Retro Image Prompt Refiner

(readme generated by gemini 2.5 flash)

This project provides an interactive Gradio web application for iteratively refining image generation prompts using DSPy and the RetroDiffusion API. Nothing serious, just proof of concept.

Features

• Iterative Prompt Refinement: Use DSPy to automatically revise your prompts based on the generated images.
• Optional User Feedback: Provide ratings and comments at each step to guide the refinement process.
• Retro Diffusion Integration: Generate images using the RetroDiffusion API with various styles, palettes, and input images.

Prerequisites

• Python 3.10+
• Access to an LLM API (e.g., OpenAI, Anthropic, Gemini) with an API key.
• Access to the Retro Diffusion API with an API token for using its features.

Setup

1. Clone this repository:

   git clone https://github.com/Archelunch/retro-refiner.git
   cd retro-refine

2. Install the required Python packages:

   pip install -r requirements.txt

3. Run the Gradio application:

   python retro-dspy.py

4. Open the provided local URL in your web browser.

Usage example

Here's an example of how it works: https://x.com/mike_pavlukhin/status/1929249804904767984