@@ -118,11 +118,11 @@ }
             }
           '';
         };
-        "swap.localhost" = {
+        "chat.localhost" = {
           logFormat = "output discard";
           extraConfig = ''
             import local_tls
-            reverse_proxy http://localhost:9292 {
+            reverse_proxy http://localhost:9090 {
               transport http {
                 dial_timeout 1s
                 compression off

@@ -5,12 +5,10 @@ lib,
   ...
 }:
 let
-  inherit (lib) removePrefix concatMapStringsSep isDerivation;
+  inherit (lib) removePrefix;
 
   nixfiles = config.home.homeDirectory + "/projects/alin.ovh/nixfiles";
   direct_symlink = path: config.lib.file.mkOutOfStoreSymlink "${nixfiles}/${removePrefix "/" path}";
-
-  makeDrvBinPath = ps: concatMapStringsSep ":" (p: if isDerivation p then "${p}/bin" else p) ps;
 in
 {
   imports = [
@@ -89,33 +87,27 @@ OLLAMA_CONTEXT_LENGTH = "131072";
         };
       };
     };
-    llama-swap = {
+    llama-server = {
       enable = true;
       config = {
         ProgramArguments = [
-          "${pkgs.llama-swap}/bin/llama-swap"
-          "--listen"
-          "0.0.0.0:9292"
-          "--config"
-          "${config.xdg.configHome}/llama-swap/config.yaml"
-          "--watch-config"
+          "${pkgs.llama-cpp}/bin/llama-server"
+          "--port"
+          "9090"
+          "--models-preset"
+          "${config.xdg.configHome}/llama/presets.ini"
         ];
         RunAtLoad = true;
         KeepAlive = true;
         WorkingDirectory = "/Users/alan";
-        StandardOutPath = "/Users/alan/Library/Logs/llama-swap.log";
-        StandardErrorPath = "/Users/alan/Library/Logs/llama-swap.log";
-        EnvironmentVariables = {
-          PATH = makeDrvBinPath [
-            pkgs.llama-cpp
-          ];
-        };
+        StandardOutPath = "/Users/alan/Library/Logs/llama-server.log";
+        StandardErrorPath = "/Users/alan/Library/Logs/llama-server.log";
       };
     };
   };
 
-  xdg.configFile."llama-swap/config.yaml" = {
-    source = direct_symlink "/user/settings/llama/swap.yaml";
+  xdg.configFile."llama/presets.ini" = {
+    source = direct_symlink "/user/settings/llama/presets.ini";
   };
 
   xdg.configFile."raycast/ai/providers.yaml" = {

@@ -0,0 +1,18 @@
+version = 1
+
+[ggml-org/gpt-oss-20b-GGUF]
+jinja = on
+
+[ggml-org/gpt-oss-120b-GGUF]
+jinja = on
+
+[henry1477/Qwen3-Embedding-4B-gguf]
+embedding = on
+pooling = cls
+ubatch-size = 8192
+
+[ggml-org/Ministral-3-14B-Reasoning-2512-GGUF]
+
+[ggml-org/Ministral-3-14B-Instruct-2512-GGUF]
+
+[ggml-org/Nemotron-Nano-3-30B-A3B-GGUF]

@@ -1,10 +0,0 @@
-# yaml-language-server: $schema=https://raw.githubusercontent.com/mostlygeek/llama-swap/refs/heads/main/config-schema.json
-models:
-  gpt-oss:20b:
-    cmd: llama-server --port ${PORT} -hf ggml-org/gpt-oss-20b-GGUF --jinja -ub 2048 -b 2048
-  gpt-oss:120b:
-    cmd: llama-server --port ${PORT} -hf ggml-org/gpt-oss-120b-GGUF --jinja -ub 2048 -b 2048
-hooks:
-  on_startup:
-    preload:
-      - "gpt-oss:120b"

@@ -1,7 +1,7 @@
 providers:
   - id: llama.cpp
     name: llama.cpp
-    base_url: https://swap.localhost/v1/
+    base_url: https://chat.localhost/v1/
     models:
       - id: gpt-oss:20b
         name: gpt-oss-20b