local, no footage leaves the network, there are no per-call API costs, and nothing depends on a third-party service staying online.

The LLM integration is the practical differentiator: instead of writing custom detection logic in code, you describe triggers in natural language — 'alert when a person enters the frame' or 'log when the package is removed' — and the model translates that into working computer vision rules. Local models handle inference, keeping the full pipeline self-contained.

Typical use cases include home and workshop automation, occupancy monitoring, DIY security systems, and any event-driven logic that should react to what a camera sees. It's aimed at developers and makers who want serious computer vision capabilities without a cloud subscription or vendor lock-in.

Developers and makers use Hearthlight to turn local camera feeds into LLM-triggered automation for home, workshop, occupancy, and DIY security workflows.