NVIDIA-NeMo/Nemotron — reverse-engineered prompt

Build me a developer resource hub for NVIDIA Nemotron models. I want it to feel like a one stop place where someone can learn what the models are, pick the right Nano, Super, or Ultra tier, then follow practical notebooks and docs to train, fine tune, evaluate, deploy, and use them in agent style workflows.

Include clear sections for reusable workflow steps, complete training recipes, usage cookbooks, and real examples like RAG, tool use, and multimodal agents if the examples fit. Add a simple Python command line experience that can discover named steps and run or chain them from YAML configs, with artifacts and outputs clearly explained. Include docs for setup, getting started, contribution, security, and a few tests so the repo feels production ready.

Please organize it cleanly, use notebooks for tutorials, and add a guided helper for Claude Code that can help users generate configs. Look up current docs online if you need to.