Qwen Image Edit API

OpenAI-compatible API for image editing using Qwen-Image-Edit model.

Features

• OpenAI-compatible endpoints for seamless integration with OpenWebUI/OpenRouter
• GPU-accelerated image editing using Qwen-Image-Edit
• Simple REST API with multipart form data support
• Docker support with CUDA

Requirements

• NVIDIA GPU with CUDA support
• Python 3.10+
• 16GB+ GPU memory recommended

Installation

Local Setup

1. Install dependencies:

pip install -r requirements.txt
pip install git+https://github.com/huggingface/diffusers

2. Run the server:

python run.py

The API will be available at http://localhost:8000

Note: On first run, the model (~15GB) will be downloaded from Hugging Face.

Docker Setup

1. Build and run with Docker Compose:

docker-compose up -d

The API will be accessible at http://localhost:200

Note: On first run, the model (~15GB) will be downloaded to ./model-cache/ directory. Subsequent restarts will use the cached model.

Kubernetes Deployment

Option 1: Use Pre-built Image from GitHub

The image is automatically built and published to GitHub Container Registry when code is pushed.

1. Deploy directly to Kubernetes:

kubectl apply -f k8s-deployment.yaml

Option 2: Build and Push Manually

1. Login to GitHub Container Registry:

echo $GITHUB_TOKEN | docker login ghcr.io -u YOUR_GITHUB_USERNAME --password-stdin

2. Build and push the image:

docker build -t ghcr.io/aihpi/recreate-goods-edit-api:latest .
docker push ghcr.io/aihpi/recreate-goods-edit-api:latest

3. Deploy to Kubernetes:

kubectl apply -f k8s-deployment.yaml

The service will be accessible on port 200 within the cluster (ClusterIP). Use kubectl port-forward or an Ingress for external access.

Note: The init container will download the model to a persistent volume on first deployment. The model is shared across pod restarts.

API Endpoints

Edit Image

POST /v1/images/edits

Send a multipart form with:

• image: Image file to edit
• prompt: Edit instruction (e.g., "Change the background to sunset")
• model: (optional) Model ID, defaults to "qwen-image-edit"

Returns base64-encoded edited image in OpenAI format.

List Models

GET /v1/models

Returns available models.

Health Check

GET /v1/health

Returns server status and model loading state.

Usage with OpenWebUI

1. Start the API server
2. In OpenWebUI settings, add custom OpenAI API endpoint:
   • URL: http://localhost:200/v1 (Docker) or http://localhost:8000/v1 (Local)
   • Model: qwen-image-edit

Usage with curl

# Docker/Kubernetes (port 200)
curl -X POST "http://localhost:200/v1/images/edits" \
  -H "Content-Type: multipart/form-data" \
  -F "image=@input.jpg" \
  -F "prompt=Make the sky purple" \
  -F "model=qwen-image-edit"

Environment Variables

Create a .env file if you want to change the default config (see .env.example):

• HOST: Server host (default: 0.0.0.0)
• PORT: Server port (default: 8000)
• DEVICE: cuda or cpu (auto-detected)

Model Information

Qwen-Image-Edit supports:

• Semantic image editing
• Text rendering and modification
• Style transfer
• Object manipulation
• Background changes

Troubleshooting

Out of Memory

Reduce image size or ensure you have 16GB+ GPU memory.

Model Loading Issues

Ensure you have latest diffusers:

pip install git+https://github.com/huggingface/diffusers

CUDA Not Available

Check NVIDIA drivers and PyTorch CUDA installation:

import torch
print(torch.cuda.is_available())