Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.cyberun.cloud/llms.txt

Use this file to discover all available pages before exploring further.

A workflow is the template for a kind of work the platform can run. It bundles:
  • An underlying runtime (for example a ComfyUI graph or a Nerfstudio pipeline).
  • The parameters callers fill in when they submit a task.
  • The agent labels required to dispatch the task.
When you submit a task, you pick a workflow and provide values for its parameters. Dispatch picks an eligible agent based on labels.

​
Where they come from

A new team starts with an empty workflow list. You add workflows two ways:
  1. Team workflows — created by team members through the dashboard, typically by importing a ComfyUI graph JSON.
  2. Wizard-built workflows — for some pipelines, Cloud provides a form-based wizard that produces a workflow without hand-editing JSON. See Run a Nerfstudio workflow.

​
Parameters

Parameters are typed inputs the workflow exposes to callers. For a ComfyUI text-to-image graph this is typically:
ParameterTypeNotes
promptstringThe positive prompt.
negativestringOptional negative prompt.
seedintegerRandom seed. Re-using a seed reproduces an output.
stepsintegerSampler steps.
modelstringCheckpoint name available on the agent.
The workflow detail page in Cloud shows the live parameter form. You can see the same shape in the API at GET /workflows/{workflowId}.

​
Label requirements

A workflow lists the labels its tasks require: 
workflow:
  name: comfy-text-to-image
  labels:
    - gpu
    - comfyui
Only agents whose label set is a superset of this list are eligible. An agent with labels gpu, comfyui, nerfstudio is eligible for the above. An agent with only gpu is not. Tasks for a workflow with unsatisfied labels stay queued indefinitely. Cloud surfaces this in the task UI as “waiting for eligible agent”. Add the missing label to an agent (or connect a new agent) to unblock.

​
Snapshots

When snapshots are enabled on your deployment, every save of a workflow auto-creates a versioned snapshot of its definition. From the History panel on the detail page you can:
  • Preview an older snapshot’s JSON and parameter shape.
  • Restore the workflow to an older snapshot (with confirmation).
Snapshots are not deleted when you restore — they pile up so you can always go back. The History panel paginates them when the list grows. See Workflow snapshots for the full walkthrough.

​
Editing and re-running

Editing a workflow’s parameters does not retroactively change past tasks — each task carries a snapshot of the parameters it was submitted with. Re-running a task uses the workflow’s current definition unless you explicitly target a snapshot.