From 97b35f64203727331dc095664934070eef9e827d Mon Sep 17 00:00:00 2001
From: Pascal Bach <pascal.bach@nextrem.ch>
Date: Thu, 5 Feb 2026 20:22:55 +0100
Subject: [PATCH 1/2] nixos/llama-cpp: add state and cache directory

This is required to allow automatic download from huggingface etc.

They are placed in LLAMA_CACHE which is now the cache directory.
---
 nixos/modules/services/misc/llama-cpp.nix | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/nixos/modules/services/misc/llama-cpp.nix b/nixos/modules/services/misc/llama-cpp.nix
index 65ed2880b0b2..df9538d5137c 100644
--- a/nixos/modules/services/misc/llama-cpp.nix
+++ b/nixos/modules/services/misc/llama-cpp.nix
@@ -77,6 +77,10 @@ in
       serviceConfig = {
         Type = "idle";
         KillSignal = "SIGINT";
+        StateDirectory = "llama-cpp";
+        CacheDirectory = "llama-cpp";
+        WorkingDirectory = "/var/lib/llama-cpp";
+        Environment = [ "LLAMA_CACHE=/var/cache/llama-cpp" ];
         ExecStart =
           let
             args = [

From 676d8fc1bc1ff02f35392ba26577236d06c68838 Mon Sep 17 00:00:00 2001
From: Pascal Bach <pascal.bach@nextrem.ch>
Date: Thu, 5 Feb 2026 20:23:11 +0100
Subject: [PATCH 2/2] nixos/llama-cpp: support models-preset

The new options allows specifying --models-preset via Nix.
---
 nixos/modules/services/misc/llama-cpp.nix | 36 +++++++++++++++++++++++
 1 file changed, 36 insertions(+)

diff --git a/nixos/modules/services/misc/llama-cpp.nix b/nixos/modules/services/misc/llama-cpp.nix
index df9538d5137c..fdf306ab5256 100644
--- a/nixos/modules/services/misc/llama-cpp.nix
+++ b/nixos/modules/services/misc/llama-cpp.nix
@@ -8,6 +8,12 @@
 
 let
   cfg = config.services.llama-cpp;
+
+  modelsPresetFile =
+    if cfg.modelsPreset != null then
+      pkgs.writeText "llama-models.ini" (lib.generators.toINI { } cfg.modelsPreset)
+    else
+      null;
 in
 {
 
@@ -32,6 +38,32 @@ in
         default = null;
       };
 
+      modelsPreset = lib.mkOption {
+        type = lib.types.nullOr (lib.types.attrsOf lib.types.attrs);
+        default = null;
+        description = ''
+          Models preset configuration as a Nix attribute set.
+          This is converted to an INI file and passed to llama-server via --model-preset.
+          See llama-server documentation for available options.
+        '';
+        example = lib.literalExpression ''
+          {
+            "Qwen3-Coder-Next" = {
+              hf-repo = "unsloth/Qwen3-Coder-Next-GGUF";
+              hf-file = "Qwen3-Coder-Next-UD-Q4_K_XL.gguf";
+              alias = "unsloth/Qwen3-Coder-Next";
+              fit = "on";
+              seed = "3407";
+              temp = "1.0";
+              top-p = "0.95";
+              min-p = "0.01";
+              top-k = "40";
+              jinja = "on";
+            };
+          }
+        '';
+      };
+
       extraFlags = lib.mkOption {
         type = lib.types.listOf lib.types.str;
         description = "Extra flags passed to llama-cpp-server.";
@@ -97,6 +129,10 @@ in
               "--models-dir"
               cfg.modelsDir
             ]
+            ++ lib.optionals (cfg.modelsPreset != null) [
+              "--models-preset"
+              modelsPresetFile
+            ]
             ++ cfg.extraFlags;
           in
           "${cfg.package}/bin/llama-server ${utils.escapeSystemdExecArgs args}";