Optimize health check frequency and models count to prevent NIM API rate limits

backend.py

@@ -467,24 +467,17 @@ def auth(authorization: str = None):
 
 async def check_models_health():
     global RECOMMENDED_MODEL
-    # Top models to test
+    # Limit test to top 3 models to prevent NVIDIA NIM quota exhaustion
     models_to_test = [
         "nvidia/nemotron-3-ultra-550b-a55b",
-        "z-ai/glm-5.1",
-        "moonshotai/kimi-k2.6",
-        "minimaxai/minimax-m3",
         "stepfun-ai/step-3.7-flash",
-        "minimaxai/minimax-m2.7",
-        "nvidia/llama-3.1-nemotron-70b-instruct",
-        "meta/llama-3.1-405b-instruct",
         "qwen/qwen2.5-coder-32b-instruct",
-        "meta/llama-3.3-70b-instruct",
     ]
     
     best_model = None
     best_latency = 999.0
     
-    log_activity("Periodic health check started: verifying 10 NIM models...")
+    log_activity("Periodic health check started: verifying top 3 NIM models...")
     for model in models_to_test:
         start_time = time.time()
         try:
@@ -523,7 +516,7 @@ async def periodic_health_check_loop():
             await check_models_health()
         except Exception as e:
             log_activity(f"Health check loop error: {e}")
-        await asyncio.sleep(300) # every 5 minutes
+        await asyncio.sleep(900) # every 15 minutes (reduce frequency to save quota)
 
 @app.on_event("startup")
 async def startup():