# gepa-ai/gepa

Optimize prompts, code, and more with AI-powered Reflective Optimization

Repository: https://github.com/gepa-ai/gepa
Canonical: https://ross.abutalabs.com/products/gepa
Homepage: https://gepa-ai.github.io/gepa/
Language: Jupyter Notebook
License: MIT
License Family: permissive
Last push: 2026-08-26T23:31:18+00:00

## Health v2 (maintenance only)
Score: 83/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 93, longevity 28
- inputs: {"age_days": 393, "days_push": 7, "days_rel": 49, "gap_med": 2, "n_releases_24m": 30}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 6254, forks 523 (observed 2026-08-28T04:09:41.004826+00:00)

## What it is
GEPA is a Python framework that optimizes textual system parameters such as prompts, code, and agent configurations using LLM-based reflection combined with Pareto-efficient evolutionary search. It reads full execution traces to diagnose failures and evolve high-performing candidates with far fewer evaluations than reinforcement learning.

## Use cases
- optimize llm prompts automatically
- improve agent accuracy with prompt tuning
- cheaper alternative to RL fine-tuning for prompts
- evolve agent architectures with AI feedback
- tune system prompts against a custom evaluation metric
- optimize code snippets with reflective mutation

## When to choose
- you have a measurable evaluation metric and want to improve prompts or agent designs
- you want prompt optimization that is cheaper and faster than RL or gradient methods
- you use DSPy or custom LLM pipelines and need automated prompt tuning

## When to avoid
- you need traditional weight-based model fine-tuning rather than textual parameter optimization
- you have no evaluation metric or dataset to score candidates against
- you want a zero-configuration one-click optimizer without wiring an evaluation loop

## Facets
- artifact type: library
- maturity: active
- function: prompt-engineering, llm-training, machine-learning, agent-framework
- domain: large-language-models, machine-learning, developer-tools
- platform: python
- tags: prompt-optimization, evolutionary-search, pareto-optimization, llm-reflection, dspy-integration, ai-agents

## Member repositories
- gepa-ai/gepa (main) score 83

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:41.004826+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:46:35.676918+00:00, confidence not recorded.
  - readme: https://github.com/gepa-ai/gepa (fetched 2026-08-28T04:09:41.004826+00:00, sha 949af958b226)
  - homepage: https://gepa-ai.github.io/gepa/ (fetched 2026-08-29T08:43:13.965467+00:00, sha 959c7629836d)
  - registry_pypi: https://pypi.org/pypi/gepa/json (fetched 2026-08-29T08:43:13.968336+00:00, sha 9bf85dfd4266)
- Data as of 2026-08-30T08:39:29.467469+00:00.