From a4bf2a0ea9cc64971b1759de80eeb0fe05d5458d Mon Sep 17 00:00:00 2001
From: Simon Gardling <titaniumtown@proton.me>
Date: Mon, 6 Oct 2025 01:42:37 -0400
Subject: [PATCH] llama.cpp: testing

---
 services/llama-cpp.nix | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/services/llama-cpp.nix b/services/llama-cpp.nix
index 208e396..dc7a368 100644
--- a/services/llama-cpp.nix
+++ b/services/llama-cpp.nix
@@ -11,8 +11,8 @@
     enable = true;
     model = builtins.toString (
       pkgs.fetchurl {
-        url = "https://huggingface.co/rodrigomt/Qwen3-30B-A3B-Thinking-Deepseek-Distill-2507-v3.1-V2-GGUF/resolve/main/Qwen3-30B-A3B-Thinking-Deepseek-Distill-2507-v3.1-V2-UD-Q4_K_XL.gguf";
-        sha256 = "1a3abffc8463041e24cdc43af26c99b6cfab1d2ee78fef0d793033ec0e5b58aa";
+        url = "https://huggingface.co/ggml-org/gpt-oss-20b-GGUF/resolve/main/gpt-oss-20b-mxfp4.gguf";
+        sha256 = "be37a636aca0fc1aae0d32325f82f6b4d21495f06823b5fbc1898ae0303e9935";
       }
     );
     port = service_configs.ports.llama_cpp;
@@ -20,14 +20,16 @@
     # vulkan broken: https://github.com/ggml-org/llama.cpp/issues/13801
     package = (
       lib.optimizePackage (
-        inputs.llamacpp.packages.${pkgs.system}.default.overrideAttrs (old: {
+        inputs.llamacpp.packages.${pkgs.system}.vulkan.overrideAttrs (old: {
           postPatch = "";
         })
       )
     );
     extraFlags = [
-      # "-ngl"
-      # "9999"
+      "-ngl"
+      "5"
+      "-c"
+      "16384"
     ];
   };