SovereignSpark
Configure
No technical expertise required

You want private AI on your own device. We handle every technical barrier.

Setting up a local AI is hard: installation, network, security, model selection. You shouldn't have to deal with any of that. Tell us your needs in our 2-minute configurator. Our Hermes agents automatically build your perfect package. When your DGX Spark arrives, run your personalized script and you're ready in seconds.

# on the DGX Spark, in the unzipped folder

$ ./install.sh

▸ [0/7] Python environment…
▸ [1/7] Hardware check…
▸ [2/7] Network hardening…
▸ [3/7] Model selection & download…
▸ [4/7] NemoClaw + Hermes + Guardrails…
▸ [5/7] Capabilities…
▸ [6/7] Private chat interface…
▸ [7/7] Knowledge base (Hermes-guided)…
✓ Sovereign Spark is ready.

How it works

From configuration to a running Spark

01

Configure

Fill the 5-step configurator: business domain, use cases, guardrails. The AI can handle it in 3 questions.

↓
02

YAML profile generated

A spark-profile.yaml file encodes your setup: LLM models, enabled capabilities, guardrails, database, language.

↓
03

Ready-to-run package

You receive a ZIP by email and direct download: YAML + install scripts + all dependencies for every activated tool.

↓
04

Run on your DGX Spark

One command: ./install.sh --profile spark-profile.yaml. Everything installs automatically. Your Spark is ready.

📦

Package contents

spark-profile.yamlinstall.shdocker-compose.ymldeps/ (models, tools)README.md

Delivery

Email + direct download

Capabilities

What gets installed per capability

Select a capability to see what gets deployed on your DGX Spark.

Internal chatbot

Open WebUI — localhost:8080
Conversations
Budget Q3 2025
Contrat fournisseur
Note RH
Rapport audit
Model
gemma4:26b
128 GB · local
Quel est le plafond de dépenses Q3 ?
D'après le budget validé, le plafond Q3 est de 142 000 €, dont 38 k€ déjà engagés. Source : budget-2025.pdf p.4
Compare avec Q2 et donne la variation.
Q2 était de 128 k€ — soit +10,9 % en Q3. Tous les chiffres proviennent de vos documents indexés localement.
Quelles lignes sont à risque de dépassement ?
Lignes Voyages (97 %) et Prestataires (91 %). Recommandation : gel immédiat.
Posez une question…↵
  • Hermes agent configured for your domain
  • Open WebUI private interface
  • Personalized system prompt
  • NeMo Guardrails (PII, off-topic, audit)

What you get

What your team gets

A private chat interface at http://<device-ip>:8080
Answers grounded in your internal indexed documents
Business guardrails enforced on every message
A browsable 3D wiki of your knowledge base
SQL queries in plain language (optional)
Fine-tuning studio to improve the model on your own data (optional)

The .yaml carries your choices — domain, sector, capabilities, guardrail level. The installer reads it and builds everything autonomously.

Configure my AI