mramangune-prog/leonardo-generative-workflow-engine

Leonardo Generative Workflow Engine describes itself as an orchestration toolkit built on top of the Leonardo AI image generation service. Rather than a simple wrapper around one API call, the README says it lets a user define multi step prompt campaigns, rotate between different Leonardo AI models such as Phoenix, Alchemy, and Canvas, and route prompts through OpenAI or Claude for refinement and critique before or after image generation. According to the README, prompts can be enriched automatically, turning a short phrase like "cat sitting on chair" into a longer, more detailed description before it reaches the image model. It describes a negative prompt library for avoiding unwanted styles, a style consistency setting meant to keep results visually similar across a batch, and an output step that can save results as PNG, JPG, WebP, PSD, or SVG files along with a metadata JSON file. A command line tool named leonardo-elite is shown running single prompts, batch jobs from a text file, and multi stage pipelines that pass a generated image through concept, refinement, variation, and upscale stages using different models at each step. Setup requires Python 3.10 or newer, or Node.js 18 or newer, along with a Leonardo AI API key, and optionally an OpenAI or Anthropic API key for the enrichment and critique features. Configuration is done through a .env file and YAML profile files that describe named art styles with specific lighting and color palette parameters. The README does not show a link to install the actual leonardo-elite package or clarify where its source lives beyond a download badge pointing to a GitHub Pages site, so readers should confirm the toolkit is what it claims to be before providing any API keys to it.