Metadata-Version: 2.5
Name: gephyra
Version: 0.0.2
Summary: GEPHYRA: benchmarks of proactive and reactive AI agents in real-world settings (in development).
Project-URL: Homepage, https://gephyra.dev
Project-URL: Repository, https://github.com/navidns/gephyra
Project-URL: Hugging Face, https://huggingface.co/gephyra
Author-email: Navid Seidi <navid@navid.tech>
License-Expression: MIT
Keywords: agents,benchmark,evaluation,llm
Classifier: Development Status :: 1 - Planning
Classifier: Intended Audience :: Science/Research
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.10
Description-Content-Type: text/markdown

# GEPHYRA

**G**rounded **E**valuation of **P**roactive, **H**elpful **Y**et **R**eactive **A**gents.

GEPHYRA (from the Greek word for bridge, pronounced JEF-ih-ruh) is a family of benchmarks that measure how AI
agents perform in real-world settings, both when they are asked to do something (reactive) and when they must
notice and act on their own (proactive).

This package is in development. The first benchmark and its documentation will be released here.

- Website: https://gephyra.dev
- Hugging Face: https://huggingface.co/gephyra
- Code: https://github.com/navidns/gephyra

Author: Navid Seidi (https://navid.tech)
