NVIDIA DGX Spark Playbooks

What NVIDIA DGX Spark Playbooks does

DGX Spark Playbooks is a collection of step-by-step guides for setting up AI and machine-learning workloads on NVIDIA DGX Spark devices with Blackwell architecture. Each playbook includes prerequisites, step-by-step instructions, troubleshooting guidance and example code. Topics range from installing AI frameworks and running inference with optimized models to setting up development environments and connecting and managing the device.

Among the agent-related playbooks is one titled Run Hermes Agent with a Local LLM. Others cover running OpenClaw and NemoClaw with a local LLM, setting up CLI coding agents and Claude Code with local inference, securing AI agents with OpenShell, and serving models with llama.cpp, LM Studio, Ollama and Open WebUI. The list also includes fine-tuning, connecting multiple DGX Sparks, and image and video generation with ComfyUI.

Key features

• Step-by-step playbooks with prerequisites, troubleshooting and example code
• Run Hermes Agent with a Local LLM playbook
• Companion playbooks for OpenClaw, NemoClaw and OpenShell agent security
• Local inference with llama.cpp, LM Studio, Ollama and Open WebUI
• Multi-device connection guides for two or three DGX Spark units

When to use it

• Run Hermes Agent against a local model on a DGX Spark
• Set up local inference for CLI coding agents
• Follow a guided fine-tuning workflow on DGX Spark hardware

Who it is for: Owners of NVIDIA DGX Spark devices who want guided setups for local AI workloads, including agents.

How it fits with Hermes Agent

Includes a playbook titled Run Hermes Agent with a Local LLM, one among many DGX Spark guides rather than a Hermes-specific project.

Requirements: An NVIDIA DGX Spark device with Blackwell architecture

FAQ