ollama: use docker

hosts/architect/llm.nix

@@ -11,18 +11,18 @@ in
 {
   environment = {
     systemPackages = [ ollamaPkg pkgs.aichat ];
-    variables = {
-      OLLAMA_ORIGINS = "*";
-    };
+    # variables = {
+    #   OLLAMA_ORIGINS = "*";
+    # };
   };
 
-  services.ollama = {
-    inherit listenAddress;
+  # services.ollama = {
+  #   inherit listenAddress;
 
-    enable = true;
-    acceleration = "cuda";
-    package = ollamaPkg;
-  };
+  #   enable = true;
+  #   acceleration = "cuda";
+  #   package = ollamaPkg;
+  # };
 
   architect.vhost.${frontendDomain} = {
     dnsInterfaces = [ "tailscale" ];
@@ -49,13 +49,32 @@ in
       extraConfig = ''
         proxy_buffering off;
         proxy_read_timeout 600s;
-        proxy_set_header Host localhost:11434;
+        proxy_set_header Host localhost:${toString ollamaPort};
       '';
     };
   };
 
   virtualisation.oci-containers = {
     containers = {
+      ollama = {
+        image = "ollama/ollama:latest";
+        autoStart = true;
+        extraOptions = [
+          "--pull=always"
+          "--gpus=all"
+        ];
+        environment = {
+          OLLAMA_ORIGINS = "*";
+        };
+        volumes = [
+          "/ollama:/root/.ollama"
+        ];
+        ports = [
+          "${listenAddress}:${toString ollamaPort}"
+          "172.17.0.1:${toString ollamaPort}:${toString ollamaPort}"
+        ];
+      };
+      
       ollama-webui = {
         image = "ghcr.io/open-webui/open-webui:main";
         autoStart = true;