Sovereign AI
AI for GLPI, running on your hardware, under your control
Self-hosted, on-premises private AI so your organisation can run open-weight models inside your own network — with no data egress, no per-token bills, and no dependence on third-party cloud AI.
What is Sovereign AI?
Sovereign AI is a hardened, documented and supported reference deployment of Open WebUI and Ollama, with a curated commercial-use model library, packaged by Echo-9 for GLPI customers.
It exposes an OpenAI-compatible API that your GLPI AI plugins connect to — replacing cloud API keys with a local endpoint. Ticket data, chat and knowledge stay on your infrastructure.
This is the AI backend — not a GLPI plugin. Echo-9 plugins such as GilpAI, ProblemAI, and AI Closure and ITAMation consume Sovereign AI so assistance runs without sending content to public LLM clouds.
Why Organisations Choose It
Data sovereignty
No egress by default. Optional fully air-gapped mode for regulated and industrial environments.
Cost certainty
Open-weight models with no per-token API invoices — predictable cost on hardware you already own or buy once.
Compliance-ready
Built for GDPR, data residency and EU AI Act concerns common in public sector, healthcare, education and finance.
GLPI-first
OpenAI-compatible API so existing Echo-9 AI plugins drop in with a local base URL and model name.
No lock-in
Swap models without new cloud contracts. Open stack: Open WebUI, Ollama and licence-clean open-weight models.
Multi-user & RBAC
Web chat with roles, optional OIDC/LDAP, invite-only hardening, audit and retention controls.
What You Get
Open WebUI
Self-hosted chat, assistants, RAG, multi-user admin and document collections for grounded answers.
Ollama runtime
Local model server with OpenAI-compatible /v1 API for chat and embeddings — CPU or GPU.
Curated models
Licence-clean catalog (Qwen, Llama, Mistral, Phi, Gemma and embeddings) sized for CPU through large GPU tiers.
Deployment kit
Automation, env templates, air-gap prep guidance, reverse-proxy TLS patterns and ops handover.
GLPI integration guide
Plugin-by-plugin configuration for base URL, model names and API-key auth — no cloud keys required.
Support
Implementation support, patching guidance and ongoing helpdesk escalation aligned to your GLPI estate.
How It Fits Together
Reference Sizing
Typical starting points — we size precisely during discovery.
Small
Up to ~15 users. 8 vCPU, 32 GB RAM, 8 GB GPU or CPU-only. Ideal for pilots and small service desks.
Medium
Up to ~60 users. 16 vCPU, 64 GB RAM, 24 GB GPU. Strong fit for most mid-size GLPI estates.
Large
Up to ~300 users. 32 vCPU, 128 GB RAM, 48 GB+ GPU. Higher concurrency and larger models.
Air-gap
Same tiers with offline install packs and extra storage for model and image reserves.
Works With Echo-9 AI Plugins
GilpAI & LiveTalk
Portal AI and live chat — point the model endpoint at Sovereign AI for on-prem assistance.
ProblemAI
Clustering and RCA drafts with PII redaction — keep inference inside your network.
AI Closure and ITAMation
Constrained close-code suggestions with PII redaction — taxonomy still enforces when AI is offline.
Bring AI to GLPI Without the Cloud Risk
Tell us about your GLPI estate, user volumes and whether you need air-gap. We will recommend a sizing and delivery plan.