Skip to main content

Getting Started

There are three main ways to run Yozakura. The native electron application is the fastest to set up for most users and is dead easy. The Docker image has some advantages for more advanced users who are familiar with Docker, especially on Linux or MacOS. Cloning from GitHub is mainly for development and modifications.

Electron​

Prerequisites:

  1. You need access to an LLM via a completions API. Check out the API setup guide if you don't have this already.

Installation:

  1. Find the latest release on GitHub and download the zip file for your operating system.
  2. Extract the zip file anywhere on your machine.
  3. Run the yozakura executable file inside. You might have to click through a warning about running an unsigned executable. Then follow the in-app instructions.
  4. Note that a yozakura_data folder will be created next to yozakura.exe, this is where generated images live if you want to access those directly.

Docker​

Prerequisites:​

  1. You need access to an LLM via a completions API.
  2. You need to have Docker installed

Installation:​

Using Docker compose: Save this docker-compose.yml file somewhere and run docker compose up -d in that directory.

Or without compose: Run docker run --network host -v ./yozakura_data:/app/data -d ghcr.io/mistval/yozakura:latest.

In both cases, your data will be saved to a yozakura_data folder in the current working directory.

Both methods use host network mode in order to have access to APIs you're running on your local machine. If you are not running text completions or image generation APIs on your local machine, you can remove the host network mode option. Note that host network mode might not work at all on Windows.

Give it just a few moments to start, then visit http://localhost:4396 in your web browser.

Git clone​

Check the README in the GitHub repo for the latest development setup instructions.