feat(image-gen): OpenRouter image picker lists every live image-output model; xAI edits honor dispatched model

- plugins/image_gen/openrouter: list_models() now queries the endpoint's
  /models catalog filtered to output_modalities containing "image"
  (per-backend 5-min cache, 10s timeout, static 2-model chain as offline
  fallback; openrouter/auto* router pseudo-models excluded). Every image
  model OpenRouter serves — including future releases — is selectable in
  `hermes tools` with no code change. Applies to Nous Portal too via the
  shared provider class.
- plugins/image_gen/xai: forward the dispatched model kwarg into
  _resolve_edit_model() so an explicitly selected edit-capable model is
  honored on /images/edits (extends the salvaged #55893 fix to the edit
  path; text-only models still fall back to quality).
- Tests: OpenRouter live-catalog filtering/exclusions/order, offline
  fallback, cache single-fetch; xAI edit-kwarg forwarding incl. the
  text-only-hijack negative case.

Live-verified against openrouter.ai: 9 image-output models returned and
rendered, matching the public models?output_modalities=image listing.

tests/plugins/image_gen/test_openrouter_compat_provider.py

@@ -129,6 +129,80 @@ class TestProviderClass:
             ]
 
 
+# ---------------------------------------------------------------------------
+# Live model catalog
+# ---------------------------------------------------------------------------
+
+
+def _mock_models_response(entries):
+    resp = MagicMock()
+    resp.status_code = 200
+    resp.raise_for_status = MagicMock()
+    resp.json.return_value = {"data": entries}
+    return resp
+
+
+class TestLiveCatalog:
+    def test_live_catalog_lists_all_image_output_models(self):
+        """Every image-output model on the endpoint is selectable — including
+        ones released after this code shipped."""
+        entries = [
+            {
+                "id": "openai/gpt-5.4-image-2",
+                "name": "GPT-5.4 Image 2",
+                "architecture": {"output_modalities": ["image"], "input_modalities": ["text", "image"]},
+            },
+            {
+                "id": "some-lab/brand-new-image-model",
+                "name": "Brand New",
+                "architecture": {"output_modalities": ["image", "text"], "input_modalities": ["text"]},
+            },
+            {
+                "id": "openai/gpt-5.4",  # text-only: excluded
+                "architecture": {"output_modalities": ["text"], "input_modalities": ["text"]},
+            },
+            {
+                "id": "openrouter/auto",  # router pseudo-model: excluded
+                "architecture": {"output_modalities": ["image", "text"], "input_modalities": ["text"]},
+            },
+        ]
+        provider = _openrouter()
+        with patch(_RUNTIME, return_value=_runtime_ok()), patch(
+            "requests.get", return_value=_mock_models_response(entries)
+        ):
+            models = provider.list_models()
+        ids = [m["id"] for m in models]
+        assert "openai/gpt-5.4-image-2" in ids
+        assert "some-lab/brand-new-image-model" in ids
+        assert "openai/gpt-5.4" not in ids
+        assert "openrouter/auto" not in ids
+        # Default chain models sort first.
+        assert ids[0] == "openai/gpt-5.4-image-2"
+
+    def test_live_failure_falls_back_to_static_chain(self):
+        provider = _openrouter()
+        with patch(_RUNTIME, side_effect=RuntimeError("no creds")):
+            models = provider.list_models()
+        from plugins.image_gen.openrouter import DEFAULT_MODEL, _FALLBACK_MODEL
+
+        assert [m["id"] for m in models] == [DEFAULT_MODEL, _FALLBACK_MODEL]
+
+    def test_live_catalog_is_cached(self):
+        provider = _openrouter()
+        entries = [
+            {
+                "id": "openai/gpt-5.4-image-2",
+                "architecture": {"output_modalities": ["image"], "input_modalities": ["text"]},
+            }
+        ]
+        with patch(_RUNTIME, return_value=_runtime_ok()), patch(
+            "requests.get", return_value=_mock_models_response(entries)
+        ) as mock_get:
+            provider.list_models()
+            provider.list_models()
+        assert mock_get.call_count == 1
+
+
 # ---------------------------------------------------------------------------
 # Helpers
 # ---------------------------------------------------------------------------

tests/plugins/image_gen/test_xai_provider.py

@@ -221,6 +221,23 @@ class TestLiveCatalog:
         monkeypatch.delenv("XAI_IMAGE_MODEL", raising=False)
         assert xai_mod._resolve_edit_model() == "grok-imagine-image-quality"
 
+    def test_edit_model_honors_caller_kwarg(self, monkeypatch):
+        """The dispatched model kwarg reaches the edit path too."""
+        import plugins.image_gen.xai as xai_mod
+
+        live = {
+            "grok-imagine-image-2.0": {"input_modalities": ["text", "image"], "aliases": []},
+            "grok-imagine-image-quality": {"input_modalities": ["text", "image"], "aliases": []},
+        }
+        monkeypatch.setattr(xai_mod, "_fetch_live_models", lambda: live)
+        monkeypatch.setattr(xai_mod, "_LIVE_CACHE", None)
+        monkeypatch.delenv("XAI_IMAGE_MODEL", raising=False)
+        assert xai_mod._resolve_edit_model("grok-imagine-image-2.0") == "grok-imagine-image-2.0"
+        # Text-only caller model must not hijack the edit path.
+        live["grok-imagine-image-2.0"]["input_modalities"] = ["text"]
+        monkeypatch.setattr(xai_mod, "_LIVE_CACHE", None)
+        assert xai_mod._resolve_edit_model("grok-imagine-image-2.0") == "grok-imagine-image-quality"
+
 
 # ---------------------------------------------------------------------------
 # Generate tests