Architecture Overview¶

BOS is organized into four layers that communicate via gRPC (internal services) and hardware protocols (drivers to devices).

System Architecture

Network Space¶

Physical and virtual infrastructure that BOS connects to: sensors, actuators, field networks (BACnet, Modbus, etc.), and compute resources (controllers and servers).

Kernel Space — Drivers¶

Drivers translate hardware protocols into a uniform gRPC interface consumed by Device Control. BOS ships drivers for:

Industrial — BACnet, Modbus
IoT — Particle, Terabee, Kasa
Web Service — REST/HTTP-based integrations
Simulation — BOPTest and synthetic data sources for testing

Each driver exposes a small gRPC service; Device Control resolves a point URI to the correct driver at runtime using xref mappings stored in the system model.

Kernel Space — System Services¶

The core services that applications build on. All are reachable over gRPC.

Service	Role
System Abstraction (Sysmod)	RDF knowledge graph of devices, points, and spaces with semantic typing (Brick ontology)
Device Control	Protocol-agnostic read/write gateway; resolves point URIs → xrefs → driver calls
History	Time-series historian; configurable per-point sample rates
Forecast	Stores and retrieves model-generated point forecasts
Message Bus (EventBus)	Kafka-backed pub/sub bus for events, alarms, and telemetry
Shared Memory	Redis-backed key-value store for inter-app data exchange
Orchestrator (Scheduler)	Runs, schedules (cron), and event-triggers containerized apps
JSON Gateway	HTTP/REST façade over the gRPC services for external integrations

Application Space¶

Apps run as containers and interact with the kernel services via gRPC (directly or through the bospy Python library). Common app patterns include:

Dashboards — real-time visualization of sensor data and system state
Analytics — fault detection, occupancy inference, energy analysis
Reports — scheduled summaries and data exports
Control Logic — setpoint adjustment, demand response, supervisory control loops