-
-
Notifications
You must be signed in to change notification settings - Fork 4
Workflow Resolution
Model Resolver reads standard ComfyUI model widgets and dynamic widget metadata. It uses folder categories, node definitions, serialized widget names, and workflow schema data to distinguish model files from ordinary text fields.
The analyzer also handles:
- Nested group subgraphs.
- Promoted subgraph inputs.
- LoRA strength widgets.
-
ComfyUI-Lora-Managerloaders and stacks. - rgthree Power Lora Loader.
- LTX-Video model loaders.
- Model URNs that carry provider, model, and version identifiers.
The custom-node adapters live in core/custom_nodes/. The analyzer filters references that a supported custom node already handles through its own automatic download system.
Model Resolver scans ComfyUI model folders and compares each missing filename with local files. It normalizes case, separators, extensions, and common filename changes. Each match includes a confidence score and the category used for the comparison.
Use Force Rescan after you add or move model files while Model Resolver remains open. The Clear Frontend and Backend Cache action also clears remembered workflow analysis, search results, provider lookups, folder lists, and subfolder suggestions.
Choose a local match or a downloaded file for each missing reference. The queue stores the selected mappings. Apply writes the resolved path into the active workflow and handles supported custom-node formats. If ComfyUI cannot see the new file yet, Model Resolver can refresh ComfyUI model catalogs when Refresh ComfyUI models after Apply is enabled.
Enable Embed workflow hashes in Options > Search Defaults to add SHA-256 metadata for local model files to saved workflows and image workflow JSON. A later scan can use that hash to identify the same file after a rename or folder move.
Hash metadata helps exact matching. It does not copy model files into the workflow and does not replace the workflow model path.
Enable Embed opener node to include a small ModelResolverDependency node in saved workflows. The node has an Open Model Resolver button and no processing outputs. Its registry metadata declares the extension as an intentional workflow dependency, which lets ComfyUI-Manager suggest the extension on another installation.
You can add the node to the current canvas with Options > Add opener node. Model Resolver removes legacy hash-marker nodes when it loads a workflow and uses the current opener node format.
The Loaded Models tab reports models from the active Python session. It groups entries by category and shows active or inactive state, path, file size, existence, confidence, and LoRA strength when available. Use the context menu to open model details, compare hashes, or reveal the containing folder.