fix(gateway): scope session-only /model context_length to configured route

gateway/slash_commands.py

@@ -1732,6 +1732,10 @@ async def _handle_model_command(self, event: MessageEvent) -> Optional[str]:
         user_provs = None
         custom_provs = None
         excluded_provs = []
+        configured_model = None
+        configured_provider = None
+        configured_base_url = None
+        configured_context_length = None
         config_path = (_command_profile_home or _hermes_home) / "config.yaml"
         try:
             cfg = _load_gateway_config()
@@ -1741,6 +1745,12 @@ async def _handle_model_command(self, event: MessageEvent) -> Optional[str]:
                     current_model = model_cfg.get("default", "")
                     current_provider = model_cfg.get("provider", current_provider)
                     current_base_url = model_cfg.get("base_url", "")
+                    configured_model = model_cfg.get("default") or model_cfg.get("model")
+                    configured_provider = model_cfg.get("provider")
+                    configured_base_url = model_cfg.get("base_url")
+                    raw_context_length = model_cfg.get("context_length")
+                    if raw_context_length is not None:
+                        configured_context_length = int(raw_context_length)
                 user_provs = cfg.get("providers")
                 try:
                     from hermes_cli.config import get_compatible_custom_providers
@@ -2016,33 +2026,17 @@ async def _on_model_selected_scoped(
                         lines.append(t("gateway.model.provider_label", provider=plabel))
                         mi = result.model_info
                         from hermes_cli.model_switch import resolve_display_context_length_async
-                        _sw_config_ctx = None
-                        _sw_model_cfg = {}
-                        try:
-                            _sw_cfg = _load_gateway_config()
-                            _sw_model_cfg = _sw_cfg.get("model", {})
-                            if isinstance(_sw_model_cfg, dict):
-                                _sw_raw = _sw_model_cfg.get("context_length")
-                                if _sw_raw is not None:
-                                    _sw_config_ctx = int(_sw_raw)
-                        except Exception:
-                            pass
-                        if not isinstance(_sw_model_cfg, dict):
-                            _sw_model_cfg = {}
                         ctx = await resolve_display_context_length_async(
                             result.new_model,
                             result.target_provider,
                             base_url=result.base_url or current_base_url or "",
                             api_key=result.api_key or current_api_key or "",
                             model_info=mi,
                             custom_providers=custom_provs,
-                            config_context_length=_sw_config_ctx,
-                            configured_model=(
-                                _sw_model_cfg.get("default")
-                                or _sw_model_cfg.get("model")
-                            ),
-                            configured_provider=_sw_model_cfg.get("provider"),
-                            configured_base_url=_sw_model_cfg.get("base_url"),
+                            config_context_length=configured_context_length,
+                            configured_model=configured_model,
+                            configured_provider=configured_provider,
+                            configured_base_url=configured_base_url,
                         )
                         if ctx:
                             lines.append(t("gateway.model.context_label", tokens=f"{ctx:,}"))
@@ -2342,33 +2336,17 @@ async def _finish_switch() -> str:
             # Copilot, and Nous-enforced caps win over the raw models.dev entry.
             mi = result.model_info
             from hermes_cli.model_switch import resolve_display_context_length_async
-            _sw2_config_ctx = None
-            _sw2_model_cfg = {}
-            try:
-                _sw2_cfg = _load_gateway_config()
-                _sw2_model_cfg = _sw2_cfg.get("model", {})
-                if isinstance(_sw2_model_cfg, dict):
-                    _sw2_raw = _sw2_model_cfg.get("context_length")
-                    if _sw2_raw is not None:
-                        _sw2_config_ctx = int(_sw2_raw)
-            except Exception:
-                pass
-            if not isinstance(_sw2_model_cfg, dict):
-                _sw2_model_cfg = {}
             ctx = await resolve_display_context_length_async(
                 result.new_model,
                 result.target_provider,
                 base_url=result.base_url or current_base_url or "",
                 api_key=result.api_key or current_api_key or "",
                 model_info=mi,
                 custom_providers=custom_provs,
-                config_context_length=_sw2_config_ctx,
-                configured_model=(
-                    _sw2_model_cfg.get("default")
-                    or _sw2_model_cfg.get("model")
-                ),
-                configured_provider=_sw2_model_cfg.get("provider"),
-                configured_base_url=_sw2_model_cfg.get("base_url"),
+                config_context_length=configured_context_length,
+                configured_model=configured_model,
+                configured_provider=configured_provider,
+                configured_base_url=configured_base_url,
             )
             if ctx:
                 lines.append(t("gateway.model.context_label", tokens=f"{ctx:,}"))

tests/gateway/test_model_command_context_scope.py

@@ -0,0 +1,151 @@
+"""Regression tests for context-length display during gateway /model switches."""
+
+from __future__ import annotations
+
+import pytest
+import yaml
+
+from gateway.config import Platform
+from gateway.platforms.base import MessageEvent, MessageType
+from gateway.run import GatewayRunner
+from gateway.session import SessionSource
+
+
+def _make_runner():
+    runner = object.__new__(GatewayRunner)
+    runner.adapters = {}
+    runner._voice_mode = {}
+    runner._session_model_overrides = {}
+    runner._running_agents = {}
+    return runner
+
+
+def _make_event(text: str) -> MessageEvent:
+    return MessageEvent(
+        text=text,
+        message_type=MessageType.TEXT,
+        source=SessionSource(
+            platform=Platform.DISCORD,
+            chat_id="12345",
+            chat_type="thread",
+            thread_id="12345",
+        ),
+    )
+
+
+def _codex_switch_result():
+    from hermes_cli.model_switch import ModelSwitchResult
+
+    return ModelSwitchResult(
+        success=True,
+        new_model="gpt-5.6-sol",
+        target_provider="openai-codex",
+        provider_changed=True,
+        api_key="codex-token",
+        base_url="https://chatgpt.com/backend-api/codex",
+        api_mode="codex_responses",
+        provider_label="OpenAI Codex",
+        is_global=False,
+    )
+
+
+def _write_pinned_profile(hermes_home) -> None:
+    """Globally pin the profile to a 1M-token model on a different route."""
+    (hermes_home / "config.yaml").write_text(
+        yaml.safe_dump({
+            "model": {
+                "default": "glm-5.2",
+                "provider": "zai",
+                "context_length": 1_000_000,
+            },
+            "providers": {},
+        }),
+        encoding="utf-8",
+    )
+
+
+def _apply_common_mocks(monkeypatch, hermes_home) -> None:
+    """Point every hermes-home lookup at *hermes_home* and stub the switch."""
+    import gateway.run as gateway_run
+
+    monkeypatch.setattr(gateway_run, "_hermes_home", hermes_home)
+    monkeypatch.setattr("hermes_constants.get_hermes_home", lambda: hermes_home)
+    monkeypatch.setattr("hermes_cli.config.get_hermes_home", lambda: hermes_home)
+    monkeypatch.setattr(
+        "hermes_cli.model_switch.switch_model", lambda **kw: _codex_switch_result()
+    )
+    monkeypatch.setattr(
+        "hermes_cli.model_cost_guard.expensive_model_warning", lambda *a, **kw: None
+    )
+
+
+def _read_profile(hermes_home) -> dict:
+    return yaml.safe_load((hermes_home / "config.yaml").read_text(encoding="utf-8"))
+
+
+@pytest.mark.asyncio
+async def test_session_model_switch_does_not_reuse_global_context_override(
+    tmp_path, monkeypatch
+):
+    """A session-only Codex switch must display Codex's real 272K cap.
+
+    The profile can be globally configured for a different 1M-token model.
+    Before the fix, the confirmation message reused that stale global
+    ``model.context_length`` and claimed Codex gpt-5.6-sol had 1,000,000 tokens.
+
+    ``--session`` opts out of persistence, so the originally configured route
+    must also survive the switch unchanged on disk.
+    """
+    hermes_home = tmp_path / ".hermes"
+    hermes_home.mkdir()
+    _write_pinned_profile(hermes_home)
+    _apply_common_mocks(monkeypatch, hermes_home)
+
+    result = await _make_runner()._handle_model_command(
+        _make_event("/model gpt-5.6-sol --provider openai-codex --session")
+    )
+
+    assert result is not None
+    assert "gpt-5.6-sol" in result
+    assert "272,000" in result
+    assert "1,000,000" not in result
+
+    # Session-only: the configured route is left untouched on disk.
+    persisted = _read_profile(hermes_home)["model"]
+    assert persisted["default"] == "glm-5.2"
+    assert persisted["provider"] == "zai"
+    assert persisted["context_length"] == 1_000_000
+
+
+@pytest.mark.asyncio
+async def test_global_model_switch_drops_stale_context_pin(tmp_path, monkeypatch):
+    """A persistent (--global) switch to a new route must not inherit the pin.
+
+    hermes-sweeper (#48187): the original guard reloaded config after the
+    switch had already written ``model.default``/``model.provider``, saw the
+    new target route, and accepted the inherited 1M ``context_length`` for
+    persistent switches. The confirmation must show Codex's real 272K cap
+    (the session-only check uses pre-switch route state), and the persisted
+    pin must be dropped because its route no longer matches — the global
+    switch stays explicitly scoped instead of inheriting the stale override.
+    """
+    hermes_home = tmp_path / ".hermes"
+    hermes_home.mkdir()
+    _write_pinned_profile(hermes_home)
+    _apply_common_mocks(monkeypatch, hermes_home)
+
+    result = await _make_runner()._handle_model_command(
+        _make_event("/model gpt-5.6-sol --provider openai-codex --global")
+    )
+
+    assert result is not None
+    assert "gpt-5.6-sol" in result
+    # Display resolves against the pre-switch route, so the stale 1M pin is cleared.
+    assert "272,000" in result
+    assert "1,000,000" not in result
+
+    # Persisted route moved to Codex; the route-mismatched pin is dropped, not inherited.
+    persisted = _read_profile(hermes_home)["model"]
+    assert persisted["default"] == "gpt-5.6-sol"
+    assert persisted["provider"] == "openai-codex"
+    assert "context_length" not in persisted