alexzhang13/rlm

General plug-and-play inference library for Recursive Language Models (RLMs), supporting various sandboxes.

What it solves

RLMs address the challenge of handling near-infinite length contexts in language models. Instead of passing a massive prompt to a single model call, this paradigm allows a model to programmatically decompose a task and recursively call itself to process information in smaller, manageable pieces.

How it works

RLMs replace the standard completion call with a system that treats the context as a variable within a REPL (Read-Eval-Print Loop) environment. The language model interacts with this environment by writing and executing code. Within this code, the model can launch sub-LM calls as functions, treating prompts and contexts as objects. This shifts the interaction from JSON-based tool calling to a CodeAct-style harness where the model manages its own execution flow and recursive sub-calls.

Who it’s for

This project is for AI researchers and developers building advanced inference engines or agents that need to process extremely long contexts or perform complex, multi-step reasoning that requires programmatic control.

Highlights

  • Flexible REPL Environments: Supports a variety of execution environments including local (exec/IPython), Docker, and cloud-based sandboxes (Modal, Prime Intellect, E2B, Daytona).
  • Recursive Inference Engine: Provides an extensible engine that enables models to call themselves recursively to solve tasks.
  • Integrated Training Harness: Includes a training environment based on prime-rl and verifiers to train models specifically for the RLM paradigm.
  • Trajectory Logging: Features a logger and a dedicated visualizer to inspect the full execution path of code, sub-LM, and root-LM calls.

Related

  • Project
    avbiswas/fast-rlm

    A minimal implementation of Recursive Language Models (RLMs) that allows LLMs to handle arbitrarily long prompts by interacting with data through an external REPL and recursive sub-agents.

  • Project
    grishahq/recursive-llm

    A Python library implementing Recursive Language Models that enables large-context analysis by keeping source text in a REPL and allowing the model to selectively query it.

  • Project
    microsoft/LMOps

    A research initiative focused on fundamental technologies to improve LLM prompting, inference speed, and context handling for building AI products.

  • Project
    Trampoline-AI/predict-rlm

    predict-rlm is a Python library that gives large language models a sandboxed, stateful runtime. Models write and execute Python code, keep mutable state, call focused sub‑models, and use tools/skills (e.g., PDF handling, spreadsheets). The framework records a full trace for observability and can feed those traces to optimisation tools like GEPA. It’s aimed at complex, multi‑step workflows such as document analysis, audit, or spreadsheet automation, and supports optional Docker sandboxing and rich callbacks.

  • Project
    rllm-org/rllm

    An open-source framework for training language agents with reinforcement learning, allowing the same agent code to be used across different backends, sandboxes, and benchmarks.