SUSE AI 1.0
- WHAT?
This document provides a comprehensive, step-by-step guide for the SUSE AI air-gapped deployment.
- WHY?
To help users successfully complete the air-gapped deployment process.
- EFFORT
Less than one hour of reading and an advanced knowledge of Linux deployment.
- GOAL
To learn enough information to deploy SUSE AI in both testing and production air-gapped environments.
Publication Date: 2026-05-13
- 1 Air-gapped environments
- 2 Installation overview
- 3 Installing the Linux and Kubernetes distribution
- 4 Preparing the cluster for AI Library
- 4.1 Installing SUSE Rancher Prime on a Kubernetes cluster in air-gapped environments
- 4.2 Installing the NVIDIA GPU Operator on the SUSE Rancher Prime: RKE2 cluster
- 4.3 Registering existing clusters
- 4.4 Assigning GPU nodes to applications
- 4.5 Installing SUSE Security
- 4.6 Setting up SUSE Observability for SUSE AI
- 5 Installing applications from AI Library
- 5.1 What is SUSE Application Collection?
- 5.2 What is SUSE Registry?
- 5.3 Installation procedure
- 5.4 Installing cert-manager
- 5.5 Installing OpenSearch
- 5.6 Installing Milvus
- 5.7 Installing Ollama
- 5.8 Installing Open WebUI
- 5.9 Installing vLLM
- 5.10 Installing mcpo
- 5.11 Installing PyTorch
- 5.12 Installing Qdrant
- 5.13 Installing LiteLLM
- 5.14 Installing MLflow
- Glossary
- A Copyright
- B GNU Free Documentation License
- B1 0. PREAMBLE
- B2 1. APPLICABILITY AND DEFINITIONS
- B3 2. VERBATIM COPYING
- B4 3. COPYING IN QUANTITY
- B5 4. MODIFICATIONS
- B6 5. COMBINING DOCUMENTS
- B7 6. COLLECTIONS OF DOCUMENTS
- B8 7. AGGREGATION WITH INDEPENDENT WORKS
- B9 8. TRANSLATION
- B10 9. TERMINATION
- B11 1. FUTURE REVISIONS OF THIS LICENSE
- B12 ADDENDUM: How to use this License for your documents
List of Figures
- 1.1 General schema of an air-gapped environment
- 2.1 SUSE AI air-gapped installation process
- 3.1 Language, keyboard and product selection
- 3.2 License agreement
- 3.3 Network settings
- 3.4 Registration
- 3.5 Extension and module selection
- 3.6 Add-on product
- 3.7 System role
- 3.8 Suggested partitioning
- 3.9 Clock and time zone
- 3.10 Local user creation
- 3.11 Password for the system administrator
root - 3.12 Installation settings
- 3.13 Confirm installation
- 3.14 Performing the installation
- 4.1 High-level overview of the SUSE Observability setup
- 4.2 Separate clusters for SUSE AI and SUSE Observability
- 4.3 New GenAI Observability menu item
- 5.1 Milvus page in the SUSE Application Collection
- 5.2 Adding a vLLM connection to Open WebUI
- 5.3 MLflow Web UI
List of Tables
List of Examples
- 4.1 Click to expand
- 5.1 Basic override file with GPU and two models pulled at startup
- 5.2 Basic override file with Ingress and no GPU
- 5.3 Open WebUI override file with Ollama included
- 5.4 Open WebUI override file with Ollama installed separately
- 5.5 Open WebUI override file with pipelines enabled
- 5.6 Open WebUI override file with a connection to vLLM
- 5.7 Stand-alone deployment of open-webui-pipelines
- 5.8 Minimal configuration
- 5.9 Validating the installation
- 5.10 Basic configuration
- 5.11 Loading prefetched models from persistent storage
- 5.12 Configuration with multiple models
- 5.13 CPU offloading
- 5.14 Shared remote KV cache storage with LMCache
- 5.15 Basic override file with GPU enabled
- 5.16 ConfigMap-based upload
- 5.17 Host-folder with files baked into the chart
- 5.18 Git repository clone: public with no authentication
- 5.19 Git repository clone: private with authentication
- 5.20 Basic override file when the cluster has no default storage class set.
- 5.21 An example where Qdrant uses GPU capabilities.
- 5.22 Basic override file with PostgreSQL deployment and master key automatically generated.