VULCA Generate (Artwork)

VULCA Generate is the "final render" step in the ComfyUI-VULCA pipeline. Where VULCA Concept throws out a handful of exploratory sketches, this node takes a brief and produces one finished image. It's the point where intent becomes an actual artwork, ready to feed into VULCA Evaluate for a score and a critique.

If you're running the pack's flagship loop - Brief → Concept → Generate → Evaluate → Update → Generate again - this is the node that sits in the middle of the loop and the one you'll be re-running most. It's also the node where you stop being free: generation is where the provider routing actually does something.

How it works

It's a wrapper over the SDK's GeneratePhase. It loads the brief from brief_path, calls the generator with a provider, writes the image into the project directory, and returns the file path. The provider field defaults to mock, which - as with the rest of this pack - gives you a deterministic placeholder so you can build and test the workflow for free, keyless, offline.

For actual artwork you switch providers. The SDK's built-in options are gemini (Google's image models), openai (OpenAI Images), comfyui (routes back to your local ComfyUI - the SDK can use ComfyUI itself as a generation backend), and mock. Choosing comfyui is the interesting one for this audience: your "provider" is literally the engine you're already running inside.

Inputs and output

• brief_path (required) - from VULCA Brief (or VULCA Update, after a revision).
• provider - mock by default; set gemini, openai, or comfyui for real output.

Single output: image_path - a file path string, not an IMAGE tensor. That's the pack's consistent design: nodes pass paths and JSON around, so to see the result in the graph you'd load the file it points to. Real generation needs whatever credential your chosen provider wants - the pack README's Google path is export GOOGLE_API_KEY=your-key, and comfyui needs a reachable local ComfyUI.

Install and troubleshooting

cd ComfyUI/custom_nodes/
git clone https://github.com/vulca-org/comfyui-vulca
pip install "vulca>=0.12.0"

Restart, or install via ComfyUI Manager (search "VULCA"). No model files to download from this pack - it's all wrapper code; the models and keys live on the provider side.

The classic confusion: you run it on the default mock provider and get a placeholder, then wonder why your "artwork" looks like a test card. Flip the provider and set the key. And if you hit ModuleNotFoundError: vulca on the first run, that's the skipped pip install - same lazy-import gotcha as every other node in this pack.