Deploy open-webui

This is a very nice web ui frontend similar to ChatGPT that can use both OpenAI
and Ollama as backends at the same time. Currently I'm just using it locally but
eventually I think I'll expose it over the internet and use it as my go-to LLM
interface.

5 files changed, 20 insertions, 1 deletions

diff --git a/Omni/Cloud/Ports.nix b/Omni/Cloud/Ports.nix
index 5b8446c..9d5bc4c 100644
--- a/Omni/Cloud/Ports.nix
+++ b/Omni/Cloud/Ports.nix
@@ -31,6 +31,8 @@
   mpd-stream = 8097;
   murmur = 64738;
   nostr-relay = 8084;
+  ollama = 11434;
+  open-webui = 8088;
   radicale = 5232;
   sabten = 8081;
   ssh = 22;
diff --git a/Omni/Dev/Beryllium.nix b/Omni/Dev/Beryllium.nix
index 0327822..9a72353 100755
--- a/Omni/Dev/Beryllium.nix
+++ b/Omni/Dev/Beryllium.nix
@@ -10,6 +10,7 @@ bild.os {
     ./Beryllium/Ollama.nix
     ./Docker.nix
     ./Vpn.nix
+    ./Beryllium/OpenWebui.nix
   ];
   networking.hostName = "beryllium";
   networking.domain = "beryl.simatime.com";
diff --git a/Omni/Dev/Beryllium/Configuration.nix b/Omni/Dev/Beryllium/Configuration.nix
index a371649..4a792ef 100644
--- a/Omni/Dev/Beryllium/Configuration.nix
+++ b/Omni/Dev/Beryllium/Configuration.nix
@@ -83,6 +83,7 @@ in {
   hardware.nvidia.powerManagement.finegrained = false;
   hardware.nvidia.open = true;
   hardware.nvidia.nvidiaSettings = true;
+  hardware.nvidia-container-toolkit.enable = true;
 
   hardware.keyboard.zsa.enable = true;
 
diff --git a/Omni/Dev/Beryllium/Ollama.nix b/Omni/Dev/Beryllium/Ollama.nix
index 0018f49..3f2398e 100644
--- a/Omni/Dev/Beryllium/Ollama.nix
+++ b/Omni/Dev/Beryllium/Ollama.nix
@@ -10,6 +10,7 @@ If you want to spend time on it, spend time over there.
 */
 let
   pkg = pkgs.unstable.ollama;
+  ports = import ../../Cloud/Ports.nix;
 in {
   systemd.services.ollama = {
     description = "ollama";
@@ -17,7 +18,7 @@ in {
     wantedBy = ["multi-user.target"];
 
     environment = {
-      OLLAMA_HOST = "localhost:11434";
+      OLLAMA_HOST = "0.0.0.0:${toString ports.ollama}";
       # Where to store LLM model files.
       HOME = "%S/ollama";
       OLLAMA_MODELS = "%S/ollama/models";
diff --git a/Omni/Dev/Beryllium/OpenWebui.nix b/Omni/Dev/Beryllium/OpenWebui.nix
new file mode 100644
index 0000000..7b95331
--- /dev/null
+++ b/Omni/Dev/Beryllium/OpenWebui.nix
@@ -0,0 +1,14 @@
+{config, ...}: let
+  ports = import ../../Cloud/Ports.nix;
+in {
+  config.virtualisation.oci-containers.backend = "docker";
+  config.virtualisation.oci-containers.containers.open-webui = {
+    image = "ghcr.io/open-webui/open-webui:main";
+    volumes = ["/var/lib/open-webui:/app/backend/data"];
+    environment = {
+      OLLAMA_BASE_URL = "http://127.0.0.1:${toString ports.ollama}";
+      PORT = ports.open-webui;
+    };
+    extraOptions = ["--network=host"];
+  };
+}