Models API by ComfyDeploy

A collection of optimized ComfyUI-based cloud inference endpoints, built on ComfyDeploy and Modal.

Features

• Optimized model loading and caching
• Cloud-based inference endpoints
• Easy workflow management through CLI
• Extensible workflow system

Installation

uv add comfy-models

Usage

CLI Commands

uv tool install comfy-models

The package provides several CLI commands to manage workflows:

# List all available workflows
models ls

# Show detailed info about a workflow
models cat <workflow_name>

# Deploy a workflow to Modal
models deploy <workflow_name>

# Run a workflow locally
models run <workflow_name>

# Test a workflow
models test <workflow_name>

Environment Setup

Create a .env file in your project root:

MODAL_VOLUME_NAME=your_modal_volume_name
HF_TOKEN=your_huggingface_token  # Optional, for private model access

Contributing New Workflows

To add a new workflow, follow these steps:

1. Create a new directory under src/comfy_models/workflows/ with your workflow name
2. Required files:

workflows/
└── your_workflow_name/
    ├── config.py          # Workflow configuration
    ├── runner.py          # Modal app definition
    ├── workflow.json      # ComfyUI workflow definition
    └── workflow_api.json  # API interface definition

Configuration (config.py)

from comfy_models.base.comfy_utils import Config

config = Config(
    name="your_workflow_name",
    models_to_cache=[
        "path/to/model1.safetensors",
        "path/to/model2.safetensors"
    ],
    warmup_workflow=True,  # Run workflow once on startup
    run_twice=False,       # Run workflow twice for testing
    nodes_to_preload=[]    # Nodes to initialize on startup
)

Runner (runner.py)

import uuid
from comfy_models.base.base_app import _ComfyDeployRunnerModelsDownloadOptimzedImports
import modal
from comfy_models.workflows.shared import get_configs
from comfy_models.workflows.your_workflow_name.config import config

APP_NAME = config.name
app = modal.App(APP_NAME)

@app.cls(
    **get_configs(APP_NAME),
    enable_memory_snapshot=True,
)
class ComfyDeployRunner(_ComfyDeployRunnerModelsDownloadOptimzedImports):
    config = config

@app.local_entrypoint()
def main():
    ComfyDeployRunner().run.remote({
        "prompt_id": str(uuid.uuid4()),
        "inputs": {
            "your_input": "your value"
        },
    })

Workflow Files

• workflow.json: Export your workflow from ComfyUI
• workflow_api.json: Define the API interface for your workflow using ComfyDeploy nodes

Example Workflows

Check out existing workflows in the workflows directory for reference:

• flux_dev: Flux model with development settings
• flux_schnell: Optimized Flux model
• sd3_5_large: Stable Diffusion 3.5 large model