Skip to main content

Versioning

Every skill in the portal is versioned. A skill has a single parent record (name, icon, color) and one or more skill versions beneath it. Each version contains the full container configuration: the Docker image reference, inputs, outputs, environment variables, and privileges.

Why versions matter​

Versioning lets you update a skill — new model weights, a new Docker image, changed inputs — without disrupting existing deployments. Workflows that are already running on version 1.0.0 continue to do so until you explicitly upgrade them to 1.1.0. This gives you control over when and which devices receive updated software.

Version numbers​

Version numbers use semantic versioning conventions (e.g., 1.0.0, 2.1.3). Each version number must be unique per skill — you cannot create two versions of the same skill with the same version string.

Adding a new version​

To add a new version to an existing skill:

  1. Open the skill from your Library.
  2. Click Add Version (or New Version, depending on context).
  3. Configure the new version — you can modify the Docker image URL, inputs, outputs, environment variables, and privileges independently from older versions.
  4. Click Save.

The new version appears alongside existing versions in the skill detail view. Existing workflows are not automatically upgraded — they remain on their current version until you update them.

Upgrading a workflow to a new version​

When a new version is available for a skill used in a workflow, the portal indicates an update is available. To upgrade:

  1. Open the workflow.
  2. Open the skill configuration within the workflow.
  3. Select the new version from the version dropdown.
  4. Review and update any input values if the inputs have changed.
  5. Save and Sync the workflow to push the new version to your devices.

Devices that are online will receive and restart with the new version during the next sync.

Deleting a skill or version​

Deleting a version​

You can delete a skill version that is not currently in use by any workflow. If the version is referenced by an active workflow, you must remove it from the workflow first.

Deleting a skill​

Deleting a skill removes all of its versions and marks any associated workflow skill records as deleted. Workflows that referenced the deleted skill will be out of sync and will need to be updated. Historical data from past deployments is preserved.

warning

Deleting a skill is permanent. Affected workflows will need to be reconfigured with a different skill before they can be synced again.

Version snapshot​

When a skill or version is deleted, the portal preserves a snapshot of its configuration at the time of deletion. This means historical workflow records and output data remain legible — you can still see what skill and version a device was running even after the original skill is gone.