Overview

View as Markdown

Welcome! We are excited to have you contribute to NeMo Gym. Whether you are adding new training environments, integrating RL frameworks, improving documentation, or fixing bugs, your contributions help advance RL training.

Ways to Contribute

New environments expand Gym’s training capabilities.

  • Novel training environments (coding, reasoning, tool use, games, and so on)
  • Benchmark integrations (such as SWE-Bench and Tau Bench)

→ Refer to contribute-environments for guidance.

Have a preferred training framework? Help integrate it with NeMo Gym.

→ Refer to training-framework-integration for integration guides.

  • New tutorials and advanced guides
  • Code examples and best practices
  • API documentation improvements

NeMo Gym ships vetted agent skills that accelerate common development workflows with AI coding assistants.

→ Refer to Agent Skills for the available skills and how to use them.

Have an idea for a new feature or improvement? Open an issue to start a discussion!

Use of AI and LLM Tools

We encourage contributors to use AI coding assistants (Cursor, Claude, Codex, OpenCode, and similar) where they genuinely help, but AI assistance does not replace human understanding, judgment, and accountability.

Guiding principle: Prefer contributions where your review and ownership clearly outweigh the maintainer review burden. You are responsible for every line of code you submit, regardless of whether you or an AI tool wrote it.

Refer to AGENTS.md for the quality bar (shared by humans and coding agents), and to Use of AI and LLM Tools for how maintainers handle low-effort submissions.

Before You Start

All contributions require DCO sign-off (git commit -s). Cryptographic commit signing is optional. Refer to development-setup for environment setup, CI/CD requirements, and DCO instructions.

Not sure where to start? Check our open issues or create a new issue to discuss your idea.