Skip to main content

​
Requirements

GEPA requires Python 3.10 or later (up to Python 3.14).
GEPA uses LiteLLM under the hood for LLM calls, which supports 100+ LLM providers including OpenAI, Anthropic, Google, Azure, AWS, and more.

​
Install via pip

Install GEPA from PyPI: 
pip install gepa
This installs the core package with minimal dependencies.

​
Install with Full Dependencies

For complete functionality including experiment tracking with W&B and MLflow: 
pip install gepa[full]
The full extra includes:
  • litellm — LLM API calls to 100+ providers
  • datasets — Dataset loading utilities
  • mlflow — Experiment tracking
  • wandb — Weights & Biases integration
  • tqdm — Progress bars

​
Install from Source

To install the latest development version from GitHub: 
pip install git+https://github.com/gepa-ai/gepa.git

​
Development Installation

If you want to contribute to GEPA or modify the source code:
1

Clone the repository

git clone https://github.com/gepa-ai/gepa.git
cd gepa
2

Install uv (recommended)

GEPA uses uv for dependency management:
curl -LsSf https://astral.sh/uv/install.sh | sh
3

Install dependencies

uv sync --extra dev
This installs all development dependencies including:
  • pytest — Testing framework
  • ruff — Linter and formatter
  • pyright — Type checker
  • pre-commit — Git hooks
4

Run tests

uv run pytest

​
Optional Extras

GEPA provides several optional dependency groups:

​
Test Dependencies

For running the test suite: 
pip install gepa[test]

​
Build Dependencies

For building and publishing packages: 
pip install gepa[build]

​
Development Dependencies

Includes everything needed for development: 
pip install gepa[dev]

​
gskill Dependencies

For the gskill coding agent optimization features: 
pip install gepa[gskill]

​
Environment Variables

GEPA uses LiteLLM to call LLMs. You’ll need to set up API keys for your chosen provider.
Set your LLM provider API keys as environment variables: 
export OPENAI_API_KEY="your-api-key"
For a complete list of supported providers and their environment variables, see the LiteLLM documentation.

​
Verify Installation

Verify that GEPA is installed correctly: 
import gepa

print(gepa.__version__)
You should see the version number printed without any errors.

​
Next Steps

Quick Start

Run your first optimization

Examples

Explore example notebooks

​
Troubleshooting

​
ImportError: No module named ‘litellm’

If you see this error, install the full dependencies: 
pip install gepa[full]

​
Python Version Compatibility

GEPA requires Python 3.10 or later but less than 3.15. Check your Python version: 
python --version
If you have an incompatible version, consider using pyenv or conda to manage multiple Python versions.

​
Windows Installation

On Windows, you may need to install Visual C++ build tools for some dependencies. Download from Microsoft’s website.

​
Getting Help

If you encounter issues:

Build docs developers (and LLMs) love

Get started for freeTalk to us