Dell Deskside Agentic AI Solutions for NVIDIA

Open-source reference architectures for enterprise agentic AI on Dell deskside systems powered by NVIDIA. Each solution pairs NVIDIA reference material with S/M/L sizing, a runnable deployment recipe, measured performance and an IT-Admin guide, so IT teams know what to deploy and what to expect.

Solutions

Solution	What it delivers
Agentic Coding	A private, on-deskside AI coding assistant — code generation, explanation, refactoring, and agentic CLI/IDE workflows served locally behind an OpenAI-compatible API.
Agentic Video Search & Summarization	Natural-language search, Q&A, and summarization over local video libraries — vision-language plus LLM serving on the same deskside tiers.
OpenClaw on NemoClaw	The OpenClaw assistant on NVIDIA's NemoClaw runtime — policy engine, OpenShell sandbox, and tool-using agents on a locally served LLM.
Hermes on NemoClaw	Nous Research' Hermes Agent harness on the NemoClaw runtime — served and sandboxed fully on Dell deskside hardware.
Higher Education AI Lab	Repeatable design for university AI teaching labs — shared serving tiers, per-course provisioning, and a curriculum built on the other solutions.

Where to look

• Docs site (GitHub Pages): the docs/ folder is the site source, served directly from /docs: solution pages, sizing, inference stack, recipes and reference material.
• Solutions: solutions/ — every solution template (README + IT-Admin guide + developer guide).
• Scope: docs/scope.md
• Operations: docs/operations.md (ports, co-tenancy, resources, prerequisites, air-gap) · docs/topology.md (TP vs DP per workload) · monitoring/ (tokenomics dashboard).
• Deployment recipes: recipes/ — self-contained docker serve scripts per platform/tier.

Getting started (Agentic Coding)

Serve the default coding model on a Dell deskside system powered by NVIDIA GB10:

recipes/gb10/vllm/docker-nemotron-3.5-lightning-nvfp4.sh --port 8000
# blocks until the endpoint is truly ready (first run downloads ~21 GB of weights), starts the
# tokenomics dashboard, then prints:  READY {"base_url":"http://localhost:8000/v1", ...}
curl -s http://localhost:8000/v1/models
recipes/gb10/vllm/health-nemotron-3.5-lightning-nvfp4.sh   # field-debug health report any time

Prerequisites: Docker + NVIDIA Container Toolkit, ~25 GB free disk, outbound HTTPS to Hugging Face (or a pre-seeded cache). Add --dry-run to preview without changing anything. Other platforms: see the recipe catalog.

Then point a CLI/IDE agent at http://<host>:8000/v1 with model nvidia/nemotron-3.5-lightning (reasoning model — use max_tokens >= 512; generate window-aligned client configs with recipes/lib/client-config.sh). See the IT-Admin guide and IDE integration for the full setup.

License

Released under the Apache License 2.0 — see LICENSE.