fix(ai): enforce ollama first for drift governance

apps/api/src/services/openclaw.py

@@ -194,14 +194,20 @@ class OpenClawService:
 
     def _cloud_fallback_allowed_for_alert(self, alert_context: dict | None) -> bool:
         """Cloud fallback is allowed after the ordered Ollama lane for alerts."""
+        if alert_context and alert_context.get("allow_cloud_fallback") is False:
+            return False
         if not self._is_incident_alert_context(alert_context):
             return True
         return bool(getattr(settings, "ALERT_AI_ALLOW_CLOUD_FALLBACK", True))
 
     def _alert_enforces_ollama_first(self, alert_context: dict | None) -> bool:
-        """Alert cards must try GCP-A/GCP-B/111 before Gemini backup."""
+        """Alert and AI-governance lanes must try GCP-A/GCP-B/111 before Gemini backup."""
         return (
-            self._is_incident_alert_context(alert_context)
+            bool(alert_context)
+            and (
+                bool(alert_context.get("enforce_ollama_first"))
+                or self._is_incident_alert_context(alert_context)
+            )
             and bool(getattr(settings, "ALERT_AI_ENFORCE_OLLAMA_FIRST", True))
         )
 
@@ -1144,9 +1150,12 @@ class OpenClawService:
                 exec_context = dict(alert_context) if alert_context else {}
                 if decision.intent == IntentType.DIAGNOSE:
                     exec_context["task_type"] = "diagnose"
-                if self._is_incident_alert_context(alert_context):
-                    exec_context["ollama_model"] = getattr(settings, "ALERT_OLLAMA_MODEL", "qwen3:14b")
-                    exec_context["allow_gcp_heavy_model"] = True
+                if self._alert_enforces_ollama_first(alert_context):
+                    exec_context.setdefault("task_type", "diagnose")
+                    exec_context.setdefault("ollama_model", getattr(settings, "ALERT_OLLAMA_MODEL", "qwen3:14b"))
+                    exec_context["allow_gcp_heavy_model"] = bool(
+                        exec_context.get("allow_gcp_heavy_model", True)
+                    )
                     exec_context["alert_requires_ollama_before_cloud"] = True
 
                 result = await executor.execute(