Skip to content

Latest commit

 

History

History
130 lines (102 loc) · 5.86 KB

development.md

File metadata and controls

130 lines (102 loc) · 5.86 KB

Development and Local Testing

This guide covers setting up your local environment for developing and testing the worker-comfyui.

Both tests will use the data from test_input.json, so make your changes in there to test different workflow inputs properly.

Setup

Prerequisites

  1. Python >= 3.10
  2. pip (Python package installer)
  3. Virtual environment tool (like venv)

Steps

  1. Clone the repository (if you haven't already): 
    git clone https://github.com/runpod-workers/worker-comfyui.git
cd worker-comfyui
  2. Create a virtual environment: 
    python -m venv .venv
  3. Activate the virtual environment:
    • Windows (Command Prompt/PowerShell): 
      .\.venv\Scripts\activate
    • macOS / Linux (Bash/Zsh): 
      source ./.venv/bin/activate
  4. Install dependencies: 
    pip install -r requirements.txt

Setup for Windows (using WSL2)

Running Docker with GPU acceleration on Windows typically requires WSL2 (Windows Subsystem for Linux).

  1. Install WSL2 and a Linux distribution (like Ubuntu) following Microsoft's official guide. You generally don't need the GUI support for this.
  2. Open your Linux distribution's terminal (e.g., open Ubuntu from the Start menu or type wsl in Command Prompt/PowerShell).
  3. Update packages inside WSL: 
    sudo apt update && sudo apt upgrade -y
  4. Install Docker Engine in WSL:
  5. Install Docker Compose (if not included with Docker Engine): 
    sudo apt-get update
sudo apt-get install docker-compose-plugin # Or use the standalone binary method if preferred
  6. Install NVIDIA Container Toolkit in WSL:
    • Follow the NVIDIA Container Toolkit installation guide, ensuring you select the correct steps for your Linux distribution running inside WSL.
    • Configure Docker to use the NVIDIA runtime as default if desired, or specify it when running containers.
  7. Enable GPU Acceleration in WSL:

After completing these steps, you should be able to run Docker commands, including docker-compose, from within your WSL terminal with GPU access.

Note

  • It is generally recommended to run the Docker commands (docker build, docker-compose up) from within the WSL environment terminal for consistency with the Linux-based container environment.
  • Accessing localhost URLs (like the local API or ComfyUI) from your Windows browser while the service runs inside WSL usually works, but network configurations can sometimes cause issues.

Testing the RunPod Handler

Unit tests are provided to verify the core logic of the handler.py.

  • Run all tests:

    python -m unittest discover tests/

  • Run a specific test file:

    python -m unittest tests.test_handler

  • Run a specific test case or method:

    # Example: Run all tests in the TestRunpodWorkerComfy class
python -m unittest tests.test_handler.TestRunpodWorkerComfy

# Example: Run a single test method
python -m unittest tests.test_handler.TestRunpodWorkerComfy.test_s3_upload

Local API Simulation (using Docker Compose)

For enhanced local development and end-to-end testing, you can start a local environment using Docker Compose that includes the worker and a ComfyUI instance.

Important

  • This currently requires an NVIDIA GPU and correctly configured drivers + NVIDIA Container Toolkit (see Windows setup above if applicable).
  • Ensure Docker is running.

Steps:

  1. Set Environment Variable (Optional but Recommended):
    • While the docker-compose.yml sets SERVE_API_LOCALLY=true by default, you might manage environment variables externally (e.g., via a .env file).
    • Ensure the SERVE_API_LOCALLY environment variable is set to true for the worker service if you modify the compose file or use an .env file.
  2. Start the services: 
    # From the project root directory
docker-compose up --build
    • The --build flag ensures the image is built locally using the current state of the code and Dockerfile.
    • This will start two containers: comfyui and worker.

Access the Local Worker API

  • With the Docker Compose stack running, the worker's simulated RunPod API is accessible at: http://localhost:8000
  • You can send POST requests to http://localhost:8000/run or http://localhost:8000/runsync with the same JSON payload structure expected by the RunPod endpoint.
  • Opening http://localhost:8000/docs in your browser will show the FastAPI auto-generated documentation (Swagger UI), allowing you to interact with the API directly.

Access Local ComfyUI

  • The underlying ComfyUI instance running in the comfyui container is accessible directly at: http://localhost:8188
  • This is useful for debugging workflows or observing the ComfyUI state while testing the worker.

Stopping the Local Environment

  • Press Ctrl+C in the terminal where docker-compose up is running.
  • To ensure containers are removed, you can run: docker-compose down