The product
What it does, and what it does not.
Every row marked Built below is working on real hardware and covered by a check that is run before it is called done. The rows that are not built say so.
Getting a message out
The quick announcement. Type it, choose where, send. One screen, from any page, with nothing created beforehand. It is expected to be the most-used thing in the product, because “the buses are running late” is far more common than any emergency.
The Standard Response Protocol, shipped ready. Hold, Secure, Lockdown, Evacuate, Shelter and All Clear exist and work on the day the server is installed — no message templates, confirmation requests or roles to assemble first.
Announcements in two languages. Any announcement can be spoken twice — once in English, then in Spanish — from one press. The second language is typed by the district and never machine-translated: a machine-written sentence in a lockdown is a building sent the wrong way.
Priority and preemption. A live voice outranks an all clear, which outranks an emergency, which outranks everything routine. A lockdown interrupts a bell; a test never interrupts anything real.
Rehearsal to one speaker. Practice on the real path, in one room, without the building hearing it and without any consequence that makes people avoid practicing twice.
Speakers
Registration and discovery. Compatible IP speakers register as they are. Their SIP settings are read off the registration socket as they change, with no reboot.
Naming a speaker files it. Calling something “HES Gym” puts it where it belongs rather than leaving a second filing job behind.
Per-speaker volume, and speaker lights per message. A message can drive the strobe and the sign as well as the audio.
Scale. Two speakers on a bench and five simulators. Hundreds across routed VLANs has not been run, and multicast across a campus link is a question a pilot answers rather than a bench.
The rest of the day
Bell schedules with a calendar and a day planner, and text-to-speech as a first-class bell tone rather than an audio file somebody has to produce.
SNMP in and out. Inbound traps fire announcements; outbound traps tell your monitoring what happened. A live trap watcher rather than a packet capture.
History and configuration change history. What was sent, who sent it, which speakers played it — and separately, what changed in the configuration and who changed it.
Permissions per person per school, readable as one sentence rather than as several independent axes that have to be combined in somebody's head.
Configuration export, import, backup and restore, and break-glass recovery at the console for a district locked out of its own system.
Reaching people, not just rooms
Text, voice call, rich messaging and email to staff, as a second fan-out that a cloud outage can degrade but never block — no broadcast to a speaker ever waits on it.
Built and verified against a local stand-in for the provider's API. The first real send waits on business registration, which is in progress.
Deployment
A virtual machine in Proxmox, Nutanix or VMware — open-vm-tools rather than VMware's own package, so nothing proprietary travels in the image — or on hardware, for a district with no virtual environment. Both are the same software.
One instance serves the whole district, sitting wherever the district's servers sit. Sizing guidance and a reference network diagram will be published here once the hardware target is settled.