AIPS BOOM

NVIDIA-AI-Workbench-Projects

2 min read

NVIDIA AI Workbench Projects

Type: Concept Tags: NVIDIA, AI Workbench, projects, Git, containers, reproducibility, development environment, spec.yaml Related: NVIDIA-AI-Workbench, NVIDIA-AI-Workbench-Locations, NVIDIA-AI-Workbench-Applications, NVIDIA-NIM, NVIDIA-Brev, NGC, NVIDIA-AI-Blueprints, NVIDIA-ChatRTX Sources: https://docs.nvidia.com/ai-workbench/user-guide/latest/concepts/project-concept.html, https://docs.nvidia.com/ai-workbench/user-guide/latest/how-to/projects/overview.html, https://docs.nvidia.com/ai-workbench/user-guide/latest/overview/introduction.html Last Updated: 2026-04-29

Summary

NVIDIA AI Workbench Projects are Git repositories with versioned configuration files that define containerized AI development environments. A project contains code plus Workbench metadata, build files, mounts, environment settings, and managed applications so the same repository can be cloned, rebuilt, and run across different NVIDIA-AI-Workbench-Locations.

Detail

Purpose

AI development often breaks when code, packages, CUDA versions, container images, data mounts, and local user settings drift across machines. AI Workbench Projects make the environment part of the repository while still letting host-specific and user-specific values be resolved per location.

Current scope

  • A project is a Git repository with Workbench configuration.
  • The project specification is stored under .project/spec.yaml and defines metadata, base environment, mounts, environment variables, and applications.
  • Build files such as apt.txt, requirements.txt, preBuild.bash, and postBuild.bash customize the container.
  • A base environment provides the starting container image, often with CUDA, Python, PyTorch, TensorFlow, or other GPU software.
  • The project directory is mounted into the container so code changes persist on the host.
  • Additional mounts can attach datasets, model files, shared resources, or output directories.
  • Projects can include single-container or multi-container development environments.
  • AI Workbench automates Git/container operations, but the project remains a normal Git repository and container workflow.

NVIDIA context

Projects are the portable unit in NVIDIA-AI-Workbench. They connect local RTX, cloud, server, and DGX-style development workflows to NGC containers, NVIDIA-NIM example projects, NVIDIA-AI-Blueprints, and app prototypes such as NVIDIA-ChatRTX customization.

Connections

  • NVIDIA-AI-Workbench - umbrella developer environment that manages projects.
  • NVIDIA-AI-Workbench-Locations - machines where projects are created, cloned, built, and run.
  • NVIDIA-AI-Workbench-Applications - project-managed tools, web apps, processes, and Compose apps.
  • NVIDIA-NIM - current Workbench example projects include downloadable NIM deployments.
  • NVIDIA-Brev - cloud GPU environments can become Workbench remote locations for project execution.
  • NGC - common source for base environments, containers, and model artifacts.
  • NVIDIA-AI-Blueprints - blueprint workflows can be packaged as reproducible Workbench projects.
  • NVIDIA-ChatRTX - local RAG customization workflow connected to AI Workbench.

Source Excerpts

  • NVIDIA docs describe a project as a Git repository with configuration files for a containerized environment.
  • Current docs state that project versioning keeps environments reproducible across machines.

Resources

Graph View

Backlinks