RLM
Inference library letting language models recursively call themselves via a REPL
What it is
A Python library implementing Recursive Language Models, where an LLM handles very long inputs by loading them into a variable inside a code REPL and launching sub-model calls on pieces of it. It wraps API-based and local models behind a drop-in completion call and can run generated code locally, in IPython, Docker, or cloud sandboxes such as Modal, E2B, Daytona, and Prime. A training environment for reinforcement learning is included, and researchers and developers use it to experiment with long-context reasoning beyond a model's native window.