This project provides an AI image generation WebUI workbench for GPT-image-2, offering both a graphical interface and a command-line interface (CLI) for automation. It caters to users seeking a streamlined workflow for image generation, editing, and reference tasks, supporting flexible API integrations and local resource management. The workbench aims to enhance productivity through features like concurrent task handling, prompt templating, and a shared asset gallery.

How It Works

The core of iLab GPT Conjure is its dual-mode authentication and API integration. Users can opt for a stable OpenAI-compatible API mode, recommended for team use or production environments, where configurations like Base URL, API Key, and model are set within the WebUI. Alternatively, an Advanced Native Mode leverages local Codex/ChatGPT OAuth sessions for personal, local workflows, though this is explicitly noted as unofficial, potentially unstable, and subject to API changes and usage restrictions. The system manages multiple tasks concurrently with a local queue, stores history with pagination and search, and utilizes "chips" for quick insertion of gallery references, color codes, or prompt snippets into prompts.

Quick Start & Requirements

• Primary Install: Clone the repository, create a Python virtual environment, and install dependencies:

  git clone https://github.com/kadevin/ilab-gpt-conjure.git
  cd ilab-gpt-conjure
  python3 -m venv .venv
  .venv/bin/python -m pip install -r requirements-webui.txt
  

• Requirements: Python 3.11 or higher. Frontend development requires Node.js tools specified in package.json.
• Execution: Launch the WebUI via provided scripts (Start WebUI.command/.bat) or manually with uvicorn. CLI usage is demonstrated with python -m codex_image --prompt "...".
• One-Click Packages: Pre-built, portable packages are available on the Releases page for immediate use.
• Links: GitHub Releases

Highlighted Details

• Supports GPT-image-2 for text-to-image, reference image generation, and image editing.
• Offers both OpenAI-compatible API and a personal Codex/ChatGPT OAuth integration mode.
• Features a public/shared gallery, multi-type prompt chips (gallery, color, snippets), and prompt templates for efficient prompt engineering.
• Includes concurrent task processing, a local queue, paginated history with search/filter, and multi-language support (Chinese/English).

Maintenance & Community

The project offers custom development services and communication via WeChat for AI programming and image generation workflow discussions. Specific details on core maintainers, sponsorships, or public community channels like Discord/Slack are not provided in the README.

Licensing & Compatibility

The project is licensed under GNU AGPLv3. This license requires that any modifications provided as a network service must also be made available under the AGPLv3. This copyleft provision primarily applies to the software itself, not to the underlying models, API services, or user-generated data/prompts.

Limitations & Caveats

The "Advanced Native Mode" using Codex/ChatGPT OAuth is explicitly discouraged for production, team, or stability-critical use due to its unofficial nature, potential for frequent changes, and susceptibility to account/usage restrictions. The AGPLv3 license's strong copyleft requirements must be carefully considered for any network-based service deployment. macOS portable packages may require manual quarantine attribute removal (xattr -dr com.apple.quarantine).