blob: 3ac3beb4de0780de5aed8d82f243924f250271ba (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
|
{ config, lib, pkgs, ... }:
let
inherit (lib) types;
cfg = config.services.ollama;
ollamaPackage = cfg.package.override {
inherit (cfg) acceleration;
linuxPackages = config.boot.kernelPackages // {
nvidia_x11 = config.hardware.nvidia.package;
};
};
in
{
options = {
services.ollama = {
enable = lib.mkEnableOption (
lib.mdDoc "Server for local large language models"
);
listenAddress = lib.mkOption {
type = types.str;
default = "127.0.0.1:11434";
description = lib.mdDoc ''
Specifies the bind address on which the ollama server HTTP interface listens.
'';
};
acceleration = lib.mkOption {
type = types.nullOr (types.enum [ "rocm" "cuda" ]);
default = null;
example = "rocm";
description = lib.mdDoc ''
Specifies the interface to use for hardware acceleration.
- `rocm`: supported by modern AMD GPUs
- `cuda`: supported by modern NVIDIA GPUs
'';
};
package = lib.mkPackageOption pkgs "ollama" { };
};
};
config = lib.mkIf cfg.enable {
systemd = {
services.ollama = {
wantedBy = [ "multi-user.target" ];
description = "Server for local large language models";
after = [ "network.target" ];
environment = {
HOME = "%S/ollama";
OLLAMA_MODELS = "%S/ollama/models";
OLLAMA_HOST = cfg.listenAddress;
};
serviceConfig = {
ExecStart = "${lib.getExe ollamaPackage} serve";
WorkingDirectory = "/var/lib/ollama";
StateDirectory = [ "ollama" ];
DynamicUser = true;
};
};
};
environment.systemPackages = [ ollamaPackage ];
};
meta.maintainers = with lib.maintainers; [ abysssol onny ];
}
|