hermes - 💡(How to fix) Fix API server ignores per-platform model config (no way to run api_server on a different model than the global default)

model = _resolve_gateway_model()

---

model:
  default: claude-opus-4-8
platform_models:
  api_server:
    default: claude-sonnet-4-6   # or a bare string: api_server: claude-sonnet-4-6

---

diff --git a/gateway/platforms/api_server.py b/gateway/platforms/api_server.py
index a18630f85..92585e6bd 100644
--- a/gateway/platforms/api_server.py
+++ b/gateway/platforms/api_server.py
@@ -963,9 +963,9 @@ class APIServerAdapter(BasePlatformAdapter):
 
         runtime_kwargs = _resolve_runtime_agent_kwargs()
         reasoning_config = GatewayRunner._load_reasoning_config()
-        model = _resolve_gateway_model()
 
         user_config = _load_gateway_config()
+        model = _resolve_gateway_model(user_config, platform="api_server")
         enabled_toolsets = sorted(_get_platform_tools(user_config, "api_server"))
 
         max_iterations = int(os.getenv("HERMES_MAX_ITERATIONS", "90"))
diff --git a/gateway/run.py b/gateway/run.py
index a2e41c609..f82a276a1 100644
--- a/gateway/run.py
+++ b/gateway/run.py
@@ -1443,14 +1443,39 @@ def _load_gateway_runtime_config() -> dict:
     return expanded if isinstance(expanded, dict) else {}
 
 
-def _resolve_gateway_model(config: dict | None = None) -> str:
+def _resolve_gateway_model(config: dict | None = None, platform: str | None = None) -> str:
     """Read model from config.yaml — single source of truth.
 
     Without this, temporary AIAgent instances (e.g. /compress) fall
     back to the hardcoded default which fails when the active provider is
     openai-codex.
+
+    Per-platform override (opt-in): when ``platform`` is supplied AND
+    ``platform_models.<platform>`` is set in config.yaml, that model wins
+    over the global ``model.default``. This lets a single platform (e.g.
+    the API server) run a cheaper/faster model without affecting any other
+    platform. Callers that omit ``platform`` — every existing call site —
+    are completely unaffected and resolve the global default as before.
+
+    The override value may be a bare model string, or a mapping with a
+    ``default`` (or ``model``) key. Any ``provider`` key in the mapping is
+    NOT consumed here — provider/credentials still come from the global
+    runtime config, so a platform override must name a model that works
+    with the active provider.
     """
     cfg = config if config is not None else _load_gateway_config()
+
+    if platform:
+        platform_models = cfg.get("platform_models")
+        if isinstance(platform_models, dict):
+            override = platform_models.get(platform)
+            if isinstance(override, str) and override:
+                return override
+            if isinstance(override, dict):
+                model = override.get("default") or override.get("model")
+                if model:
+                    return model
+
     model_cfg = cfg.get("model", {})
     if isinstance(model_cfg, str):
         return model_cfg