nixos/nextjs-ollama-llm-ui: init module

NixOS already has good support for the Ollama
backend service. Now we can benefit from
having a convenient web frontend as well for it.

nixos/modules/services/web-apps/nextjs-ollama-llm-ui.nix

@@ -0,0 +1,87 @@
+{
+  config,
+  pkgs,
+  lib,
+  ...
+}:
+let
+  cfg = config.services.nextjs-ollama-llm-ui;
+  # we have to override the URL to a Ollama service here, because it gets baked into the web app.
+  nextjs-ollama-llm-ui = cfg.package.override { ollamaUrl = "https://ollama.lambdablob.com"; };
+in
+{
+  options = {
+    services.nextjs-ollama-llm-ui = {
+      enable = lib.mkEnableOption ''
+        Simple Ollama web UI service; an easy to use web frontend for a Ollama backend service.
+        Run state-of-the-art AI large language models (LLM) similar to ChatGPT locally with privacy
+        on your personal computer.
+        This service is stateless and doesn't store any data on the server; all data is kept
+        locally in your web browser.
+        See https://github.com/jakobhoeg/nextjs-ollama-llm-ui.
+
+        Required: You need the Ollama backend service running by having
+        "services.nextjs-ollama-llm-ui.ollamaUrl" point to the correct url.
+        You can host such a backend service with NixOS through "services.ollama".
+      '';
+      package = lib.mkPackageOption pkgs "nextjs-ollama-llm-ui" { };
+
+      hostname = lib.mkOption {
+        type = lib.types.str;
+        default = "127.0.0.1";
+        example = "ui.example.org";
+        description = ''
+          The hostname under which the Ollama UI interface should be accessible.
+          By default it uses localhost/127.0.0.1 to be accessible only from the local machine.
+          Change to "0.0.0.0" to make it directly accessible from the local network.
+
+          Note: You should keep it at 127.0.0.1 and only serve to the local
+          network or internet from a (home) server behind a reverse-proxy and secured encryption.
+          See https://wiki.nixos.org/wiki/Nginx for instructions on how to set up a reverse-proxy.
+        '';
+      };
+
+      port = lib.mkOption {
+        type = lib.types.port;
+        default = 3000;
+        example = 3000;
+        description = ''
+          The port under which the Ollama UI interface should be accessible.
+        '';
+      };
+
+      ollamaUrl = lib.mkOption {
+        type = lib.types.str;
+        default = "127.0.0.1:11434";
+        example = "https://ollama.example.org";
+        description = ''
+          The address (including host and port) under which we can access the Ollama backend server.
+          !Note that if the the UI service is running under a domain "https://ui.example.org",
+          the Ollama backend service must allow "CORS" requests from this domain, e.g. by adding
+          "services.ollama.environment.OLLAMA_ORIGINS = [ ... "https://ui.example.org" ];"!
+        '';
+      };
+    };
+  };
+
+  config = lib.mkIf cfg.enable {
+    systemd.services = {
+
+      nextjs-ollama-llm-ui = {
+        wantedBy = [ "multi-user.target" ];
+        description = "Nextjs Ollama LLM Ui.";
+        after = [ "network.target" ];
+        environment = {
+          HOSTNAME = cfg.hostname;
+          PORT = toString cfg.port;
+          NEXT_PUBLIC_OLLAMA_URL = cfg.ollamaUrl;
+        };
+        serviceConfig = {
+          ExecStart = "${lib.getExe nextjs-ollama-llm-ui}";
+          DynamicUser = true;
+        };
+      };
+    };
+  };
+  meta.maintainers = with lib.maintainers; [ malteneuss ];
+}