PozzettiAndrea-ComfyUI-GeometryPack/tests/workflow_tests

Workflow-Based Testing

This directory contains infrastructure for executing and testing ComfyUI workflows via the ComfyUI API.

Overview

Unlike traditional unit tests that instantiate nodes directly, these tests execute actual workflow JSON files through a running ComfyUI server. This provides end-to-end validation of the complete workflow execution path.

Architecture

Core Components

WorkflowExecutor (workflow_executor.py)

• Connects to ComfyUI server via websocket
• Queues workflow prompts for execution
• Waits for completion and retrieves outputs
• Usage:

  with WorkflowExecutor() as executor:
      outputs = executor.execute_workflow_file("workflow.json")
  

WorkflowConverter (workflow_converter.py)

• Converts UI workflow JSON format to API format
• Maps widget values to input parameters
• Resolves link connections between nodes
• Usage:

  from workflow_converter import WorkflowConverter
  api_prompt = WorkflowConverter.convert_to_api_format(workflow_json)
  

MeshValidator (mesh_validators.py)

• Validates mesh geometry and topology
• Checks vertex/face counts, bounding boxes, UV coordinates
• Provides assertions for test validation
• Usage:

  is_valid, error = MeshValidator.is_valid_mesh(mesh)
  assert is_valid, error
  

Running Tests

Prerequisites

1. Start ComfyUI server:

   cd /path/to/ComfyUI
   python main.py --listen 127.0.0.1 --port 8188
   

2. Install test dependencies:

   pip install pytest websocket-client
   

Run Workflow Tests

# Run all workflow tests
pytest tests/workflow_tests/ -v -m workflow

# Run only simple workflow tests
pytest tests/workflow_tests/ -v -m workflow_simple

# Run a specific workflow test
pytest tests/workflow_tests/test_simple_workflows.py::test_preview_mesh_workflow -v

# Run unit tests (no ComfyUI server needed)
pytest tests/workflow_tests/ -v -m unit

Test Markers

Tests use pytest markers for categorization:

• @pytest.mark.workflow - All workflow execution tests
• @pytest.mark.workflow_simple - Simple workflows (2-3 nodes)
• @pytest.mark.workflow_medium - Medium workflows (4-5 nodes)
• @pytest.mark.workflow_complex - Complex workflows (6-7 nodes)
• @pytest.mark.unit - Unit tests (no server required)

Adding New Workflow Tests

1. Add workflow JSON to /workflows/ directory

2. Add to test parameters:

   SIMPLE_WORKFLOWS = [
       "preview_mesh.json",
       "your_new_workflow.json",  # Add here
   ]
   

3. Optionally add specific test:

   @pytest.mark.workflow
   def test_your_workflow(workflow_executor):
       outputs = workflow_executor.execute_workflow_file("your_workflow.json")
       assert outputs is not None
       # Add custom validations
   

Workflow Format

Workflows can be in two formats:

UI Format (from ComfyUI save):

{
  "nodes": [
    {
      "id": 1,
      "type": "GeomPackLoadMesh",
      "widgets_values": ["bunny.stl"],
      "inputs": [],
      "outputs": [{"name": "mesh", "links": [1]}]
    }
  ],
  "links": [[1, 1, 0, 2, 0, "TRIMESH"]]
}

API Format (converted automatically):

{
  "1": {
    "class_type": "GeomPackLoadMesh",
    "inputs": {
      "filename": "bunny.stl"
    }
  }
}

Troubleshooting

"ComfyUI server not available"

• Ensure ComfyUI is running at 127.0.0.1:8188
• Check firewall settings
• Verify ComfyUI started without errors

"Workflow execution failed"

• Check ComfyUI server logs for errors
• Verify all required nodes are installed
• Check that input files exist in ComfyUI input directory

"websocket-client package required"

• Install: pip install websocket-client

CI/CD Integration

For GitHub Actions, see .github/workflows/test-workflows.yml:

• Installs ComfyUI
• Starts server in background
• Runs workflow tests
• Uploads output artifacts

Future Enhancements

• Golden reference comparison (compare outputs to baseline meshes)
• Visual diff generation (render and compare images)
• Performance benchmarking
• Parallel workflow execution
• Test result dashboard