ALLSmartLLM on your own hardware
You want the full stack running in-house — GDPR on your network, data never leaves the premises. We deliver setup and operations.
What we set up
The same stack we run under allsmartllm.com — on your server, in your network.
- Gateway (routing, priority lanes, usage tracking)
- Portal + Zentrale (customer management, keys, BYOK vault)
- Privacy sidecar (GLiNER detector, LLM check)
- Metrics collector + hardware dashboard
What you bring
Hardware and a network context. Models run on your GPU infrastructure (vLLM/Ollama) — we integrate them.
- One or more GPU servers (DGX Spark, RTX 4090+, or similar)
- Public VPS for the domain (can be an existing one)
- SMTP access for registration mails
How it goes
Four steps, usually 2-3 weeks calendar time. We set up every instance personally.
- Kick-off: model catalogue, billing flow, network setup
- Installation: gateway, proxy, sidecar on your hardware
- Cutover: customer keys migrated, DNS switched
- Handover + optional maintenance contract
What stays with you
Everything. Prompts, responses, customer data, invoices — all on your hardware. We don't see any of it. Updates ship as signed bundles you can install yourself.
Sound like your path?
A 30-minute kickoff call — we'll figure out whether the scenario fits and what a realistic effort would look like.
Send enquiryFrequently asked questions
Is this a self-service wizard?
No. Each instance is set up personally: kick-off, install, cutover. Self-service rollout is in the works.
What does the setup cost?
Usually 2-3 person-days plus kick-off consulting. Concrete number after the first call — depends on network setup and model catalogue.
Do I get the source code?
Yes. Gateway and portal run on your hardware, you have access. License model and update policy discussed during setup.
Can I run both, with you and on my own hardware?
Yes, the recommended transition path. Start with us via allsmartllm.com, finish development, then switch a URL in the app once your instance is live.