ComfyUI-GenAsset

ComfyUI nodes for saving and loading GenAsset generation memory.

GenAsset treats generated images as cached previews and metadata as the source of truth. These nodes let ComfyUI save a generation with its prompt, seed, model, workflow JSON, image quality metrics, and version metadata, then load that version back later for continued editing.

Nodes

Nodes appear under the ComfyUI category:

genasset

Save To GenAsset

Input:

• IMAGE
• base_url
• workspace_token
• asset_name

Output:

• image passthrough
• asset_id
• version_id
• status_json

What it captures automatically when the graph exposes it:

• prompt and negative prompt
• checkpoint/model name
• seed, steps, cfg, sampler, scheduler, denoise
• image size and batch size
• upstream node ids
• ComfyUI API prompt and workflow JSON, with token-like fields redacted
• basic tags derived from the prompt/model
• image quality metrics

It refuses to save blank or near-black frames. In that case it returns a status JSON explaining the rejection instead of creating a bad version.

Test GenAsset Connection

Input:

• base_url
• workspace_token

Output:

• workspace name
• status JSON

Use this tiny node first when helping a user. It checks that ComfyUI can reach GenAsset and that the token is valid before they run a full image workflow.

Load From GenAsset

Input:

• base_url
• workspace_token
• version_id

Output:

• preview image
• workflow JSON
• metadata JSON
• status JSON

Use this when you copied an exact version id from GenAsset.

Load Asset From GenAsset

Input:

• base_url
• workspace_token
• asset_query

Output:

• current-version preview image
• asset_id
• version_id
• workflow JSON
• metadata JSON
• status JSON

Use this as a lightweight asset browser. asset_query can be an exact asset id, an exact asset name, or search text.

Install

From your ComfyUI folder:

cd custom_nodes
git clone https://github.com/steliosot/ComfyUI-GenAsset.git

Restart ComfyUI. The nodes should appear under genasset.

No separate Python dependencies are required beyond the normal ComfyUI runtime.

Connect To GenAsset

In GenAsset:

1. Open Settings.
2. Select Tokens.
3. Create a workspace token.
4. Paste it into Test GenAsset Connection.
5. If the test succeeds, paste the same token into Save To GenAsset or Load Asset From GenAsset.

For local testing:

base_url = http://127.0.0.1:3010
workspace_token = ga_...
asset_name = your reusable asset name

For deployed GenAsset, replace base_url with your hosted app URL.

Example Workflows

Example workflows are in workflows/:

• genasset_sdxl_save_generation.json
  • Generate an image and save it to GenAsset.
• genasset_load_version.json
  • Load an exact GenAsset version and preview it.
• genasset_img2img_load_edit_save_woman_cafe.json
  • Load an asset, run img2img, and save the edit as the next version.

Typical round trip:

Generate in ComfyUI
  -> Save To GenAsset
  -> Browse asset/version in GenAsset
  -> Load Asset From GenAsset
  -> VAE Encode
  -> KSampler
  -> VAE Decode
  -> Save To GenAsset

Security Notes

• Workspace tokens are sent only to the configured GenAsset API as bearer tokens.
• Saved workflow JSON redacts fields named like workspace_token, token, authorization, api_key, secret, and service_role_key.
• Do not commit workflow files after pasting real workspace tokens into them.
• Example workflows in this repo use placeholders only.

Smoke Test

From this repo:

python scripts/smoke_import.py

Expected result:

Loaded GenAsset nodes: GenAssetLoadAsset, GenAssetLoadVersion, GenAssetSaveGeneration

License

MIT.