Sensia Hypervision — run your estate in real time, with no physical station

Sensia Hypervision is the full tier: it carries over the entire regulatory foundation of Sensia Compliance and adds day-to-day operations.

• Real-time supervision of sensors and equipment, from portfolio down to the room;
• multi-site hypervision with health indicators, MTBF, MTTR and availability;
• an alarm engine implementing the ISA-18.2 / EEMUA 191 lifecycle, wired to an intervention cycle;
• a 3D BIM/IFC digital twin, with sensors anchored to model elements;
• cross-cutting AI: anomaly detection, heat maps, maintenance anticipation.

All of it without a physical station to install and without a mandatory integrator — though working with yours remains possible, and often advisable on large estates.

LoRaWAN via The Things Network and ChirpStack, with auto-discovery on the first uplink. Helium and Loriot are wired for downlink acknowledgements; uplink ingestion for those two networks is on the roadmap. Zigbee via Zigbee2MQTT. Generic MQTT for everything else. BACnet integration is under way.

The stance is open protocols and off-the-shelf hardware: no proprietary gateway imposed, no per-device billing.

The alarm engine implements the ISA-18.2 / EEMUA 191 lifecycle: Normal, Unacknowledged, Acknowledged, RTN Unacknowledged, Shelved, Suppressed, Out-of-Service.

The difference is concrete: a control room can acknowledge, shelve and audit. An alert is not an email that gets lost, it is an object with a state, an owner and an attached intervention.

Import your IFC model and anchor sensors onto its elements. 3D navigation becomes a control surface: click a room, see its sensors, its alarms and its interventions.

The digital twin is included in the Hypervision tier, at no extra cost — a change from the previous paid add-on model.

AI is not a chatbot bolted onto the side of the product. It is spread across the use cases: anomaly detection on time series, cross-sensor correlation, intervention recommendations, alarm summarisation, natural-language assistance over your own data.

It abstains when it does not know. That is a design decision: a fabricated recommendation costs more than no recommendation.

Inference runs on OVHcloud in France, on open Qwen3 models. Your operational data and inference requests do not leave the Union, and production depends on no US AI provider.

The platform ships an AI Act compliance module: an inventory of the product's AI uses mapped to the obligations of Regulation (EU) 2024/1689. It is a deliverable you can reuse directly in tenders and governance reviews.