kadevin/ilab-gpt-conjure

iLab GPT Conjure is a web interface for generating images with OpenAI's GPT-image-2 model, running locally on your own machine. You open it in a browser, type a prompt, add reference images if you want, and submit. The app sends the request to an image generation API and shows you the results in a task list on the left with a preview panel on the right. It is built with Python and FastAPI and can be installed via a git clone or downloaded as a portable zip that requires no installation. The prompt editor supports three types of quick-insert shortcuts. Typing at-sign searches your local reference image library and inserts a selected image as a reference for the generation. Typing a hash sign inserts a hex color code, which is useful for pinning a brand color or background shade. Typing a tilde inserts a saved prompt snippet, which shows as a short tag in the editor but expands into the full text when the request is sent. Longer reusable prompt structures can be saved as full templates and recalled from a template library. The app runs multiple generation tasks in parallel and keeps a queue for pending ones. A history page backed by a local database lets you browse, search, and filter past results with thumbnail previews. Individual results can be retried, downloaded, archived, or added to your shared reference image library for use in future generations. Two connection modes are available. The recommended mode connects to any OpenAI-compatible image API by configuring a base URL, API key, and model name in the settings. An advanced mode can reuse a local Codex or ChatGPT login session to call internal endpoints, but this is flagged as unstable and only suitable for personal local use. The interface supports switching between Chinese and English. A command-line interface is also included for scripting and batch use. The license is AGPL-3.0.