I built an open source framework for building RL environments. Named "Seahaven" after the fake town in The Truman Show.

A developer has created an open-source framework called "Seahaven" for building Reinforcement Learning (RL) environments, named after the fictional town in The Truman Show. This framework is designed to optimize long-running agents by allowing developers to rewrite prompts, tools, skills, and subagents, retaining improvements that yield better scores. An example environment, "Stripe World," mimics Stripe's API with 24 tables and 155 API operations, enabling the official Stripe SDK to function without modification.