# Springhouse: Off-Grid AI Cluster & Automated Greenhouse

## 1. Project Summary

Springhouse is a closed-loop, off-grid technical ecosystem built around solar power, local automation, and edge AI. The full build plan includes 2,800 watts of solar capacity. The current live deployment is now 500 watts on the 24V system plus 200 watts on the primary 12V system, for 700 watts actively in use while the remaining panels wait on permanent ground-mount placement.

The project is transitioning toward a standalone DC microgrid that can power environmental sensors, cameras, greenhouse and coop automation, pumps, fans, tool charging, and compute loads including an Intel i9 + NVIDIA GPU cluster. The SFF Home Assistant node is already running directly from the 24V battery bank through the charge controller’s DC load circuit.

As of the late-April build logs, the 2x 8D Duracell lead-acid batteries are topped up and recovering after winter, the first 5-panel solar string is installed on a temporary ground mount, and permanent placement testing is underway. The coop and greenhouse build have also moved from layout into field construction: posts and footings are going in, the coop has been moved into place, and the old flat roof is being reshaped into a curved greenhouse-plastic canopy.

## 2. The Green-Powered Solution: DC Distribution

The core problem with running traditional tech off-grid is the DC-AC-DC penalty. Solar panels output DC, batteries store DC, and many sensors, cameras, controllers, and small computers ultimately run on DC. A conventional off-grid setup often sends that energy through an inverter to make AC, then through wall adapters to convert it back to DC.

Springhouse is built to avoid that wherever practical.

Current power architecture:

- Harvest: 500W live on the 24V system, 200W live on the primary 12V system, with 2,800W total panel capacity in the project plan.
- Storage: Reconditioned 24V lead-acid battery bank using 2x 8D Duracell batteries, currently topped up and recovering capacity after winter.
- Distribution: DC rails and DC-DC converters for low-voltage loads such as sensors, cameras, automation nodes, and control hardware.
- Heavy Compute: Larger compute loads remain isolated through a pure sine wave inverter path while the DC transition continues.
- Field Work: Solar is already supporting practical build work, including cordless tool charging through a smaller inverter.

The goal is not just to prove that solar can blink a light. The goal is to run a working rural technical system from harvested energy: automation, documentation, security, compute, and fabrication support.

## 3. Pushing the Limits: The Solar AI Cluster

Many solar projects aim for “least power.” Springhouse also asks a “most power” question: how much useful local AI and automation work can be done from a solar-powered, off-grid system?

The current cluster and control stack includes:

- Main compute node: Intel i9 desktop-class system with NVIDIA GPU hardware for local AI inference.
- Automation hub: SFF Home Assistant node now running directly from the 24V battery bank.
- Local AI command surface: software for local-first AI operations, planning, documentation, and workflow coordination.
- Edge terminals: Laptops and rugged field hardware for project work and greenhouse/coop documentation.
- Sensor swarm: Raspberry Pi and ESP32-class nodes for local monitoring and automation.

The software side is advancing alongside the physical build. The local-first command surface is being tested as a practical control layer for AI planning, reports, charts, photos, build logs, hardware documentation, and longer overnight planning runs.

The purpose of the software stack is to keep the project usable without depending entirely on cloud services. It supports local project memory, local reports, local automation notes, field documentation, and AI-assisted planning while the physical system continues to grow.

## 4. The Automated Green...