Quickstart Guide

This guide assumes a virtualized infrastructure with appropriate NVIDIA vGPU-supported hypervisor. Please ensure all required hardware, software, and development tools are properly installed and configured prior to continuing. This toolkit details both Windows and Linux options to complete this Agentic AI Toolkit example.

Before you get started, go through the following to acquire the necessary software and hardware components.

Required Infrastructure and Supporting Software

• NVIDIA vGPU Software, version 18.0 or later

• RTX Virtual Workstation (vWS) license enablement Activate, download, and install

• NVIDIA NGC and NIM access: Join the NVIDIA Developer Program

• NVIDIA vGPU-supported hypervisor Supported Products - NVIDIA Docs

• Microsoft Windows Server 2025 + Hyper-V (if using Windows Guests)

• Windows or Linux guest Virtual Machine configured with16 vCPU, 64 GB system memory, 500 GB disk, 24Q minimum NVIDIA vGPU profile

• NVIDIA vGPU Certified Server with NVIDIA RTX PRO 6000 Blackwell Server Edition, L40S, L40, or L4. NVIDIA vGPU Certified Servers

• Tavily API Key Tavily Account and API Key

• Internet connectivity, for NIM, NGC, and github access

Required Software for Windows Guests

• Windows 11 guest VM + Ubuntu 22.04.5 LTS WSL

• Remote desktop agent (RDP, other)

• Docker Desktop Windows | Docker Docs (for Windows guest)

• NVIDIA AI Workbench: Download here

Required Software for Ubuntu Linux Guests

• Ubuntu 22.04.5 LTS guest VM

• Remote desktop agent (RDP, other)

• NVIDIA AI Workbench: Download here

• Git, text editor (nano, vim, etc.)