feat(wolf): show Hermes GPU fallback state

backend/atlas_portal/ariadne_client.py

@@ -43,6 +43,7 @@ def request_raw(
     *,
     payload: Any | None = None,
     params: dict[str, Any] | None = None,
+    timeout_sec: float | None = None,
 ) -> httpx.Response:
     """Send one authenticated request to Ariadne and return the raw response.
 
@@ -55,9 +56,10 @@ def request_raw(
 
     url = _url(path)
     attempts = max(1, settings.ARIADNE_RETRY_COUNT)
+    request_timeout = timeout_sec if timeout_sec is not None else settings.ARIADNE_TIMEOUT_SEC
     for attempt in range(1, attempts + 1):
         try:
-            with httpx.Client(timeout=settings.ARIADNE_TIMEOUT_SEC) as client:
+            with httpx.Client(timeout=request_timeout) as client:
                 resp = client.request(
                     method,
                     url,
@@ -78,7 +80,7 @@ def request_raw(
                     "method": method,
                     "path": path,
                     "attempt": attempt,
-                    "timeout_sec": settings.ARIADNE_TIMEOUT_SEC,
+                    "timeout_sec": request_timeout,
                     "error": str(exc),
                 },
             )
@@ -93,6 +95,7 @@ def proxy(
     *,
     payload: Any | None = None,
     params: dict[str, Any] | None = None,
+    timeout_sec: float | None = None,
 ) -> tuple[Any, int]:
     """Proxy an Ariadne response through Flask as JSON plus status code.
 
@@ -101,7 +104,7 @@ def proxy(
     """
 
     try:
-        resp = request_raw(method, path, payload=payload, params=params)
+        resp = request_raw(method, path, payload=payload, params=params, timeout_sec=timeout_sec)
     except AriadneError as exc:
         return jsonify({"error": str(exc)}), exc.status_code
 

backend/atlas_portal/routes/account_wolf.py

@@ -5,7 +5,7 @@ from typing import Any
 
 from flask import jsonify, request
 
-from .. import ariadne_client
+from .. import ariadne_client, settings
 from ..keycloak import require_auth, require_account_access
 
 
@@ -142,7 +142,12 @@ def register_account_wolf(app) -> None:
         ok, resp = _require_account()
         if not ok:
             return resp
-        return ariadne_client.proxy("POST", "/api/admin/game-mode/start", payload=_json_payload())
+        return ariadne_client.proxy(
+            "POST",
+            "/api/admin/game-mode/start",
+            payload=_json_payload(),
+            timeout_sec=settings.ARIADNE_GAME_MODE_TIMEOUT_SEC,
+        )
 
     @app.route("/api/account/wolf/game-mode/stop", methods=["POST"])
     @require_auth
@@ -150,7 +155,12 @@ def register_account_wolf(app) -> None:
         ok, resp = _require_account()
         if not ok:
             return resp
-        return ariadne_client.proxy("POST", "/api/admin/game-mode/stop", payload=_json_payload())
+        return ariadne_client.proxy(
+            "POST",
+            "/api/admin/game-mode/stop",
+            payload=_json_payload(),
+            timeout_sec=settings.ARIADNE_GAME_MODE_TIMEOUT_SEC,
+        )
 
     @app.route("/api/account/wolf/admin/firewall/unlock", methods=["POST"])
     @require_auth

backend/atlas_portal/settings.py

@@ -80,6 +80,7 @@ KEYCLOAK_ADMIN_REALM = os.getenv("KEYCLOAK_ADMIN_REALM", KEYCLOAK_REALM)
 
 ARIADNE_URL = os.getenv("ARIADNE_URL", "").strip()
 ARIADNE_TIMEOUT_SEC = float(os.getenv("ARIADNE_TIMEOUT_SEC", "10"))
+ARIADNE_GAME_MODE_TIMEOUT_SEC = float(os.getenv("ARIADNE_GAME_MODE_TIMEOUT_SEC", "900"))
 ARIADNE_RETRY_COUNT = int(os.getenv("ARIADNE_RETRY_COUNT", "2"))
 ARIADNE_RETRY_BACKOFF_SEC = float(os.getenv("ARIADNE_RETRY_BACKOFF_SEC", "0.2"))
 

backend/tests/test_account_wolf.py

@@ -8,13 +8,20 @@ from atlas_portal.routes import account_wolf
 class DummyAriadne:
     def __init__(self, enabled: bool = True) -> None:
         self._enabled = enabled
-        self.calls: list[tuple[str, str, object | None, dict | None]] = []
+        self.calls: list[tuple[str, str, object | None, dict | None, float | None]] = []
 
     def enabled(self) -> bool:
         return self._enabled
 
-    def proxy(self, method: str, path: str, payload: object | None = None, params: dict | None = None):
+    def proxy(
-        self.calls.append((method, path, payload, params))
+        self,
+        method: str,
+        path: str,
+        payload: object | None = None,
+        params: dict | None = None,
+        timeout_sec: float | None = None,
+    ):
+        self.calls.append((method, path, payload, params, timeout_sec))
         return jsonify({"path": path, "payload": payload, "params": params})
 
 
@@ -42,7 +49,7 @@ def test_wolf_status_proxies_source_ip(monkeypatch) -> None:
     )
 
     assert resp.status_code == 200
-    assert ariadne.calls == [("GET", "/api/game-stream/status", None, {"source_ip": "1.2.3.4"})]
+    assert ariadne.calls == [("GET", "/api/game-stream/status", None, {"source_ip": "1.2.3.4"}, None)]
 
 
 def test_wolf_status_prefers_public_query_ip(monkeypatch) -> None:
@@ -54,7 +61,7 @@ def test_wolf_status_prefers_public_query_ip(monkeypatch) -> None:
     )
 
     assert resp.status_code == 200
-    assert ariadne.calls == [("GET", "/api/game-stream/status", None, {"source_ip": "181.1.87.186"})]
+    assert ariadne.calls == [("GET", "/api/game-stream/status", None, {"source_ip": "181.1.87.186"}, None)]
 
 
 def test_wolf_unlock_uses_current_source_ip(monkeypatch) -> None:
@@ -68,7 +75,7 @@ def test_wolf_unlock_uses_current_source_ip(monkeypatch) -> None:
     )
 
     assert resp.status_code == 200
-    assert ariadne.calls == [("POST", "/api/game-stream/firewall/unlock", {"ttl_seconds": 120, "ip": "5.6.7.8"}, None)]
+    assert ariadne.calls == [("POST", "/api/game-stream/firewall/unlock", {"ttl_seconds": 120, "ip": "5.6.7.8"}, None, None)]
 
 
 def test_wolf_unlock_prefers_public_payload_ip(monkeypatch) -> None:
@@ -81,7 +88,7 @@ def test_wolf_unlock_prefers_public_payload_ip(monkeypatch) -> None:
     )
 
     assert resp.status_code == 200
-    assert ariadne.calls == [("POST", "/api/game-stream/firewall/unlock", {"ip": "181.1.87.186"}, None)]
+    assert ariadne.calls == [("POST", "/api/game-stream/firewall/unlock", {"ip": "181.1.87.186"}, None, None)]
 
 
 def test_wolf_unlock_ignores_private_payload_ip(monkeypatch) -> None:
@@ -94,7 +101,7 @@ def test_wolf_unlock_ignores_private_payload_ip(monkeypatch) -> None:
     )
 
     assert resp.status_code == 200
-    assert ariadne.calls == [("POST", "/api/game-stream/firewall/unlock", {"ip": "5.6.7.8"}, None)]
+    assert ariadne.calls == [("POST", "/api/game-stream/firewall/unlock", {"ip": "5.6.7.8"}, None, None)]
 
 
 def test_wolf_source_ip_prefers_nearest_public_proxy_value(monkeypatch) -> None:
@@ -108,7 +115,7 @@ def test_wolf_source_ip_prefers_nearest_public_proxy_value(monkeypatch) -> None:
     )
 
     assert resp.status_code == 200
-    assert ariadne.calls == [("POST", "/api/game-stream/firewall/unlock", {"ip": "181.1.87.186"}, None)]
+    assert ariadne.calls == [("POST", "/api/game-stream/firewall/unlock", {"ip": "181.1.87.186"}, None, None)]
 
 
 def test_wolf_pairing_and_admin_actions_proxy(monkeypatch) -> None:
@@ -125,11 +132,11 @@ def test_wolf_pairing_and_admin_actions_proxy(monkeypatch) -> None:
     client.post("/api/account/wolf/admin/firewall/unlock", json={"ip": "8.8.8.8", "target_user": "olya"})
 
     assert ariadne.calls == [
-        ("GET", "/api/game-stream/pairing/status", None, {"source_ip": "181.1.87.186"}),
+        ("GET", "/api/game-stream/pairing/status", None, {"source_ip": "181.1.87.186"}, None),
-        ("POST", "/api/game-stream/pairing/submit-pin", {"pair_secret": "secret", "pin": "1234", "source_ip": "181.1.87.186"}, None),
+        ("POST", "/api/game-stream/pairing/submit-pin", {"pair_secret": "secret", "pin": "1234", "source_ip": "181.1.87.186"}, None, None),
-        ("POST", "/api/admin/game-mode/start", {"game": "steam"}, None),
+        ("POST", "/api/admin/game-mode/start", {"game": "steam"}, None, account_wolf.settings.ARIADNE_GAME_MODE_TIMEOUT_SEC),
-        ("POST", "/api/admin/game-mode/stop", {"game": "steam"}, None),
+        ("POST", "/api/admin/game-mode/stop", {"game": "steam"}, None, account_wolf.settings.ARIADNE_GAME_MODE_TIMEOUT_SEC),
-        ("POST", "/api/admin/game-stream/firewall/unlock", {"ip": "8.8.8.8", "target_user": "olya"}, None),
+        ("POST", "/api/admin/game-stream/firewall/unlock", {"ip": "8.8.8.8", "target_user": "olya"}, None, None),
     ]
 
 
@@ -143,7 +150,7 @@ def test_wolf_user_revoke_accepts_public_payload_ip(monkeypatch) -> None:
     )
 
     assert resp.status_code == 200
-    assert ariadne.calls == [("POST", "/api/game-stream/firewall/revoke", {"ip": "9.9.9.9"}, None)]
+    assert ariadne.calls == [("POST", "/api/game-stream/firewall/revoke", {"ip": "9.9.9.9"}, None, None)]
 
 
 def test_wolf_routes_require_account_and_ariadne(monkeypatch) -> None:

frontend/src/account/useWolfDashboard.js

@@ -47,8 +47,13 @@ export function useWolfDashboard() {
     sourceIp: "",
     unlockTtlSeconds: 28800,
     gpuPriority: "unknown",
+    gpuOwner: "unknown",
     gameModeStatus: "unknown",
     gameModeActive: false,
+    localModel: "unknown",
+    localModelLoaded: false,
+    localInferenceReady: false,
+    inferencePath: "fallback",
     selectedGame: "steam",
     note: "",
     manualIp: "",
@@ -82,8 +87,14 @@ export function useWolfDashboard() {
       wolf.sourceIp = data.moonlight?.source_ip || "";
       wolf.unlockTtlSeconds = Number(data.moonlight?.unlock_ttl_seconds || 28800);
       wolf.gpuPriority = data.gpu?.priority || "unknown";
-      wolf.gameModeStatus = data.gpu?.game_mode?.status || "unknown";
+      const gameMode = data.gpu?.game_mode || {};
-      wolf.gameModeActive = Boolean(data.gpu?.game_mode?.active);
+      wolf.gpuOwner = gameMode.gpu_owner || "unknown";
+      wolf.gameModeStatus = gameMode.status || "unknown";
+      wolf.gameModeActive = Boolean(gameMode.active);
+      wolf.localModel = gameMode.model || "unknown";
+      wolf.localModelLoaded = Boolean(gameMode.model_loaded);
+      wolf.localInferenceReady = Boolean(gameMode.local_inference_ready);
+      wolf.inferencePath = gameMode.inference_path || "fallback";
       wolf.clients = Array.isArray(data.wolf?.clients) ? data.wolf.clients : [];
       wolf.pendingPairRequests = Array.isArray(data.wolf?.pending_pair_requests) ? data.wolf.pending_pair_requests : [];
       wolf.sessions = Array.isArray(data.wolf?.sessions) ? data.wolf.sessions : [];

frontend/src/components/WolfAccountCard.vue

@@ -35,6 +35,12 @@
         <span class="k mono">GPU</span>
         <strong class="mono">{{ wolf.gpuPriority }}</strong>
       </div>
+      <div class="wolf-summary-item">
+        <span class="k mono">Hermes</span>
+        <strong class="mono" :class="wolf.localInferenceReady ? 'ok-text' : 'warn-text'">
+          {{ wolf.inferencePath }}
+        </strong>
+      </div>
       <div class="wolf-summary-item">
         <span class="k mono">Firewall</span>
         <strong class="mono" :class="currentIpUnlocked ? 'ok-text' : 'warn-text'">
@@ -60,6 +66,14 @@
         <span class="k mono">Paired devices</span>
         <span class="v mono">{{ pairedDeviceNames }}</span>
       </div>
+      <div class="row">
+        <span class="k mono">GPU owner</span>
+        <span class="v mono">{{ wolf.gpuOwner }}</span>
+      </div>
+      <div class="row">
+        <span class="k mono">Local model</span>
+        <span class="v mono">{{ wolf.localModel }} · {{ wolf.localModelLoaded ? "loaded" : "unloaded" }}</span>
+      </div>
     </div>
 
     <div class="actions">

testing/frontend/unit/account-dashboard.spec.js

@@ -215,7 +215,18 @@ describe("account dashboard", () => {
         return jsonResponse({
           can_control_gpu: true,
           moonlight: { host: "moonlight.bstein.dev", source_ip: "181.1.87.186", unlock_ttl_seconds: 28800 },
-          gpu: { priority: "ai", game_mode: { status: "idle", active: false } },
+          gpu: {
+            priority: "ai",
+            game_mode: {
+              status: "idle",
+              active: false,
+              gpu_owner: "hermes",
+              model: "gpt-oss:20b",
+              model_loaded: true,
+              local_inference_ready: true,
+              inference_path: "local",
+            },
+          },
           wolf: {
             api_enabled: true,
             clients: [{ name: "Desktop" }],
@@ -236,6 +247,11 @@ describe("account dashboard", () => {
 
     expect(dashboard.wolf.status).toBe("ready");
     expect(dashboard.wolf.canControlGpu).toBe(true);
+    expect(dashboard.wolf.gpuOwner).toBe("hermes");
+    expect(dashboard.wolf.localModel).toBe("gpt-oss:20b");
+    expect(dashboard.wolf.localModelLoaded).toBe(true);
+    expect(dashboard.wolf.localInferenceReady).toBe(true);
+    expect(dashboard.wolf.inferencePath).toBe("local");
     expect(dashboard.wolf.clients[0].name).toBe("Desktop");
     expect(dashboard.wolf.pendingPairRequests[0].pair_secret).toBe("secret-1");