fix: support GPT-5 request contract (max_completion_tokens, reasoning_effort) and switch aimaq voice decisions to gpt-5-nano

services/ai_orchestrator_service/app.py

@@ -2357,13 +2357,18 @@ def _request_structured_model_decision(
     if not _ai_api_base() or not _ai_api_key():
         raise RuntimeError("AI_API_BASE / AI_API_KEY are required for openai_compatible provider")
     started = time.perf_counter()
-    payload = {
-        "model": _ai_model(),
-        "temperature": 0.2,
-        "max_tokens": _ai_decision_max_tokens(),
+    model = _ai_model()
+    payload: dict[str, Any] = {
+        "model": model,
         "response_format": {"type": "json_object"},
         "messages": messages,
     }
+    if model.startswith("gpt-5"):
+        payload["max_completion_tokens"] = _ai_decision_max_tokens()
+        payload["reasoning_effort"] = "minimal"
+    else:
+        payload["temperature"] = 0.2
+        payload["max_tokens"] = _ai_decision_max_tokens()
     effective_timeout = timeout_seconds if timeout_seconds is not None else _ai_timeout_seconds()
     with httpx.Client(timeout=effective_timeout) as client:
         response = client.post(