Feat: Updated llama-cpp and switch mirai to llama-cpp

flake.lock

@@ -1790,6 +1790,23 @@
         "type": "github"
       }
     },
+    "llama-cpp": {
+      "flake": false,
+      "locked": {
+        "lastModified": 1754628423,
+        "narHash": "sha256-KZpYyDBCg/qZGpzCFILDdV8x7wgPetJsZp8i6o0u63E=",
+        "owner": "ggml-org",
+        "repo": "llama.cpp",
+        "rev": "aaa3d07ae749b781d6135eaff23c7fa8a4ab404a",
+        "type": "github"
+      },
+      "original": {
+        "owner": "ggml-org",
+        "ref": "b6116",
+        "repo": "llama.cpp",
+        "type": "github"
+      }
+    },
     "music-player": {
       "inputs": {
         "advisory-db": "advisory-db_4",
@@ -2621,6 +2638,7 @@
         "ik_llama": "ik_llama",
         "ironbar": "ironbar",
         "lanzaboote": "lanzaboote",
+        "llama-cpp": "llama-cpp",
         "music-player": "music-player",
         "musnix": "musnix",
         "navigator": "navigator",

flake.nix

@@ -192,6 +192,11 @@
       # submodules = true;
       flake = false;
     };
+    llama-cpp = {
+      # url = "https://github.com/ggml-org/llama.cpp";
+      url = "github:ggml-org/llama.cpp/b6116?submodules=1";
+      flake = false;
+    };
   };
 
   outputs = {

nixos/mirai/services/llama.nix

@@ -26,7 +26,7 @@
         url = "https://huggingface.co/lmstudio-community/gpt-oss-20b-GGUF/resolve/main/gpt-oss-20b-MXFP4.gguf";
         sha256 = "65d06d31a3977d553cb3af137b5c26b5f1e9297a6aaa29ae7caa98788cde53ab";
       };
-      package = pkgs.ik_llama;
+      # package = pkgs.ik_llama;
     };
     ollama = {
       enable = true;

overlays.nix

@@ -125,6 +125,10 @@
       src = inputs.ik_llama;
       version = "5995";
     });
+    llama-cpp = prev.llama-cpp.overrideAttrs (oldAttrs: {
+      src = inputs.llama-cpp;
+      version = "b6116";
+    });
     python312 = prev.python312.override {
       packageOverrides = final: prev: {
         pysaml2 = prev.pysaml2.overridePythonAttrs (orig: {