NixOS Module Options

The Nurvus NixOS module configures everything Demi needs: Ollama, Podman, sudo rules, and environment variables.

Options

Option	Default	Description
services.nurvus.enable	false	Enable nurvus and its dependencies
services.nurvus.model	auto-detect	Ollama model name (auto-selects by RAM if empty)
services.nurvus.llmUrl	"http://localhost:11434"	LLM server URL (Ollama, LM Studio, etc.)
services.nurvus.contextSize	8192	LLM context window size in tokens
services.nurvus.openFirewall	false	Open Ollama port for remote access
services.nurvus.maxCpuCores	null (no limit)	Cap Ollama to N CPU cores
services.nurvus.maxMemoryGB	null (no limit)	Cap Ollama memory usage in GB
services.nurvus.acceleration	null (CPU only)	GPU acceleration: "rocm" for AMD, "cuda" for NVIDIA
services.nurvus.daemon.enable	true	Enable daemon mode with web UI
services.nurvus.daemon.host	"0.0.0.0"	Address to bind the web server to
services.nurvus.daemon.port	3364	Port for the web UI (DEMI on a phone dial pad)
services.nurvus.daemon.openFirewall	true	Open the daemon port in the firewall

See Daemon Mode for detailed daemon configuration and setup.

Remote LLM

When llmUrl points to a non-localhost address, the module automatically skips installing and running the local Ollama service. This is useful when you want to run the LLM on a more powerful machine in your network, or use a different server like LM Studio.

# Remote Ollama
services.nurvus = {
  enable = true;
  model = "qwen3:30b-a3b";
  llmUrl = "http://my-server:11434";
  contextSize = 32768;
};

# LM Studio on a Mac (OpenAI-compatible API, auto-detected from port)
services.nurvus = {
  enable = true;
  model = "qwen/qwen3-30b-a3b";
  llmUrl = "http://192.168.64.1:1234";
  contextSize = 32768;
};

User Access

The module creates a nurvus system group. Add users to this group to let them run demi without root:

users.users.yourname = {
  isNormalUser = true;
  extraGroups = [ "nurvus" ];
};

Root users can always run demi without extra setup.