Ollama
Ollama to run local LLMs (optional model pull).
AI / ML
NVIDIA drivers + CUDA
NVIDIA drivers and CUDA toolkit. Destructive: kernel modules, reboot likely.
A recipe is not a frozen script: it is a verified playbook that drives an AI run. Reconnaissance first, idempotent by design, and every step proves itself before the next one starts.
- Category
- AI / ML
- Risk
- destructive
- Verified steps
- 2
- Estimated
- ~7 min
- Systems
- Ubuntu · Debian
01
Reconnaissance — before touching anything
The run starts read-only. Before a single change, Servor checks the real state of your server — a real machine is rarely clean, and only what is actually there decides what happens next.
- 01Is an NVIDIA GPU present?
- 02Driver already installed and working?
02
The steps — each one proves itself
The playbook below guides the run; the model adapts each command to the distribution and state found during reconnaissance. A step only counts as done when its verification passes.
- 01
Install the driversdestructive
Verified by
nvidia-smi >/dev/null 2>&1 && echo ok || echo reboot-needed - 02
Verify the GPU
Verified by
nvidia-smi -L | grep -c GPU
03
How Servor runs this recipe
Servor is zero-knowledge: the platform cannot execute anything on its own. When you launch this recipe, an AI run reads the reconnaissance, plans the steps, and asks for your approval. Each approved command is signed by your browser — with a key the server never sees — then relayed to the agent, which verifies the signature locally before executing.
- Read-only reconnaissance before any change
- Every command signed by your browser, approved by you
- Each step verified before the run moves on
More ai / ml recipes
Run this recipe on your server
Connect a server, launch the recipe, and approve each step as Servor executes and verifies it. Free plan, no card required.