Ollama Image Generator

Overview

This project is a simple proxy server for the Ollama Image Generator, which serves an HTML interface and forwards API requests to the Ollama API. The purpose of this is to overcome the CORS policy of many browserswhich prevents direct access to the Ollama API from a local HTML file. The main components include:

• server.py: The backend server that handles API requests and serves the HTML interface.
• index.html: The frontend interface for users to interact with the image generation functionality.

Latest Updates

• 2026-01-26: Search Ollama log to find causes of failed image creation, as well as the log entry of time taken for a successful run.
• 2026-01-24: Added support for additional models.

Prerequisites

• Python 3.9 or later
• Ollama CLI - ensure you have the latest version of Ollama installed

  ollama --version

  should be at least '0.15.0'
• (right now) macOS

Whilst you can clone the repo, you only have to download server.py and index.html to the same folder to use the application.

Architecture

• The server listens on port 8080 and forwards requests to the Ollama API located at http://localhost:11434/api/generate.
• The server acts as a proxy, forwarding requests from the HTML interface to the Ollama API and returning the generated images to the HTML interface in 'stream' format so that the progress of the image generation is displayed as it is being generated.
• The HTML interface is served directly from the server, allowing users to generate images through a web interface.
• Model Selection: Users can select from available image generation models via a dropdown menu. The selected model is automatically saved to the browser's localStorage and will be remembered on subsequent visits.
• Image History: All generated images are automatically saved to the server's history/ folder created as a sub-folder to the folder where you run server.py, providing persistent storage across browser sessions and unlimited capacity.
• The /models API endpoint returns a filtered list of available image generation models from the Ollama API.
• Tested to work with Python 3.9 and later

Choose one or more models to download.

At the time of writing the model choices are these. Your choice will depend the size of your VRAM / Unified RAM. Follow each link to see the available model sizes.

• https://ollama.com/x/flux2-klein
• https://ollama.com/x/z-image-turbo

Examples:

ollama pull x/z-image-turbo:bf16
ollama pull x/z-image-turbo:pf8
ollama pull x/z-image-turbo:latest
ollama pull x/flux2-klein:4b
ollama pull x/flux2-klein:9b
ollama pull x/flux2-klein:latest

To set up and start the server

• Clone the repository - OR - just download server.py and index.html to the same folder (<-- links to raw source files).
• From the terminal run at least one of the above ollama pull commands to pull the model(s) you want to use.
• At the top of server.py alter the value for the OLLAMA_LOG_LOCATION if necessary (by default it is set to default MacOS location)
• Run python3 server.py in the terminal.

To use the application

• Open http://localhost:8080 in your web browser.
• Type your prompt into the 'Prompt' field, adjust the controls then click 'Generate Image' to generate an image. You will see message 'Step N of X' as the image is being generated.
• If you set the 'Random' option to checked, the application will generate a random seed when you click 'Generate Image'.
• Right-click the image and select 'Save image as...' to save the image
• You can also drag and drop the image from the browser window into a folder to save it.

Image History

• All generated images are automatically saved to the history/ folder on the server.
• Each image is saved with a timestamp-based filename in the format YYYYMMDDhhmmss.png (e.g., 20260123143239.png).
• Along with each image, a corresponding JSON file is saved containing:
  • The image prompt
  • Generation settings (model, seed, width, height, steps)
  • Timestamp and unique ID
• History is persistent: Images remain available even after closing the browser or restarting the server.
• Unlimited storage: No browser storage limits - only limited by available disk space.
• Image history sidebar: View all your generated images on the left side of the interface.
• Double-click an image to load it into the interface along with the prompt and settings used to create it.
• Delete individual images: Hover over an image and click the '×' button to permanently delete it from the server.
• Delete all images: Click the 'Delete All History' button to permanently delete all images from the server.

History Folder Structure

history/
├── 20260123143239.png    # Generated image
├── 20260123143239.json   # Metadata (prompt, settings, timestamp)
├── 20260123143512.png
├── 20260123143512.json
└── ...

Backup and Restore

You can easily backup your image history:

# Backup
tar -czf history_backup_$(date +%Y%m%d).tar.gz history/

# Restore
tar -xzf history_backup_20260123.tar.gz

Key Files

• server.py: Main server logic and API handling.
• index.html: User interface for image generation.

Maintenance Notes

• server.py filters the model list to only include models capable of image generation. You will need to add any new models to the IMAGE_GEN_MODEL_LIST in server.py as well as use 'ollama pull model_name' to pull the model from the Ollama registry. Only successfully pulled models will be included in the model list on the web interface.

Author

• Nick Lansley
• GitHub

License

• MIT License