Control Seat V0.1.7

V0.1.7 completes the MQTT loop. Control Seat could already read from MQTT and Sparkplug; this release adds the write path too, so operators can send values back through MQTT directly from Control Seat. It also introduces the first version of the hosted MQTT broker, making small deployments easier to bring online without separate broker infrastructure. We also tightened alarm delivery and kept the Trends Workbench responsive on very large tag systems.

What's in V0.1.7

MQTT and Sparkplug write path

Tags that source from an MQTT subscription are now first-class writable targets. In the Tags Panel, a writable MQTT tag has a value field you can edit, and a confirmation step explains where the write is going. Operators can see the pending write state in the editor until the round trip completes.

• Sparkplug-aware writes — device and node metrics are written in the correct Sparkplug format.
• Safer ownership rules — Control Seat avoids conflicting with inbound values on the same topic.
• Clearer disconnect behavior — if a subscription disappears or a broker drops, affected tags fall to bad quality and recover cleanly when data returns.
• Better write feedback — pending writes are visible immediately in the editor.

Hosted MQTT broker (edge deployments)

Small sites that do not already run an MQTT broker can now enable one inside Control Seat. It is configured from the same Gateway → Network → MQTT Brokers page as external brokers and is designed to behave cleanly in multi-node deployments.

Password rotation is a first-class action on the broker page. One click issues a new broker password and updates connected clients cleanly so old credentials are not left hanging around.

Two shapes of workspace to think about:

• Edge-only site. No existing broker. Enable the hosted broker, point your PLCs and Sparkplug edge nodes at it, and Control Seat writes to itself for tags, alarms, and history.
• Site with an existing broker. Nothing changes — external brokers keep working exactly as before. The hosted option is opt-in per tenant.

Alarm notifications: serialized, no more duplicates

Alarm notifications used to double-fire in narrow windows. V0.1.7 tightens delivery so a reopened incident or a restart edge case does not send the same page twice.

While in there, the evaluator also stopped tripping on numeric string tags — tags whose value came through the wire as "42.7" used to fail some threshold predicates because the comparison ran as a string; now they round-trip through the same numeric coercion the rest of the platform uses.

Trends Workbench at 100k+ tags

The Workbench is now much more comfortable on very large tag systems. Large tag trees load incrementally, repeated analysis is more efficient, and memory use is more predictable on workspaces with tens of thousands of bound tags.

Fit-and-finish

• Ignition connector — Java module connection edge cases around reconnect were smoothed over.
• Secret store — read/write against the Postgres secret store handles a few race conditions we saw with concurrent writers.
• Toast stack — long toast messages now wrap instead of overflowing the corner.
• Editor bindings — a bound tag with no value in the store now REST-seeds its last-known value on load, so a page opens with numbers filled in instead of blanks waiting on the WebSocket subscription.
• Cloud networking — WebSocket subscriptions are more robust on pages with very large numbers of MQTT-bound tags.
• Automatic upgrades — the data model changes in this release apply with no downtime.