A Battery Guardrail for Nights Without Enough Sun

A hybrid solar-and-battery system does not need to reach empty before it needs help. The useful question is whether the energy left in the battery can cover the house until solar production is expected again, with some reserve left over. That became a practical question for MARA, the PowMr inverter at my family’s home in Jamaica. The system has a 24 V inverter and two 24 V, 200 Ah REDODO LiFePO4 batteries. Together, the batteries store about 9.6 kWh at their nominal ratings. The inverter’s local telemetry is available in Home Assistant through the Siseli bridge. ...

November 2, 2026 · 4 min · Shane Greaves

Bringing Siseli Inverter Telemetry into Home Assistant Locally

While I was in Jamaica, I set up a PowMr inverter with a Siseli-compatible Wi-Fi dongle. The vendor app could see the inverter, but I wanted useful, current telemetry in Home Assistant without making the cloud app stop working. I already had the Siseli Inverter Bridge integration in Home Assistant. The missing piece was a reliable local path for live readings. I used Codex as a pair programmer while I captured and inspected the traffic, worked out the Modbus request format, and built a small read-only polling path. ...

October 28, 2026 · 5 min · Shane Greaves
Two graphics cards connected to a local language model node

Two RTX 3060s, One Local 27B Model, and a GPU Unlock Experiment

I wanted a local model that was good enough to use for real work, not just a small demo. My current favorite is a 27B Qwen model, so instead of swapping it out for something smaller, I put two 12 GB RTX 3060s to work in the same inference box. The machine is called Moqorro. It runs Debian 12 and serves the model through the CUDA build of llama.cpp. Marwyn, my Hermes agent, can select this local model for private tasks while keeping its usual cloud model as the default. ...

October 26, 2026 · 4 min · Shane Greaves

Fixing Empty Responses from a Local LLM

The Symptom I spent some time chasing a frustrating failure mode in a self-hosted agent stack: the model was clearly alive, but some requests came back empty, or with enough hidden reasoning overhead that the whole system felt sluggish. The confusing part was that the usual “is the service up?” checks all looked fine. The API responded. The model was loaded on the GPU. Short prompts worked. Health checks passed. But once the prompts got larger, the system started to misbehave in ways that were hard to separate: ...

May 31, 2026 · 3 min · Shane Greaves