Open Source Projects

  • ffunes/Marstek-Venus-Energy-Manager — A Home Assistant custom integration for monitoring and controlling Marstek Venus solar-battery inverters (E v2, E v3, A, D) over IoT. Engineers can extend residential energy management tooling and apply embedded systems or local API skills to distributed solar storage.
  • NatLabRockies/sup3r — Open-source Python package that applies generative adversarial networks to downscale coarse wind and solar resource grids into high-resolution spatiotemporal synthetic data. ML engineers can adapt GAN training pipelines to improve renewable resource assessment without needing denser physical sensor networks.
  • gwittebolle/claude-carbon — A CLI tool that estimates the carbon emissions of Claude Code sessions by tracking token usage and compute. Software engineers concerned with AI sustainability can instrument their workflows and quantify the energy cost of LLM-assisted development.

Today’s Synthesis

The three tools published today sketch a coherent path for an engineer to build a self-hosted, low-carbon residential energy stack. Start with NatLabRockies/sup3r : its GAN-based downscaling of coarse solar and wind grids produces high-resolution local generation forecasts without deploying physical sensors. Feed those synthetic profiles into ffunes/Marstek-Venus-Energy-Manager via its local IoT API to automate battery charge/discharge scheduling — for example, pre-charging when sup3r predicts a midday solar surplus and discharging during evening peak tariffs. This is a concrete control-loop problem: time-series forecasting feeding a state machine against a device API, with clear ROI in reduced grid draw. Throughout development, run gwittebolle/claude-carbon to quantify the token-compute emissions of your LLM-assisted coding sessions and keep the build itself accountable. The actionable next step is a small middleware service that polls sup3r outputs on a cron, computes an optimal dispatch plan, and calls the Marstek integration’s control endpoints — a weekend-scale project that exercises ML inference, async I/O, and embedded control skills simultaneously.