Skip to content

Workflow Model Importance

Control which models are required, flexible, or optional for your workflows.

Overview

ComfyDock lets you mark models in workflows as:

  • Required - Must be present to run the workflow
  • Flexible - Can substitute with similar models
  • Optional - Nice to have but not critical

This helps when sharing workflows or importing environments on machines with different model collections.

Why model importance?

When you share a workflow that uses 10 different models, not everyone may have all of them. By marking models appropriately:

  • Recipients know what's critical vs. nice-to-have
  • ComfyDock can warn about missing required models
  • Imports can succeed even without optional models
  • Team collaboration improves with clear expectations

Setting model importance

cfd workflow model importance

Walks you through:

  1. Select workflow - Choose from tracked workflows
  2. Select model - Pick model to configure
  3. Set importance - Choose required/flexible/optional

Example session:

Select workflow:
  1. portrait-generation.json
  2. anime-style.json
  3. sdxl-upscale.json

Choice: 1

Models in portrait-generation.json:
  1. sd_xl_base_1.0.safetensors (no importance set)
  2. perfectEyesXL.safetensors (no importance set)
  3. DetailedEyes.safetensors (no importance set)

Select model: 1

Set importance for sd_xl_base_1.0.safetensors:
  1. required
  2. flexible
  3. optional

Choice: 1

✓ Set sd_xl_base_1.0.safetensors to required

Non-interactive mode

Specify everything on the command line:

cfd workflow model importance portrait-generation sd_xl_base_1.0.safetensors required

Arguments:

  1. Workflow name (without .json)
  2. Model identifier (filename or hash)
  3. Importance level (required/flexible/optional)

Importance levels explained

Required

Use when:

  • Workflow won't run without this model
  • Model is core to the workflow's purpose
  • No reasonable substitutes exist

Example: Base SDXL checkpoint for SDXL workflow

cfd workflow model importance my-workflow sd_xl_base_1.0.safetensors required

Flexible

Use when:

  • Model can be swapped with alternatives
  • Any checkpoint in the same category works
  • Style/quality varies but workflow still runs

Example: Style LoRA that can be replaced

cfd workflow model importance my-workflow anime-style-lora.safetensors flexible

Optional

Use when:

  • Model enhances output but isn't necessary
  • Workflow has fallback without it
  • Nice-to-have refinement

Example: Eye detail LoRA for portraits

cfd workflow model importance my-workflow detail-eyes.safetensors optional

Viewing model importance

Use cfd status to see workflow model status:

cfd status

Output shows models by importance:

📋 Workflows:
  ⚠️  portrait-generation.json (synced)
      2 models with importance set, 1 missing optional model

For detailed view:

cfd workflow resolve portrait-generation.json --auto

Shows model resolution with importance indicators.

How ComfyDock uses importance

During cfd commit

  • Required models missing → Commit blocked (unless --allow-issues)
  • Flexible models missing → Warning shown, commit allowed
  • Optional models missing → No warning

During cfd import

  • Required models missing → Download attempted or error shown
  • Flexible models missing → Warning, suggests alternatives
  • Optional models missing → Silently skipped

During cfd repair

With --models flag:

# Download all models (default)
cfd repair --models all

# Download only required models
cfd repair --models required

# Skip model downloads entirely
cfd repair --models skip

During cfd workflow resolve

Shows importance in model resolution output:

Models:
  ✓ sd_xl_base_1.0.safetensors (required) - found
  ⚠️  anime-style.safetensors (flexible) - missing, suggest alternatives
  • detail-eyes.safetensors (optional) - missing, can proceed

Batch setting importance

For workflows with many models, you can script importance:

# Set all checkpoints as required
for model in sd_xl_base_1.0 sd15_base; do
  cfd workflow model importance my-workflow "$model.safetensors" required
done

# Set all LoRAs as flexible
for lora in style1 style2 style3; do
  cfd workflow model importance my-workflow "$lora.safetensors" flexible
done

Best practices

Recommended

  • Mark base checkpoints as required - They're essential
  • Mark style LoRAs as flexible - Users can substitute
  • Mark refinement models as optional - Nice but not critical
  • Document in commit messages - "Set model importance for sharing"

Avoid

  • Marking everything required - Makes imports fragile
  • No importance set - ComfyDock assumes required by default
  • Inconsistent across workflows - Confuses recipients

Common patterns

Portrait workflow

# Base checkpoint - required
cfd workflow model importance portrait sd_xl_base_1.0.safetensors required

# Style LoRA - flexible
cfd workflow model importance portrait portrait-style.safetensors flexible

# Eye detail - optional
cfd workflow model importance portrait perfect-eyes.safetensors optional

# Skin detail - optional
cfd workflow model importance portrait skin-detail.safetensors optional

Anime workflow

# Anime checkpoint - required
cfd workflow model importance anime anime-xl.safetensors required

# Character LoRA - flexible
cfd workflow model importance anime character-style.safetensors flexible

# Background LoRA - optional
cfd workflow model importance anime detailed-bg.safetensors optional

Clearing importance

To remove importance setting (revert to default):

Edit .cec/pyproject.toml manually and remove the importance annotation, or set it again to change the value.

Future enhancement

A cfd workflow model clear-importance command may be added for easier management.

Troubleshooting

"Model not found in workflow"

Problem: Model identifier doesn't match workflow

Solution: Use exact filename or hash from cfd status output

Changes not reflected in status

Problem: Set importance but status doesn't show it

Solution: Commit changes first: 

cfd commit -m "Set model importance"

Import still fails with optional models

Problem: Import fails even though models are optional

Solution: Check if required models are missing - those will block imports

Next steps