Overview

Iteratively improve prompts using evaluation-driven feedback and optimization algorithms for higher-quality, more consistent AI responses.

About

Optimization is Future AGI’s prompt improvement engine. It takes a prompt, runs it against your data, scores the outputs using evaluations, and iteratively generates better versions. Instead of manually tweaking prompts, you pick an algorithm and let it explore the prompt space systematically.

Future AGI supports 6 optimization algorithms: Random Search, Bayesian Search, Meta-Prompt, ProTeGi, GEPA, and PromptWizard. Each takes a different approach to exploring and improving prompts. You can run optimization from the platform UI or programmatically via the agent-opt Python SDK.

How Optimization Connects to Other Features

  • Evaluation: Optimization uses eval scores as its objective function. Better evals lead to better optimization. Learn more
  • Datasets: Optimization runs against dataset rows. Your input/output pairs are the training ground. Learn more
  • Experiments: Compare optimized prompts against baselines using dataset experiments. Learn more

Getting Started

Understanding Optimization

How optimization works, available algorithms, and how to choose the right one.

Using the Python SDK

Run optimization programmatically with the agent-opt library.

Using the Platform

Run optimizations from the Future AGI UI with datasets and evals.
Was this page helpful?

Questions & Discussion