SUSE AI

What is SUSE AI?

SUSE AI is an enterprise platform for deploying Generative AI - allows you to run large language models (LLM) on-premises, without sending data to external clouds. Based on open source, integrated with SUSE ecosystem (Rancher, SLES).

Key Functions:

• On-Premises LLM - Llama, Mistral, other models locally
• RAG - Retrieval Augmented Generation with your own data
• Kubernetes Native - deployment on Rancher
• Air-Gap Ready - for regulated industries

What problem does it solve?

flowchart LR
    subgraph Cloud AI Problem
        A[Company data] --> B[OpenAI/Azure]
        B --> C[Data in cloud]
        C --> D[Compliance risk]
        D --> E[Vendor lock-in]
    end

    subgraph SUSE AI
        F[Company data] --> G[SUSE AI On-Prem]
        G --> H[Data stays local]
        H --> I[Compliance OK]
        I --> J[Open source]
    end

    style A fill:#6366f1,stroke:#4f46e5,color:#fff
    style B fill:#dc2626,stroke:#b91c1c,color:#fff
    style C fill:#dc2626,stroke:#b91c1c,color:#fff
    style D fill:#dc2626,stroke:#b91c1c,color:#fff
    style E fill:#dc2626,stroke:#b91c1c,color:#fff
    style F fill:#6366f1,stroke:#4f46e5,color:#fff
    style G fill:#22c55e,stroke:#16a34a,color:#fff
    style H fill:#22c55e,stroke:#16a34a,color:#fff
    style I fill:#22c55e,stroke:#16a34a,color:#fff
    style J fill:#22c55e,stroke:#16a34a,color:#fff

Common Problems:

• Company data sent to OpenAI/Azure AI = compliance risk
• Regulations (banking, healthcare) prohibit cloud AI
• Vendor lock-in to OpenAI/Azure
• No control over model and training data
• API call costs in cloud

SUSE AI Architecture

flowchart TD
    A[Users/Apps] --> B[SUSE AI API]
    B --> C[Inference Engine]
    C --> D[LLM Models]
    D --> E[Llama 3.x]
    D --> F[Mistral]
    D --> G[Custom Models]

    B --> H[RAG Pipeline]
    H --> I[Vector DB]
    H --> J[Enterprise Data]

    K[Rancher] --> C
    L[NVIDIA GPU] --> C

    style A fill:#6366f1,stroke:#4f46e5,color:#fff
    style B fill:#f59e0b,stroke:#d97706,color:#fff
    style C fill:#8b5cf6,stroke:#7c3aed,color:#fff
    style D fill:#22c55e,stroke:#16a34a,color:#fff
    style H fill:#22c55e,stroke:#16a34a,color:#fff
    style K fill:#f59e0b,stroke:#d97706,color:#fff
    style L fill:#22c55e,stroke:#16a34a,color:#fff

Key Features

Use Cases

SUSE AI vs Cloud AI

Aspect	SUSE AI	OpenAI/Azure AI
Data location	On-premises	Cloud
Compliance	Air-gap ready	Depends
Cost model	Infrastructure	Per token
Vendor lock-in	Open source	Proprietary
Customization	Full control	Limited
Latency	Low (local)	Network dependent
Models	Open source	OpenAI only
Scaling	Your infrastructure	Automatic

SUSE AI Advantage:

• Data stays on-premises
• Compliance for regulated industries
• Open source models - no vendor lock-in
• Predictable costs (infra vs per-token)
• Full customization and fine-tuning

Who is it for?

Requirements

Parameter	Minimum	Recommended
GPU	NVIDIA RTX 3090 / A10	NVIDIA A100 / H100
VRAM	24 GB	80 GB+
RAM	64 GB	256 GB+
Storage	500 GB SSD	2 TB NVMe
OS	SLES 15 SP5+ / Linux Micro	SLES 15 SP6
Kubernetes	K3s / RKE2 / Rancher	Rancher managed

Specification

Parameter	Value
Models	Llama 3.x, Mistral, Mixtral, custom
API	OpenAI-compatible REST API
Deployment	Kubernetes (Rancher, K3s, RKE2)
GPU	NVIDIA CUDA
Quantization	4-bit, 8-bit, FP16
RAG	Vector DB integration
Security	Air-gap, RBAC, audit

FAQ

What models does it support? Open source LLMs: Llama 3.x, Mistral, Mixtral, CodeLlama, and others. You can also use your own fine-tuned models.

Do I need GPU? Yes for production. NVIDIA GPUs with minimum 24GB VRAM. For dev/test can use CPU (slower).

Is the API compatible with OpenAI? Yes. SUSE AI offers OpenAI-compatible API - easy migration from cloud AI.

How does RAG work? You import documents, system creates vector embeddings, on question searches relevant fragments and adds to LLM context.

Can models be fine-tuned? Yes. Platform supports fine-tuning on your own data.

Does nFlo deploy SUSE AI? Yes. Deployments on Rancher/K8s, GPU configuration, RAG pipeline setup, application integration.