Sol OS
Compute that knows what the sun and the battery are doing
Sol OS sits between your site's energy data and your AI workloads. It does not control electrical equipment — it decides what compute runs, when, and under whose policy.
Operator view
A dashboard that answers one question honestly
How much of today's work ran on energy you generated — and what did the rest cost?
Site generation
4.2 kW
Battery state of charge
61%
Node draw
410 W
Work on local energy today
76%
Sample data shown for illustration. Live figures depend on your site and hardware.
Architecture
Four layers, no hidden control
Every decision the scheduler makes is traceable to a rule you can read and change.
Telemetry
Read-only feeds from inverter, meter and battery, normalised to a single sign convention with staleness handling.
Policy
Thresholds you write: battery reserve floor, surplus trigger, tariff windows, maximum deferral time.
Scheduler
Classifies each job as interactive, batch or background and decides run, defer or pause each interval.
Reporting
Energy per job, share of work on local generation, deferrals honoured and missed, exported as CSV.
Policies
Choose the objective, then live with the trade-off
The same node produces very different outcomes depending on what you optimise for.
Solar first
Batch work waits for surplus generation. Interactive work always runs.
72% local energy (sample site)
Lowest cost
Work shifts to the cheapest tariff window, using stored energy when export value is low.
58% local energy (sample site)
Lowest carbon
Deferrable work follows grid carbon intensity as well as on-site generation.
64% local energy (sample site)