diff --git a/src-tauri/src/core/state.rs b/src-tauri/src/core/state.rs
index dab29aa85..b59aa0a3d 100644
--- a/src-tauri/src/core/state.rs
+++ b/src-tauri/src/core/state.rs
@@ -8,18 +8,18 @@ use tokio::task::JoinHandle;
 
 /// Server handle type for managing the proxy server lifecycle
 pub type ServerHandle = JoinHandle<Result<(), Box<dyn std::error::Error + Send + Sync>>>;
+use tokio::{process::Child, sync::Mutex};
 
 #[derive(Default)]
 pub struct AppState {
     pub app_token: Option<String>,
     pub mcp_servers: Arc<Mutex<HashMap<String, RunningService<RoleClient, ()>>>>,
     pub download_manager: Arc<Mutex<DownloadManagerState>>,
-    pub cortex_restart_count: Arc<Mutex<u32>>,
-    pub cortex_killed_intentionally: Arc<Mutex<bool>>,
     pub mcp_restart_counts: Arc<Mutex<HashMap<String, u32>>>,
     pub mcp_active_servers: Arc<Mutex<HashMap<String, serde_json::Value>>>,
     pub mcp_successfully_connected: Arc<Mutex<HashMap<String, bool>>>,
     pub server_handle: Arc<Mutex<Option<ServerHandle>>>,
+    pub llama_server_process: Arc<Mutex<Option<Child>>>,
 }
 pub fn generate_app_token() -> String {
     rand::thread_rng()
diff --git a/src-tauri/src/core/utils/extensions/inference_llamacpp_extension/server.rs b/src-tauri/src/core/utils/extensions/inference_llamacpp_extension/server.rs
index 4efa725cc..dc7b10757 100644
--- a/src-tauri/src/core/utils/extensions/inference_llamacpp_extension/server.rs
+++ b/src-tauri/src/core/utils/extensions/inference_llamacpp_extension/server.rs
@@ -1,8 +1,7 @@
 use std::path::PathBuf;
-use std::sync::Arc;
+use tauri::path::BaseDirectory;
 use tauri::{AppHandle, Manager, State}; // Import Manager trait
-use tokio::process::{Child, Command};
-use tokio::sync::Mutex;
+use tokio::process::Command;
 
 use crate::core::state::AppState;
 
@@ -23,6 +22,16 @@ pub enum ServerError {
     Tauri(#[from] tauri::Error),
 }
 
+// impl serialization for tauri
+impl serde::Serialize for ServerError {
+    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: serde::Serializer,
+    {
+        serializer.serialize_str(self.to_string().as_ref())
+    }
+}
+
 type ServerResult<T> = Result<T, ServerError>;
 
 // --- Helper function to find the server binary ---
@@ -38,26 +47,26 @@ fn get_server_path(app_handle: &AppHandle) -> ServerResult<PathBuf> {
 
     app_handle
         .path()
-        .resolve_resource(relative_path)
-        .map_err(|e| ServerError::ResourcePathError(e.to_string()))?
-        .ok_or_else(|| {
-            ServerError::BinaryNotFound(format!(
-                "Could not resolve resource path for '{}'",
-                if cfg!(windows) {
-                    "engines/llama-server.exe"
-                } else {
-                    "engines/llama-server"
-                } // TODO: ADJUST THIS PATH
-            ))
-        })
+        .resolve(relative_path, BaseDirectory::Resource)
+        .map_err(|e| ServerError::ResourcePathError(e.to_string()))
+    // .ok_or_else(|| {
+    //     ServerError::BinaryNotFound(format!(
+    //         "Could not resolve resource path for '{}'",
+    //         if cfg!(windows) {
+    //             "engines/llama-server.exe"
+    //         } else {
+    //             "engines/llama-server"
+    //         } // TODO: ADJUST THIS PATH
+    //     ))
+    // })
 }
 
 // --- Load Command ---
 #[tauri::command]
 pub async fn load(
-    app_handle: AppHandle,            // Get the AppHandle
-    state: State<'_, AppState>,       // Access the shared state
-    args: Vec<String>,                // Arguments from the frontend
+    app_handle: AppHandle,      // Get the AppHandle
+    state: State<'_, AppState>, // Access the shared state
+    args: Vec<String>,          // Arguments from the frontend
 ) -> ServerResult<()> {
     let mut process_lock = state.llama_server_process.lock().await;
 
@@ -71,8 +80,14 @@ pub async fn load(
     log::info!("Using arguments: {:?}", args);
 
     if !server_path.exists() {
-         log::error!("Server binary not found at expected path: {:?}", server_path);
-         return Err(ServerError::BinaryNotFound(format!("Binary not found at {:?}", server_path)));
+        log::error!(
+            "Server binary not found at expected path: {:?}",
+            server_path
+        );
+        return Err(ServerError::BinaryNotFound(format!(
+            "Binary not found at {:?}",
+            server_path
+        )));
     }
 
     // Configure the command to run the server
diff --git a/src-tauri/src/lib.rs b/src-tauri/src/lib.rs
index fa6ccaea4..4d46d19a2 100644
--- a/src-tauri/src/lib.rs
+++ b/src-tauri/src/lib.rs
@@ -88,19 +88,18 @@ pub fn run() {
             core::hardware::get_system_info,
             core::hardware::get_system_usage,
             // llama-cpp extension
-            core::utils::extensions::inference_llamacpp_extension::load, 
-            core::utils::extensions::inference_llamacpp_extension::unload
+            core::utils::extensions::inference_llamacpp_extension::server::load,
+            core::utils::extensions::inference_llamacpp_extension::server::unload,
         ])
         .manage(AppState {
             app_token: Some(generate_app_token()),
             mcp_servers: Arc::new(Mutex::new(HashMap::new())),
             download_manager: Arc::new(Mutex::new(DownloadManagerState::default())),
-            cortex_restart_count: Arc::new(Mutex::new(0)),
-            cortex_killed_intentionally: Arc::new(Mutex::new(false)),
             mcp_restart_counts: Arc::new(Mutex::new(HashMap::new())),
             mcp_active_servers: Arc::new(Mutex::new(HashMap::new())),
             mcp_successfully_connected: Arc::new(Mutex::new(HashMap::new())),
             server_handle: Arc::new(Mutex::new(None)),
+            llama_server_process: Arc::new(Mutex::new(None)),
         })
         .setup(|app| {
             app.handle().plugin(