nixos/anubis: init module

nixos/doc/manual/redirects.json

@@ -2,6 +2,15 @@
   "book-nixos-manual": [
     "index.html#book-nixos-manual"
   ],
+  "module-services-anubis": [
+    "index.html#module-services-anubis"
+  ],
+  "module-services-anubis-configuration": [
+    "index.html#module-services-anubis-configuration"
+  ],
+  "module-services-anubis-quickstart": [
+    "index.html#module-services-anubis-quickstart"
+  ],
   "module-services-crab-hole": [
     "index.html#module-services-crab-hole"
   ],

nixos/doc/manual/release-notes/rl-2505.section.md

@@ -168,6 +168,8 @@
 
 - [PDS](https://github.com/bluesky-social/pds), Personal Data Server for [bsky](https://bsky.social/). Available as [services.pds](option.html#opt-services.pds).
 
+- [Anubis](https://github.com/TecharoHQ/anubis), a scraper defense software. Available as [services.anubis](options.html#opt-services.anubis).
+
 - [synapse-auto-compressor](https://github.com/matrix-org/rust-synapse-compress-state?tab=readme-ov-file#automated-tool-synapse_auto_compressor), a rust-based matrix-synapse state compressor for postgresql. Available as [services.synapse-auto-compressor](#opt-services.synapse-auto-compressor.enable).
 
 - [mqtt-exporter](https://github.com/kpetremann/mqtt-exporter/), a Prometheus exporter for exposing messages from MQTT. Available as [services.prometheus.exporters.mqtt](#opt-services.prometheus.exporters.mqtt.enable).

nixos/modules/module-list.nix

@@ -1038,6 +1038,7 @@
   ./services/networking/adguardhome.nix
   ./services/networking/alice-lg.nix
   ./services/networking/amuled.nix
+  ./services/networking/anubis.nix
   ./services/networking/aria2.nix
   ./services/networking/asterisk.nix
   ./services/networking/atftpd.nix

nixos/modules/services/networking/anubis.md

@@ -0,0 +1,61 @@
+# Anubis {#module-services-anubis}
+
+[Anubis](https://anubis.techaro.lol) is a scraper defense software that blocks AI scrapers. It is designed to sit
+between a reverse proxy and the service to be protected.
+
+## Quickstart {#module-services-anubis-quickstart}
+
+This module is designed to use Unix domain sockets as the socket paths can be automatically configured for multiple
+instances, but TCP sockets are also supported.
+
+A minimal configuration with [nginx](#opt-services.nginx.enable) may look like the following:
+
+```nix
+{ config, ... }: {
+  services.anubis.instances.default.settings.TARGET = "http://localhost:8000";
+
+  # required due to unix socket permissions
+  users.users.nginx.extraGroups = [ config.users.groups.anubis.name ];
+  services.nginx.virtualHosts."example.com" = {
+    locations = {
+      "/".proxyPass = "http://unix:${config.services.anubis.instances.default.settings.BIND}";
+    };
+  };
+}
+```
+
+If Unix domain sockets are not needed or desired, this module supports operating with only TCP sockets.
+
+```nix
+{
+  services.anubis = {
+    instances.default = {
+      settings = {
+        TARGET = "http://localhost:8080";
+        BIND = ":9000";
+        BIND_NETWORK = "tcp";
+        METRICS_BIND = "127.0.0.1:9001";
+        METRICS_BIND_NETWORK = "tcp";
+      };
+    };
+  };
+}
+```
+
+## Configuration {#module-services-anubis-configuration}
+
+It is possible to configure default settings for all instances of Anubis, via {option}`services.anubis.defaultOptions`.
+
+```nix
+{
+  services.anubis.defaultOptions = {
+    botPolicy = { dnsbl = false; };
+    settings.DIFFICULTY = 3;
+  };
+}
+```
+
+Note that at the moment, a custom bot policy is not merged with the baked-in one. That means to only override a setting
+like `dnsbl`, copying the entire bot policy is required. Check
+[the upstream repository](https://github.com/TecharoHQ/anubis/blob/1509b06cb921aff842e71fbb6636646be6ed5b46/cmd/anubis/botPolicies.json)
+for the policy.

nixos/modules/services/networking/anubis.nix

@@ -0,0 +1,314 @@
+{
+  config,
+  lib,
+  pkgs,
+  ...
+}:
+let
+  inherit (lib) types;
+  jsonFormat = pkgs.formats.json { };
+
+  cfg = config.services.anubis;
+  enabledInstances = lib.filterAttrs (_: conf: conf.enable) cfg.instances;
+  instanceName = name: if name == "" then "anubis" else "anubis-${name}";
+
+  commonSubmodule =
+    isDefault:
+    let
+      mkDefaultOption =
+        path: opts:
+        lib.mkOption (
+          opts
+          // lib.optionalAttrs (!isDefault && opts ? default) {
+            default =
+              lib.attrByPath (lib.splitString "." path)
+                (throw "This is a bug in the Anubis module. Please report this as an issue.")
+                cfg.defaultOptions;
+            defaultText = lib.literalExpression "config.services.anubis.defaultOptions.${path}";
+          }
+        );
+    in
+    { name, ... }:
+    {
+      options = {
+        enable = lib.mkEnableOption "this instance of Anubis" // {
+          default = true;
+        };
+        user = mkDefaultOption "user" {
+          default = "anubis";
+          description = ''
+            The user under which Anubis is run.
+
+            This module utilizes systemd's DynamicUser feature. See the corresponding section in
+            {manpage}`systemd.exec(5)` for more details.
+          '';
+          type = types.str;
+        };
+        group = mkDefaultOption "group" {
+          default = "anubis";
+          description = ''
+            The group under which Anubis is run.
+
+            This module utilizes systemd's DynamicUser feature. See the corresponding section in
+            {manpage}`systemd.exec(5)` for more details.
+          '';
+          type = types.str;
+        };
+
+        botPolicy = lib.mkOption {
+          default = null;
+          description = ''
+            Anubis policy configuration in Nix syntax. Set to `null` to use the baked-in policy which should be
+            sufficient for most use-cases.
+
+            This option has no effect if `settings.POLICY_FNAME` is set to a different value, which is useful for
+            importing an existing configuration.
+
+            See [the documentation](https://anubis.techaro.lol/docs/admin/policies) for details.
+          '';
+          type = types.nullOr jsonFormat.type;
+        };
+
+        extraFlags = mkDefaultOption "extraFlags" {
+          default = [ ];
+          description = "A list of extra flags to be passed to Anubis.";
+          example = [ "-metrics-bind \"\"" ];
+          type = types.listOf types.str;
+        };
+
+        settings = lib.mkOption {
+          default = { };
+          description = ''
+            Freeform configuration via environment variables for Anubis.
+
+            See [the documentation](https://anubis.techaro.lol/docs/admin/installation) for a complete list of
+            available environment variables.
+          '';
+          type = types.submodule [
+            {
+              freeformType =
+                with types;
+                attrsOf (
+                  nullOr (oneOf [
+                    str
+                    int
+                    bool
+                  ])
+                );
+
+              options = {
+                # BIND and METRICS_BIND are defined in instance specific options, since global defaults don't make sense
+                BIND_NETWORK = mkDefaultOption "settings.BIND_NETWORK" {
+                  default = "unix";
+                  description = ''
+                    The network family that Anubis should bind to.
+
+                    Accepts anything supported by Go's [`net.Listen`](https://pkg.go.dev/net#Listen).
+
+                    Common values are `tcp` and `unix`.
+                  '';
+                  example = "tcp";
+                  type = types.str;
+                };
+                METRICS_BIND_NETWORK = mkDefaultOption "settings.METRICS_BIND_NETWORK" {
+                  default = "unix";
+                  description = ''
+                    The network family that the metrics server should bind to.
+
+                    Accepts anything supported by Go's [`net.Listen`](https://pkg.go.dev/net#Listen).
+
+                    Common values are `tcp` and `unix`.
+                  '';
+                  example = "tcp";
+                  type = types.str;
+                };
+                SOCKET_MODE = mkDefaultOption "settings.SOCKET_MODE" {
+                  default = "0770";
+                  description = "The permissions on the Unix domain sockets created.";
+                  example = "0700";
+                  type = types.str;
+                };
+                DIFFICULTY = mkDefaultOption "settings.DIFFICULTY" {
+                  default = 4;
+                  description = ''
+                    The difficulty required for clients to solve the challenge.
+
+                    Currently, this means the amount of leading zeros in a successful response.
+                  '';
+                  type = types.int;
+                  example = 5;
+                };
+                SERVE_ROBOTS_TXT = mkDefaultOption "settings.SERVE_ROBOTS_TXT" {
+                  default = false;
+                  description = ''
+                    Whether to serve a default robots.txt that denies access to common AI bots by name and all other
+                    bots by wildcard.
+                  '';
+                  type = types.bool;
+                };
+
+                # generated by default
+                POLICY_FNAME = mkDefaultOption "settings.POLICY_FNAME" {
+                  default = null;
+                  description = ''
+                    The bot policy file to use. Leave this as `null` to respect the value set in
+                    {option}`services.anubis.instances.<name>.botPolicy`.
+                  '';
+                  type = types.nullOr types.path;
+                };
+              };
+            }
+            (lib.optionalAttrs (!isDefault) (instanceSpecificOptions name))
+          ];
+        };
+      };
+    };
+
+  instanceSpecificOptions = name: {
+    options = {
+      # see other options above
+      BIND = lib.mkOption {
+        default = "/run/anubis/${instanceName name}.sock";
+        description = ''
+          The address that Anubis listens to. See Go's [`net.Listen`](https://pkg.go.dev/net#Listen) for syntax.
+
+          Defaults to Unix domain sockets. To use TCP sockets, set this to a TCP address and `BIND_NETWORK` to `"tcp"`.
+        '';
+        example = ":8080";
+        type = types.str;
+      };
+      METRICS_BIND = lib.mkOption {
+        default = "/run/anubis/${instanceName name}-metrics.sock";
+        description = ''
+          The address Anubis' metrics server listens to. See Go's [`net.Listen`](https://pkg.go.dev/net#Listen) for
+          syntax.
+
+          The metrics server is enabled by default and may be disabled. However, due to implementation details, this is
+          only possible by setting a command line flag. See {option}`services.anubis.defaultOptions.extraFlags` for an
+          example.
+
+          Defaults to Unix domain sockets. To use TCP sockets, set this to a TCP address and `METRICS_BIND_NETWORK` to
+          `"tcp"`.
+        '';
+        example = "127.0.0.1:8081";
+        type = types.str;
+      };
+      TARGET = lib.mkOption {
+        description = ''
+          The reverse proxy target that Anubis is protecting. This is a required option.
+
+          The usage of Unix domain sockets is supported by the following syntax: `unix:///path/to/socket.sock`.
+        '';
+        example = "http://127.0.0.1:8000";
+        type = types.str;
+      };
+    };
+  };
+in
+{
+  options.services.anubis = {
+    package = lib.mkPackageOption pkgs "anubis" { };
+
+    defaultOptions = lib.mkOption {
+      default = { };
+      description = "Default options for all instances of Anubis.";
+      type = types.submodule (commonSubmodule true);
+    };
+
+    instances = lib.mkOption {
+      default = { };
+      description = ''
+        An attribute set of Anubis instances.
+
+        The attribute name may be an empty string, in which case the `-<name>` suffix is not added to the service name
+        and socket paths.
+      '';
+      type = types.attrsOf (types.submodule (commonSubmodule false));
+    };
+  };
+
+  config = lib.mkIf (enabledInstances != { }) {
+    users.users = lib.mkIf (cfg.defaultOptions.user == "anubis") {
+      anubis = {
+        isSystemUser = true;
+        group = cfg.defaultOptions.group;
+      };
+    };
+
+    users.groups = lib.mkIf (cfg.defaultOptions.group == "anubis") {
+      anubis = { };
+    };
+
+    systemd.services = lib.mapAttrs' (
+      name: instance:
+      lib.nameValuePair "${instanceName name}" {
+        description = "Anubis (${if name == "" then "default" else name} instance)";
+        wantedBy = [ "multi-user.target" ];
+        after = [ "network-online.target" ];
+        wants = [ "network-online.target" ];
+
+        environment = lib.mapAttrs (lib.const (lib.generators.mkValueStringDefault { })) (
+          lib.filterAttrs (_: v: v != null) instance.settings
+        );
+
+        serviceConfig = {
+          User = instance.user;
+          Group = instance.group;
+          DynamicUser = true;
+
+          ExecStart = lib.concatStringsSep " " (
+            (lib.singleton (lib.getExe cfg.package)) ++ instance.extraFlags
+          );
+          RuntimeDirectory =
+            if
+              lib.any (lib.hasPrefix "/run/anubis") (
+                with instance.settings;
+                [
+                  BIND
+                  METRICS_BIND
+                ]
+              )
+            then
+              "anubis"
+            else
+              null;
+
+          # hardening
+          NoNewPrivileges = true;
+          CapabilityBoundingSet = null;
+          SystemCallFilter = [
+            "@system-service"
+            "~@privileged"
+          ];
+          SystemCallArchitectures = "native";
+          MemoryDenyWriteExecute = true;
+
+          PrivateUsers = true;
+          PrivateTmp = true;
+          PrivateDevices = true;
+          ProtectHome = true;
+          ProtectClock = true;
+          ProtectHostname = true;
+          ProtectKernelLogs = true;
+          ProtectKernelModules = true;
+          ProtectKernelTunables = true;
+          ProtectProc = "invisible";
+          ProtectSystem = "strict";
+          ProtectControlGroups = "strict";
+          LockPersonality = true;
+          RestrictRealtime = true;
+          RestrictSUIDSGID = true;
+          RestrictNamespaces = true;
+          RestrictAddressFamilies = [
+            "AF_UNIX"
+            "AF_INET"
+            "AF_INET6"
+          ];
+        };
+      }
+    ) enabledInstances;
+  };
+
+  meta.maintainers = with lib.maintainers; [ soopyc ];
+  meta.doc = ./anubis.md;
+}

nixos/tests/all-tests.nix

@@ -174,6 +174,7 @@ in {
   amd-sev = runTest ./amd-sev.nix;
   angie-api = runTest ./angie-api.nix;
   anki-sync-server = runTest ./anki-sync-server.nix;
+  anubis = runTest ./anubis.nix;
   anuko-time-tracker = runTest ./anuko-time-tracker.nix;
   apcupsd = runTest ./apcupsd.nix;
   apfs = runTest ./apfs.nix;

nixos/tests/anubis.nix

@@ -0,0 +1,98 @@
+{ lib, ... }:
+{
+  name = "anubis";
+  meta.maintainers = [ lib.maintainers.soopyc ];
+
+  nodes.machine =
+    {
+      config,
+      pkgs,
+      ...
+    }:
+    {
+      services.anubis.instances = {
+        "".settings.TARGET = "http://localhost:8080";
+
+        "tcp" = {
+          user = "anubis-tcp";
+          group = "anubis-tcp";
+          settings = {
+            TARGET = "http://localhost:8080";
+            BIND = ":9000";
+            BIND_NETWORK = "tcp";
+            METRICS_BIND = ":9001";
+            METRICS_BIND_NETWORK = "tcp";
+          };
+        };
+
+        "unix-upstream" = {
+          group = "nginx";
+          settings.TARGET = "unix:///run/nginx/nginx.sock";
+        };
+      };
+
+      # support
+      users.users.nginx.extraGroups = [ config.users.groups.anubis.name ];
+      services.nginx = {
+        enable = true;
+        recommendedProxySettings = true;
+        virtualHosts."basic.localhost".locations = {
+          "/".proxyPass = "http://unix:${config.services.anubis.instances."".settings.BIND}";
+          "/metrics".proxyPass = "http://unix:${config.services.anubis.instances."".settings.METRICS_BIND}";
+        };
+
+        virtualHosts."tcp.localhost".locations = {
+          "/".proxyPass = "http://localhost:9000";
+          "/metrics".proxyPass = "http://localhost:9001";
+        };
+
+        virtualHosts."unix.localhost".locations = {
+          "/".proxyPass = "http://unix:${config.services.anubis.instances.unix-upstream.settings.BIND}";
+        };
+
+        # emulate an upstream with nginx, listening on tcp and unix sockets.
+        virtualHosts."upstream.localhost" = {
+          default = true; # make nginx match this vhost for `localhost`
+          listen = [
+            { addr = "unix:/run/nginx/nginx.sock"; }
+            {
+              addr = "localhost";
+              port = 8080;
+            }
+          ];
+          locations."/" = {
+            tryFiles = "$uri $uri/index.html =404";
+            root = pkgs.runCommand "anubis-test-upstream" { } ''
+              mkdir $out
+              echo "it works" >> $out/index.html
+            '';
+          };
+        };
+      };
+    };
+
+  testScript = ''
+    for unit in ["nginx", "anubis", "anubis-tcp", "anubis-unix-upstream"]:
+      machine.wait_for_unit(unit + ".service")
+
+    for port in [9000, 9001]:
+      machine.wait_for_open_port(port)
+
+    for instance in ["anubis", "anubis-unix-upstream"]:
+      machine.wait_for_open_unix_socket(f"/run/anubis/{instance}.sock")
+      machine.wait_for_open_unix_socket(f"/run/anubis/{instance}-metrics.sock")
+
+    # Default unix socket mode
+    machine.succeed('curl -f http://basic.localhost | grep "it works"')
+    machine.succeed('curl -f http://basic.localhost -H "User-Agent: Mozilla" | grep anubis')
+    machine.succeed('curl -f http://basic.localhost/metrics | grep anubis_challenges_issued')
+    machine.succeed('curl -f -X POST http://basic.localhost/.within.website/x/cmd/anubis/api/make-challenge | grep challenge')
+
+    # TCP mode
+    machine.succeed('curl -f http://tcp.localhost -H "User-Agent: Mozilla" | grep anubis')
+    machine.succeed('curl -f http://tcp.localhost/metrics | grep anubis_challenges_issued')
+
+    # Upstream is a unix socket mode
+    machine.succeed('curl -f http://unix.localhost/index.html | grep "it works"')
+  '';
+}