Iteratively optimize prompts and text-based agent configs against scored eval sets with GEPA

Use reflective search to improve prompts or text-configured agent components against a real eval set instead of manual prompt tweaking.

Prerequisites

Python environment, GEPA package, train and validation examples with a scoring function, model provider credentials or local models, target prompt or text configuration to optimize

Installation

Use the upstream install or setup path that matches your environment:

• pip install gepa
• pip install git+https://github.com/gepa-ai/gepa.git

Requirements and caveats from upstream:

• <a href="https://pypi.org/project/gepa/"><img src="https://img.shields.io/pypi/v/gepa?logo=python&logoColor=white&color=3776ab" alt="PyPI"></a>

Basic usage or getting-started notes:

• <a href="https://gepa-ai.github.io/gepa/guides/quickstart/"><strong>Quick Start</strong></a>  | 

• bash

• To install the latest from main:

• Source: https://github.com/gepa-ai/gepa

• Extracted from upstream docs: https://raw.githubusercontent.com/gepa-ai/gepa/HEAD/README.md

Documentation

• https://gepa-ai.github.io/gepa/

Source

• Agent Skill Exchange