Cloud or on-premise:
we bring
to your data.
RIV4 runs as a cloud SaaS subscription or as a single-tenant dedicated installation on your own servers or private cloud. Same platform, same capabilities — you decide where your documents are processed and stored.
Pick where the platform lives
Both models run the same engine, the same document models and the same integrations. What changes is where the platform lives, who operates it and how capacity is licensed.
Model 01
Cloud
Managed by Velion, hosted in the EU or in another standard data zone. You subscribe and process — everything else is our job.
- Live in days, nothing to install
- You pay according to your consumption tier
- Updates arrive automatically
- Scales instantly with your volume
- Bring Your Own Model supported
Model 02
Dedicated installation
Runs entirely on your own infrastructure, on-premises or in your private cloud. One installation, one organisation.
- Documents never leave your perimeter
- Flat fee based on capacity, not volume
- Single-tenant, no shared resources
- Your brand, from capacity tier T2
- Run it yourself or let Velion manage it
Cloud vs on-premise,
at a glance
Neither model wins everywhere. The check marks the side with the structural advantage.
Structural advantage
Find your deployment model
in 2 steps
Two questions, no personal data required.
How many documents do you process per year?
A rough figure is enough.
What kind of documents, mostly?
When do those documents arrive?
Capacity is sized on your busiest minute, not on the yearly total.
Your recommended deployment model
—
Estimated peak load
0 PU / min
1 Processing Unit ≈ 1 page
Indicative capacity tier
—
—
Recommended
—
Use your own AI
and OCR subscription
Available in both deployment models: optical character recognition and AI inference run on a provider account that belongs to you, chosen together with Velion from the compatible list. RIV4 orchestrates — you own the engine.
- You pick the provider and the processing region, on Velion's recommendation
- You control retention and model-training terms directly
- Consumption is billed to you by the provider, with no margin in between
- Switch or add providers without changing the platform
- Native digital documents skip OCR entirely and cost less to process
Capacity tiers
for dedicated deployment
A dedicated installation is licensed on processing capacity, measured in Processing Units per minute at peak. One PDF page or one scanned page is approximately one Processing Unit.
T1
40 PU / min
Daily flow
- 8–12 vCPU
- 32 GB memory
- 500 GB SSD
- 10–50 Mbit/s outbound
T2
120 PU / min
Batches and peaks
- 12–16 vCPU
- 48–64 GB memory
- 1 TB NVMe SSD
- 50–200 Mbit/s outbound
T3
250 PU / min
Tight bursts
- 16–24 vCPU
- 64–96 GB memory
- 2 TB NVMe SSD
- 200+ Mbit/s outbound
T3+
250+ PU / min
Defined case by case
- Custom sizing
- Dedicated architecture review
- Multi-node options
- Let's talk about it
Inside a dedicated installation
The platform ships as containerised components on a single dedicated infrastructure. Only a few outbound connections leave it — and none of them carries your documents to Velion.
Your dedicated infrastructure
On-premises or in your private cloud · single tenant
Outbound
HTTPS only
External services
Reached outbound · no inbound access from Velion
Who manages the infrastructure
A dedicated installation can run on infrastructure you provide, or on infrastructure provided and fully managed by Velion. Switch between the two to see how responsibilities shift.
From assessment to go-live
A dedicated deployment is a project, not a signup. Here is what it looks like end to end.
Request information
Tell us about your documents and your constraints. We will come back with a sizing, the right deployment model and a concrete plan — no obligation.
FAQ
Everything you need to know about RIV4 deployment
Answers to the most common questions about cloud subscriptions, dedicated installations, licensing and data residency.
-
Yes. RIV4 is available as a dedicated single-tenant installation running on your own servers or in your private cloud. The platform is delivered as containerised components, and the infrastructure can be provided and managed by you or by Velion.
-
A dedicated installation is typically the right choice above roughly 250,000 documents per year, or whenever internal policy or regulation requires that documents never leave your perimeter — for example in defence, public administration, healthcare, banking and legal. Below that threshold, and without specific constraints, the cloud subscription is usually faster to deploy and simpler to operate.
-
No. Every document processing component — application, orchestrator, workers, databases, object store and vector index — runs on your infrastructure. Documents are neither transmitted to nor stored on Velion systems. The only data that reaches Velion is licence metadata: installation identifier, licence key identifier, software version, capacity level, timestamp, status and aggregate load figures.
-
Yes — a dedicated installation is a connected installation. It needs outbound HTTPS access to the licence verification service, the software image registry, the OCR and AI inference provider and an NTP server, plus a mail server if you use email intake or notifications. No inbound connection from Velion is required. Fully air-gapped installations are not supported today.
-
The installation confirms its licence every 24 hours, retrying every 6 hours after a failure. If no confirmation succeeds, it keeps operating normally for 21 days while showing a notice in the interface. Capacity limits are always enforced locally from the signed licence key, so a lost connection never changes your throughput. In no state does the software delete, encrypt or make your data inaccessible.
-
Bring Your Own Model means the OCR and AI inference runs on a subscription you own, with a provider selected together with Velion from the compatible list. You control the endpoints, the processing region and the provider's data-retention and model-training terms, and you pay consumption directly to that provider with no intermediary margin. It is available in both deployment models.
-
On dedicated installations white label is available from capacity tier T2 upward: your name, your colours, your visual identity. On the cloud platform white label is reserved for OEM partners.
-
Dedicated installations receive scheduled releases, typically three or four per year, plus security fixes as needed. Updates are installed in an agreed window, usually 60 to 90 minutes outside peak hours, with release notes and a documented rollback procedure. No unattended automatic updates are performed. Support covers the current and the previous version.
-
Yes. It is the same platform in both models, so document models, workflows and integrations carry over. Many organisations validate the process on the cloud subscription and migrate to a dedicated installation as volumes grow or compliance requirements tighten.
