Regulated sectors
Healthcare, defence, finance, energy and public bodies bound by sector-specific rules.
Elium On-Premise
Your knowledge. Your infrastructure. Your rules.
By default, Elium is a sovereign European SaaS. On-premise is optional: the same platform, inside your own infrastructure.
- Full data sovereigntynothing leaves your perimeter
- AI on your own modelcalled directly from your infrastructure
- Source-code accessfront and back, auditable
Built and run in Belgium · Independently owned for over a decade · ISO 27001 · GDPR
Who it’s for
Built for organisations that cannot compromise on sovereignty.
Regulated and sovereignty-sensitive organisations that have outgrown what a SaaS deployment can offer them, whatever the hosting country. More than 1,200 multinational teams run on Elium, including public bodies and regulated enterprises.
Reinforced obligations
Reinforced GDPR duties or strict internal security policies you cannot bend.
Real independence
Genuine independence from cloud providers, not a contractual promise of it.
Managed-platform builders
IT services firms building a managed knowledge platform for their own clients.
Not everyone needs this. On-premise adds operational weight. It earns its place when sovereignty, auditability or regulation make it non-negotiable. If European hosting already meets your rules, our sovereign European SaaS is the lighter path. Explore the platform.
What you get
The full platform, inside your perimeter.
Elium On-Premise is the complete trusted knowledge platform, not a reduced edition. Your knowledge stays verified, current and governed.
Data sovereignty
Your content never leaves your perimeter.
Content, metadata and processing flows all stay on your infrastructure. Nothing transits external services, and you control where your data physically sits and which network flows exist around it.
- Content, metadata and processing stay on your infrastructure
- You control where data sits and which flows exist
- Nothing transits external services
AI on your terms
The AI runs on the model you choose.
This is where most sovereignty claims fall apart. Our AI features, from the Smart Assistant to semantic search and automatic summaries, run against the model you pick, called directly from your infrastructure.
- Azure OpenAI, Mistral, Claude, or any OpenAI-compatible API
- Cloud API keys, or a model on your own GPUs (vLLM, Ollama)
- Calls runfrom your infrastructure. No data reaches Elium.
Compliance
It fits the framework you already run.
Reinforced GDPR requirements, sector-specific standards, internal security policies: Elium On-Premise works with the compliance architecture you have, rather than asking you to change it.
- Reinforced GDPR and sector-specific standards
- Workswith your compliance architecture
- Customer-managed encryption keys
Transparency
Read the source. Audit everything.
You get the platform source code through a private GitHub repository, front end and back end. Your technical teams and your auditors can review it, run their own security analysis and validate compliance against your standards.
- Front end and back end, in a private GitHub repository
- Your auditors run their own security analysis
- Validate compliance against your standards
Architecture at a glance
A containerised architecture your teams already know.
Elium On-Premise is a modern, containerised stack built on standards your teams use every day.
| Container orchestration | Kubernetes or equivalent |
|---|---|
| Database | PostgreSQL |
| Search | Vector and full-text index, shipped with the platform |
| Object storage | S3-compatible (MinIO, Ceph, or any compatible system) |
| Reverse proxy and TLS | nginx or equivalent |
| Cache and message queue | Redis or equivalent |
| File scanning | Antivirus (self-hosted or cloud equivalent) |
Managed services welcome. Database, cache, search, object storage and antivirus can each run as managed cloud equivalents. The Elium application services stay containerised.
Sizing. Indicative footprints, high-availability patterns and bandwidth needs are shared during the feasibility phase and matched to your volumes.
Who does what
Clear ownership from day one.
On-premise only works when the responsibility model is unambiguous. Whoever runs it, an integration partner or your own IT team, keeps full autonomy over their methods; what they commit to is the outcome.
Your integration partner or IT team
- Infrastructure provisioning and sizing
- Security, access management, backups and disaster recovery
- Deployment, updates, monitoring and availability
- Full technical support, from L1 to L3
Elium
- Platform source code
- Complete operational documentation and deployment guides
- Software releases as Docker images and Helm charts
- Release notes with every version
Your teams
- Access policies
- Content governance
- Functional administration of the platform
One thing we never delegate: critical security patches reach every on-premise customer, with no conditions attached.
An optional publisher support package covers analysis and correction of confirmed bugs in the Elium source code.
How we get there
Three phases, one clear plan.
Feasibility, together
Elium, your teams and whoever will operate it check the project holds up before anyone commits: technical feasibility, security and compliance, target architecture, and support arrangements.
Implementation
Your integration partner or IT team provisions the infrastructure, deploys from the images and manifests we provide, and configures storage, TLS, your model API and single sign-on. Our team can validate the target architecture in an optional, time-boxed review.
Operation, in your hands
Monitoring, support and updates follow the cadence you agree with whoever operates it, inside your own maintenance windows. You keep control of the schedule and of every change to production.
FAQ
Questions we hearabout on-premise
Yes. Your own data centre, a private cloud, or dedicated infrastructure operated by your integration partner or IT team.
Yes. Front end and back end, through a private GitHub repository, along with the operational documentation.
No. That includes AI processing, which runs against the model you configure, called directly from your infrastructure.
Yes. Any model with an OpenAI-compatible API works, including one you host yourself on your own GPUs.
The full platform is there, but a few features that depend on public cloud services change. Integrations that rely on external webhooks, such as Slack and Microsoft Teams, need adaptation. Automatic video transcription is not available, because it relies on third-party APIs. Connectors that need a publicly reachable URL work only where your network allows one. These rarely matter on-premise, where the whole point is autonomy from public cloud. Your integration partner or IT team can build the specific connectors you need, and a detailed compatibility matrix is part of the feasibility phase.
Your integration partner, or your own IT team if you run it in-house, across infrastructure, application and migrations. We back them with source code, documentation, releases and, optionally, publisher-level support on confirmed bugs.
Not always. If your rules are met by European hosting, our SaaS, with a SecNumCloud-qualified hosting option, is simpler to run. On-premise is for cases where sovereignty, auditability or regulation make hosting inside your own perimeter non-negotiable.
Let’s find out if on-premise is right for you.
See it on your own knowledge.
A feasibility call with our team is the fastest way to know. No commitment, and a clear answer at the end of it.