Know which model made every result.

A model is more than a filename in a folder.

The same checkpoint shows up in dozens of workflows under slightly different names, and the best result you ever made is only as reproducible as your memory of what produced it. Numonic reads the workflow behind each asset, resolves every model reference to a single library entry keyed by hash, and keeps the link between the model and the work. So the answer to which model made this is never a guess.

From import to browsable library.

Every import runs the same path, so the library stays accurate without any manual bookkeeping.

• Step 1
  Detect
  Import a ComfyUI image and Numonic reads the embedded workflow graph, pulling out every model reference: LoRAs, checkpoints, VAEs, CLIP encoders, ControlNets. No manual tagging.
• Step 2
  Register
  Each model becomes a library entry with a hash and a version, deduplicated by hash so the same file only appears once no matter how many workflows used it.
• Step 3
  Link
  Every entry is linked to the assets it produced, so you can open a model and see the exact results it made, and open an asset and see the models behind it.

Every model reference in the graph.

Numonic reads the embedded ComfyUI workflow JSON and identifies model references across node types. Names are normalized so the same file resolves to one entry however it was written.

• LoRA
• Checkpoint
• UNET
• VAE
• ControlNet
• CLIP
• Upscaler
• Embedding
• IP-Adapter

Built for how creative work actually happens.

The model library sits inside a larger picture.