FrostLogic Explore
Best for teams that want the deepest read on what's happening, with a human deciding what happens next
Explore is ours, so read this entry knowing that. It runs six anomaly-detection methods with causal filtering, forecasts every tracked metric one hour to seven days ahead with confidence bounds, and tracks BREEAM, LEED, and Nordic Swan criteria against live data continuously rather than at audit time. The Frostdynamics engine lets you simulate a change before anyone touches the building. Hosting is in the EU on Hetzner or in your own cluster, and your data and trained models are exportable from day one. Writes start off; you grant them one scope at a time.
The boundary: BrainBox writes setpoints unattended. Explore's writes start off; you grant a scope, and each proposal waits for approval or runs automatically inside that scope. If the priority is a model that adjusts setpoints on its own, that's BrainBox, below.