Frok

Appearance

A ComfyUI image pipeline ​

The SDXL Turbo example is a ComfyUI API graph plus metadata. It uses the same folder convention as a Vpipe pipeline, with run.json instead of run.vpipeline.

Install the example ​

Create ~/frok/pipelines/image/my-sdxl.local/ and copy:

When keeping the bundled workflow alongside it, change the copied metadata ID to comfyui:my-sdxl, give it a distinct name, and set default to false.

Start ComfyUI and connect it in Settings → Services. Refresh Pipelines, select the new image workflow and prepare any missing dependencies. The metadata declares the checkpoint location and its download source, so a separate preparation file is not required for this known download.

API format matters ​

An API graph maps node IDs to their class and inputs:

json
{
  "10": {
    "class_type": "CLIPTextEncode",
    "inputs": { "text": "A quiet mountain lake", "clip": ["4", 1] }
  }
}

This is a structural excerpt, not a complete generation graph. A Frok binding such as {"node":"10","field":"text"} targets that node's input. Visual editor coordinates, links and UI state are not an API graph. Export in API format or use the complete downloadable example above.

Add another checkpoint or sampler ​

Edit your native graph, then align the model reference and download requirements in meta.json. Keep negative prompt bindings separate from the main user prompt. Bind every output so Frok can collect job-specific files. Required custom node classes must exist on the connected ComfyUI service before the workflow becomes ready.

The bundled Z-Image-Turbo workflow follows the same structure. Use its own model and encoder requirements; swapping a checkpoint filename in an unrelated graph is not a complete conversion.

Import a workflow with Utilities →