From the Deployments page, update your existing deployment to meet changing requirements, such as upgrading to a new workflow version, fine-tuning autoscaling settings, toggling it on or off for maintenance, or deleting it entirely.

​
Update Workflow Version

If you’ve updated your workflow (for example, fixing bugs, optimizing nodes, or adding features) and cloud-saved a new version, your deployment will still run the original version it was created with. To use the new version, edit the deployment and select the updated workflow version. Before upgrading, test the new version thoroughly in a ComfyUI session to confirm it runs without errors. Also, check whether any modified nodes change the inputs used in your API overrides. If they do, update your overrides accordingly to avoid broken API calls or unexpected results.

​
Change Hardware

You can adjust the GPU hardware configuration at any time to match changing requirements, such as upgrading to a higher VRAM tier for larger models or more intensive tasks.

​
Change Autoscaling

Adjust autoscaling rules, such as minimum and maximum instances, queue size, or keep-warm duration, at any time to better align with your app’s traffic patterns and usage demands, allowing you to optimize for lower latency during peaks or reduced costs in quieter times without recreating the deployment.

​
Rollout Process

Changes to your deployment, whether updating the workflow version, modifying hardware, or adjusting autoscaling rules, roll out gradually via a rolling update to avoid disruptions. For workflow version or hardware updates, existing instances continue handling in-flight requests under the previous configuration, while new instances launch with the updated setup; once ready, new requests are routed to the new instances. For autoscaling adjustments without workflow or hardware changes, the system modifies the current instance pool directly. This ensures seamless transitions, with minimal impact on ongoing operations and no need for downtime.

​
Disable Deployment

Disabling deployment immediately fails all requests, both new and ongoing, while shutting down all instances to halt billing, and preserves the configuration for easy reactivation later.

​
Enable Deployment

Enabling the deployment restarts it fully, including autoscaling based on your configured rules, so it can begin accepting and processing requests again while resuming normal billing based on usage.

​
Delete Deployment

If the deployment is no longer useful, delete it permanently from the page, this removes all settings and stops any associated costs, but your underlying workflow stays intact for recreating if desired. Before deleting, jot down your settings for reference, as the action is irreversible—always confirm before proceeding.