fix(#3517): metrics are not optional, and the endpoint is denied by default

Two review findings, folded together.

mara: a swarm integrated auto deployed forge always has metrics, so the
toggle is gone. The endpoint follows behindGateway instead, which is the
swarm-integrated shape and the condition the protected location lives
under. Serving it without that location would put it on a listener
openFirewall can expose with nothing in front.

argus: /metrics matched no access_control rule, so default_policy
one_factor governed it. That is any authenticated subject, which today
means any operator and tomorrow any agent. My audience argument covered
the Bearer path only; the same endpoint also accepts CookieSession, and a
cookie carries no audience at all, so the audience was never what stood
between a browser session and this data.

The rule is deny rather than a client-scoped allow because the collector's
client does not exist yet. Authelia refuses a subject naming an
unregistered client, and does so in a preStart validator rather than at
build time, so naming one early yields a green nixos-rebuild and dead
swarm SSO on the next restart. Denying until the client is registered
makes publishing the endpoint safe on its own; registering it is a
one-line change from deny to that allow.

Rule order is load-bearing: authelia takes the first match.

nix/host-modules/hive-forge/default.nix

@@ -450,15 +450,54 @@ in
       "${cfg.domain}" = (gatewayCfg.lib.tlsFor cfg.domain) // {
         listen = gatewayCfg.lib.listen;
         extraConfig = gatewayCfg.lib.securityHeaders;
-        locations."/" = {
-          proxyPass = "http://127.0.0.1:${toString cfg.httpPort}/";
-          proxyWebsockets = true;
-          extraConfig = ''
-            proxy_buffering off;
-            client_max_body_size 1G;
-            proxy_read_timeout 1h;
-            proxy_send_timeout 1h;
-          '';
+        locations = {
+          "/" = {
+            proxyPass = "http://127.0.0.1:${toString cfg.httpPort}/";
+            proxyWebsockets = true;
+            extraConfig = ''
+              proxy_buffering off;
+              client_max_body_size 1G;
+              proxy_read_timeout 1h;
+              proxy_send_timeout 1h;
+            '';
+          };
+
+          # ⚠️ EXACT match, and that is what makes this safe. Forgejo
+          # serves `/metrics` on the same listener the `/` prefix above
+          # already proxies, so without a more specific location the
+          # endpoint would ride that catch-all to anyone. `= /metrics`
+          # outranks the `/` prefix in nginx, so this location — and its
+          # auth — is the one that runs.
+          "= /metrics" = {
+            proxyPass = "http://127.0.0.1:${toString cfg.httpPort}/metrics";
+            extraConfig = ''
+              auth_request /__forge_metrics_authz;
+            '';
+          };
+
+          # The subrequest. Same implementation and header set as
+          # `swarm-ui.nix` uses, for the same reason: `X-Original-URL`
+          # and `X-Original-Method` are what authelia's `auth-request`
+          # implementation reads.
+          #
+          # ⚠️ NO `error_page 401 =302` here, and its absence is the
+          # whole point. The swarm UI redirects an unauthenticated
+          # browser to a login page; a scraper handed that 302 would
+          # follow it and parse an HTML page as metrics. A machine-facing
+          # location lets the 401 reach the client unchanged.
+          "= /__forge_metrics_authz" = {
+            proxyPass = "http://127.0.0.1:${toString autheliaCfg.port}/api/authz/auth-request";
+            extraConfig = ''
+              internal;
+              proxy_pass_request_body off;
+              proxy_set_header Content-Length "";
+              proxy_set_header X-Original-Method $request_method;
+              proxy_set_header X-Original-URL $scheme://$http_host$request_uri;
+              proxy_set_header X-Forwarded-Proto $scheme;
+              proxy_set_header X-Forwarded-Host $http_host;
+              proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+            '';
+          };
         };
       };
     };
@@ -674,6 +713,20 @@ in
                 DEFAULT_BRANCH = "main";
                 DEFAULT_PRIVATE = "private";
               };
+              # Not an option: a swarm-integrated, auto-deployed forge
+              # always has metrics. Tied to `behindGateway` because that
+              # IS the swarm-integrated shape — it is the condition under
+              # which the protected `= /metrics` location below exists.
+              # Serving the endpoint without that location would put it on
+              # a listener `openFirewall` can expose, with nothing in
+              # front of it.
+              #
+              # No `TOKEN` here on purpose. Forgejo can guard this itself
+              # with a static bearer, but the swarm authenticates the
+              # scraper at the gateway, so a second credential system per
+              # service would buy nothing and would be the one that stops
+              # getting rotated.
+              metrics.ENABLED = cfg.behindGateway;
               # Repo migrations / pull-mirrors fetch from the source
               # URL *inside* Forgejo. hyperhive code is synced from
               # `localhost` (and the host LAN), which Forgejo's

nix/host-modules/swarm-authelia.nix

@@ -42,6 +42,7 @@ let
   hyperhiveDomain = hyperhiveCfg.domain;
   swarmDomain = hyperhiveCfg.swarm.domain;
   uiCfg = hyperhiveCfg.swarm.ui;
+  forgeCfg = hyperhiveCfg.swarm.forge;
 
   # Group an account must hold to reach operator-only surfaces. Named
   # here because this module writes the rule that enforces it and
@@ -1078,6 +1079,35 @@ in
             settings = {
               theme = "dark";
               server.address = "tcp://127.0.0.1:${toString cfg.port}";
+
+              # Let a machine present an OAuth2 access token to the same
+              # `auth_request` endpoint browsers use, so a scraper can be
+              # authenticated by the gateway instead of every service
+              # growing its own static bearer.
+              #
+              # ⚠️ `authn_strategies` REPLACES the defaults rather than
+              # adding to them, so `CookieSession` is listed explicitly.
+              # Dropping it does not fail to evaluate and does not fail to
+              # start — it silently ends every operator session on the
+              # swarm UI, which rides this same endpoint.
+              #
+              # Unconditional, and not keyed to whichever service is
+              # currently scraped: this only makes a *scheme* available.
+              # Authorisation is the audience — authelia refuses a token
+              # that carries no audience for the requested URL, and a
+              # client may only be issued audiences it is registered for.
+              # So enabling the scheme grants nobody anything until a
+              # client is registered for a specific URL.
+              server.endpoints.authz.auth-request = {
+                implementation = "AuthRequest";
+                authn_strategies = [
+                  {
+                    name = "HeaderAuthorization";
+                    schemes = [ "Bearer" ];
+                  }
+                  { name = "CookieSession"; }
+                ];
+              };
               log.level = "info";
 
               # `watch` is load-bearing, not a convenience: authelia reads
@@ -1105,11 +1135,41 @@ in
               # the account to disagree silently.
               access_control = {
                 default_policy = "one_factor";
-                rules = lib.optional uiCfg.enable {
-                  domain = uiCfg.domain;
-                  subject = [ "group:${operatorGroup}" ];
-                  policy = "one_factor";
-                };
+                # ⚠️ ORDER MATTERS — authelia takes the FIRST matching rule.
+                # The metrics rule is listed first so it cannot be shadowed
+                # by a broader domain rule added later.
+                rules =
+                  # The forge's metrics endpoint. `deny` is deliberate and
+                  # is the whole protection right now: the endpoint is
+                  # always served (a swarm-integrated forge always has
+                  # metrics), and `default_policy` is `one_factor`, which
+                  # means *any* authenticated subject — every operator
+                  # today, every agent once they hold authelia accounts.
+                  #
+                  # Being reachable by a Bearer token is not sufficient on
+                  # its own: `authn_strategies` on this endpoint also
+                  # accepts `CookieSession`, and a cookie carries no
+                  # audience, so the audience is not what stands between a
+                  # browser session and this data.
+                  #
+                  # The collector gets in by REPLACING this with a
+                  # client-scoped allow (`subject = ["oauth2:client:<id>"]`)
+                  # once such a client is registered. Denying until then is
+                  # what makes publishing the endpoint safe on its own —
+                  # authelia refuses a subject naming a client that is not
+                  # registered, and it does so in a `preStart` validator,
+                  # so naming one early takes the whole SSO service down on
+                  # the next restart rather than failing the build.
+                  lib.optional forgeCfg.behindGateway {
+                    domain = forgeCfg.domain;
+                    resources = [ "^/metrics$" ];
+                    policy = "deny";
+                  }
+                  ++ lib.optional uiCfg.enable {
+                    domain = uiCfg.domain;
+                    subject = [ "group:${operatorGroup}" ];
+                    policy = "one_factor";
+                  };
               };
 
               # The cookie domain is the SWARM's domain, NOT authelia's