fix(cli): use custom provider default_model when --provider is set

hermes chat --provider <name> without -m sent the global model.default
to the custom endpoint. Named custom entries already expose
default_model via _get_named_custom_provider(); honor that when the
user selected the provider and did not pass an explicit model.

Fixes #86978

cli.py

@@ -4856,6 +4856,21 @@ class HermesCLI(CLIAgentSetupMixin, CLICommandsMixin, CLIBillingMixin):
             or os.getenv("HERMES_INFERENCE_PROVIDER")
             or "auto"
         )
+        # `--provider <custom>` without `-m` must use that entry's
+        # default_model. Otherwise the global model.default is sent to the
+        # custom endpoint and the compressor inherits the wrong context
+        # length (#86978). Explicit `-m` still wins.
+        if not model and provider:
+            try:
+                from hermes_cli.runtime_provider import _get_named_custom_provider
+
+                _named_custom = _get_named_custom_provider(provider)
+            except Exception:
+                _named_custom = None
+            _provider_default = str((_named_custom or {}).get("model") or "").strip()
+            if _provider_default:
+                self.model = _provider_default
+                self._model_is_default = False
         self._provider_source: Optional[str] = None
         self.provider = self.requested_provider
         self.api_mode = "chat_completions"

tests/cli/test_cli_provider_resolution.py

@@ -123,6 +123,66 @@ def _import_cli():
     return importlib.import_module("cli")
 
 
+def test_provider_flag_uses_named_custom_default_model(monkeypatch):
+    """`--provider <custom>` without `-m` uses that entry's default_model (#86978)."""
+    cli = _import_cli()
+    monkeypatch.setitem(
+        cli.CLI_CONFIG,
+        "model",
+        {"default": "tencent/hy3:free", "provider": "nous"},
+    )
+    config = {
+        "model": {"default": "tencent/hy3:free", "provider": "nous"},
+        "providers": {
+            "gmk-lan": {
+                "name": "GMK Local",
+                "base_url": "http://gmk.lan:9931/v1",
+                "api_key": "not-needed",
+                "default_model": "/models/gemma.gguf",
+            }
+        },
+    }
+    monkeypatch.setattr("hermes_cli.config.load_config", lambda: config)
+    monkeypatch.setattr("hermes_cli.runtime_provider.load_config", lambda: config)
+
+    shell = cli.HermesCLI(provider="gmk-lan", compact=True, max_turns=1)
+
+    assert shell.model == "/models/gemma.gguf"
+    assert shell.requested_provider == "gmk-lan"
+
+
+def test_explicit_model_wins_over_provider_default_model(monkeypatch):
+    """`-m` still wins when `--provider` also names a custom default_model."""
+    cli = _import_cli()
+    monkeypatch.setitem(
+        cli.CLI_CONFIG,
+        "model",
+        {"default": "tencent/hy3:free", "provider": "nous"},
+    )
+    config = {
+        "model": {"default": "tencent/hy3:free", "provider": "nous"},
+        "providers": {
+            "gmk-lan": {
+                "name": "GMK Local",
+                "base_url": "http://gmk.lan:9931/v1",
+                "api_key": "not-needed",
+                "default_model": "/models/gemma.gguf",
+            }
+        },
+    }
+    monkeypatch.setattr("hermes_cli.config.load_config", lambda: config)
+    monkeypatch.setattr("hermes_cli.runtime_provider.load_config", lambda: config)
+
+    shell = cli.HermesCLI(
+        provider="gmk-lan",
+        model="explicit-id",
+        compact=True,
+        max_turns=1,
+    )
+
+    assert shell.model == "explicit-id"
+
+
 def test_hermes_cli_init_does_not_eagerly_resolve_runtime_provider(monkeypatch):
     cli = _import_cli()
     calls = {"count": 0}