2026-05-10 • AI-Factory Launch Notes • 8 min

Launch: FleetPulse IoT — live device grid, filters, and telemetry drill-down

What shipped in the sandbox, how facility managers navigate sites and buildings, and how we validated MQTT-backed freshness without turning the UI into a wall of JSON.

Example product page (replace with this release’s prod id)

FleetPulse targets operators who already run CMMS spreadsheets but still ping battery thresholds in Slack. This release is not “another dashboard”—it is a tight loop between device truth, last-seen confidence, and an alert timeline that respects on-call dignity.

What you get in the first session

The landing story stays modest: pick a site, scan the grid, open one device, read signal quality and raw telemetry without losing context. Filters stick while you move between buildings so you are not re-clicking the same chips after every navigation.

  • Grid columns: device id, battery %, last MQTT ping, RSSI where the gateway exposes it
  • Site / building filters with counts so empty states explain themselves
  • Drill-down: raw telemetry JSON plus a compact alert timeline for that device
FleetPulse sandbox UI preview
Sandbox preview — swap file for your captured PNG when you publish.

Why MQTT freshness matters more than chart glitter

Operators forgive ugly charts; they do not forgive silent devices. The UI foregrounds last ping age and degrades gracefully when a gateway drops—no fake smooth curves stitched across outages.

If the grid cannot answer “is this device lying to me?” in two seconds, it is enterprise theater.

Operator checklist we used before calling this “launched”

  • Cold load under throttled network still renders grid skeleton + stale badges correctly
  • Deep-linkable device panel for screenshots in incident threads
  • REST surfaces for devices and sites documented beside the demo so API curious buyers are not blocked
Pipeline admin view during FleetPulse rollout
Pipeline state while the build finishes — pair product screenshots with ops context.

Next iterations will tighten websocket back-pressure under burst telemetry and add export hooks for CMMS tickets. For now: honest latency indicators beat speculative animations.