EuLLM Pro: your company's AI, on your own servers, with an account for every person.
The same open-source engine, plus what you need to run it across a whole organization: accounts, control, a usage log and monitoring.
The questions it answers
Personal account for everyone
Learn more
Do several people need to use it, each with their own account?
Each person signs in with their own name and password. There are three roles: administrator, auditor (reads the log and statistics but changes nothing) and user. The session closes after 8 hours of inactivity, and after 5 failed attempts the account locks for 15 minutes.
Self-service people management
Learn more
Want to manage people from the browser, without knowing their passwords?
The administrator adds a person from the "People" page and sends them a link: the person chooses their own password, and the link only works once. From there the administrator changes roles, deactivates or removes an account, and sees who is connected, from where and with which browser; a session can be closed with one click. If someone forgets their password, they get a new link and the old one stops working immediately.
Admin-only model control
Learn more
Want only the administrator to choose the models?
Only the administrator downloads, deletes or changes models; everyone else uses what the administrator has made available. No one can turn off the model the whole office is using.
Full usage log
Learn more
Do you need to know who used what, and when?
For every request, the log records who made it, when, with which model, how long it took and how it ended. The person's name also appears in the engine's technical log, built for the traceability the AI Act requires.
Per-person performance
Learn more
Want to see if it's slow, and for whom?
For every person you can see time to first token, speed, the median and the worst cases, over the last 24 hours, 7 or 30 days. That makes it possible to tell whether the problem is the server, the model, or the type of requests.
24/7 monitoring
Learn more
Do you need to monitor the server around the clock?
The monitor records everything minute by minute, even when no one is watching: load, memory, GPU temperature and power draw, speed, requests. It keeps history for 30 days (up to a year, if the administrator decides so), with peaks and the time they happened. If you already use Prometheus or Grafana, it connects there with a token.
Engine guardrails
Learn more
Do you need guardrails on the engine?
A list of allowed models, operations reserved for the administrator, API keys with a per-minute request limit, allowed network addresses, and the sites the web tool is allowed to browse.
Data stays in-house
Learn more
Does data need to stay inside the company?
Everything runs on your own servers, with no telemetry. Even the license is verified locally, so it works on a closed network, with no Internet access.
Docker install in minutes
Learn more
Want to install it in a few minutes?
Ready-made Docker images are available, for CPU and for NVIDIA GPUs, with a sample file that starts them. Accounts, the log and the license live in a separate volume that's easy to back up.
Ollama & OpenAI-compatible API
Learn more
Do your own programs need to use it?
The API is compatible with Ollama and OpenAI, with dedicated keys for programs. It works with RAG Enterprise.
Graceful license expiry
Learn more
If the license expires, does everything stop?
No: accounts, chats and the log keep working. Only new users are blocked, until the new license arrives.
Single sign-on (SSO)
Learn more
Want to sign in with the company account?
Single sign-on (SSO) with Microsoft Entra ID, Google Workspace or Keycloak.
Per-department groups
Learn more
Want different rules per department?
Groups of people, with different models and limits, and a company system prompt that can't be changed.
Cross-device chat history
Learn more
Want to find your conversations on every computer?
Chat history saved on the server, with its own retention rules.
Multi-GPU load balancing
Learn more
Do you have more than one GPU server?
A single address in front of multiple servers, with the load balanced between them.
Open and Pro compared
| Feature | EuLLM (open, free) | EuLLM Pro |
|---|---|---|
| Engine, chat, Ollama- and OpenAI-compatible API | β | β |
| 24/7 monitoring with history and Prometheus | β | β |
| Models managed by the administrator only | β | β |
| Personal accounts and roles | β | β |
| Browser-based people management, with invite links | β | β |
| Usage log (who used what) | β | β |
| Per-person performance | β | β |
| Ready-made Docker images | β | β |
| Company-account sign-in (SSO), groups, multiple servers | β | Coming soon |
How it installs
Install EuLLM Pro on your server, with or without Docker.
Create the first administrator account.
The administrator invites people, who sign in from the browser with their name and a password they choose themselves.
No data leaves the company, and the license is verified locally β no Internet connection required.
Privacy and employees
Logging what employees write is a form of remote monitoring of workers. That's why EuLLM Pro, by default, never logs the text of questions and answers: only who, when, which model, and how long it took. No password is ever known to the administrator: everyone chooses their own.
This is not a substitute for advice from your legal counsel or DPO.
Frequently asked questions
- Is this the same as EuLLM?
- It's the same engine, plus the features your organization needs. If you already use EuLLM, nothing changes in how you use it.
- Where does it run?
- On your own Linux servers, with Docker or directly, with or without an NVIDIA GPU.
- Can I try it?
- Yes: without a license it works for up to 5 users.
- How much does it cost?
- Contact us.
