WEDA Node v1.2

v1.2.0_20261016.2

This is the first public WEDA Node v1.2 release. These notes combine the v1.2 changes since the previous public v1.1 installer, v1.1.1_20260902.1.

Download: WEDA Node Activator installers.

What's New

• Remote device operations: Added device power-management commands and support for invoking commands exposed by a WEDA Node or SubNode application. See Remote Commands.
• Telemetry validation: Added validation of telemetry configuration against the device's supported sensor settings. See Telemetry Configuration.
• Heartbeat reporting: Added the heartbeat sensor. When the System Agent configuration does not declare an hb sensor, the agent supplies a default heartbeat with a 60-second reporting interval and local recording disabled.
• Software and SDK versions: Device capabilities now report the WEDA Node software version and SDK version. SubNodes can provide their own versions during registration, and the node reports them in the SubNode list.

Behavior Changes

• Container metrics: CPU, memory, network and disk metrics are reported once per minute instead of once per second.
• Missed sensor readings: Historical recordings and batch reports retain the original time slots. A missed reading is represented as null instead of shifting later values to earlier timestamps.

Bug Fixes

Device Configuration and Identity

• Deleting a device and registering it again now resets the telemetry configuration retained on the device. Reset requests made while the device is offline are applied after reconnection, and an older reset does not overwrite a newer configuration change.
• Telemetry configuration changes and commands sent to SubNodes are no longer forwarded repeatedly after reconnection. A connected SubNode with its devices disabled can still receive the change that enables them again.
• A device disabled through telemetry configuration can be enabled again without restarting System Agent. A failed capability upload no longer blocks the later configuration change.
• The default heartbeat sensor includes its schema when System Agent reports or applies the configuration, fixing failures when reading or resetting telemetry settings.
• The node retains the identity established by the Activator instead of deriving a new identity when the primary MAC address changes. This requires the updated Activator and device management agent together.
• Device configuration is no longer echoed back to Cloud through the device-management shadow, avoiding duplicate state updates.

Container and Command Operations

• Existing container images are reused when available, avoiding unnecessary downloads during retries, agent restarts and device reboots.
• Adopting an already running stack after a restart or reboot preserves its environment variables.
• Updating a stack revision no longer allows cleanup of the previous revision to remove containers belonging to the new revision.
• Checking one stack's status no longer blocks commands for other stacks.
• Commands that have timed out in Cloud are no longer executed later by background retries. Completed commands no longer count toward the command queue limit.
• Scheduled work continues after the device clock is corrected during boot, without a burst of catch-up runs or a stalled schedule.

AI Model and File Deployment

• Concurrent model and file downloads no longer consume each other's data chunks.
• Removing a model during download cancels that download, allowing a later deployment of the same file to start cleanly.
• Concurrent status updates no longer leave a file stuck at downloading after the transfer has finished.
• A file refused by a lifecycle hook remains REJECTED, including after an agent restart, until the deployment request changes. A checksum or size validation failure is reported on the first attempt. Download failures that can recover are still retried.
• The first model deployment after an agent restart no longer waits for the periodic retry before proceeding.

For deployment steps and result interpretation, see Deploy an AI Model.

WEDA Node Activator

• Restart Service preserves the shared network used by AI model secured volumes and no longer leaves the node without service containers on Docker Compose v1 hosts. Restart failures include error E4003 and the Docker Compose output.
• x64 Advantech devices register with their actual board model instead of ADVANTECH-EDGE. Installing again after Remove in the same session also preserves the model.
• Update Authentication Credentials applies the selected credential. If the node cannot use a selected Device Credentials file, the update fails with E5005 and restores the previous credential.

Upgrade Notes

Caution: Check credentials when the board model changes

An x64 node that previously registered as ADVANTECH-EDGE cannot reuse a saved credential protected with that model after switching to its actual model. A node with an Auto Provision master key can provision again. If you use Device Credentials only, obtain a new credential for the corrected model.

See Register WEDA Node for the credential and activation steps.

Caution: Handle missing historical values

Applications that consume recorded sensor values or batch reports must accept null for a missed reading. Keep each value associated with its reported timestamp.

Package Versions

The installer package and the agents it contains have separate version numbers.

Component	Version
WEDA Node Activator installer	v1.2.0_20261016.2
WEDA Node device management agent	v1.2.0_20261016.1
System Agent	v1.2.0_20261016.1