feat: disable dynamic completions health checks to preserve NIM key RPM

backend.py

@@ -558,70 +558,10 @@ def auth(authorization: str = None):
 
 
 async def check_models_health():
-    global RECOMMENDED_MODEL
-    # Test only the unstable frontier models (the rest). 
-    # The stable ones (Step 3.7 Flash, Nemotron 3 Ultra, Qwen 2.5 Coder) are always free/working.
-    models_to_test = [
-        "moonshotai/kimi-k2.6",
-        "z-ai/glm-5.1",
-        "minimaxai/minimax-m3",
-        "minimaxai/minimax-m2.7",
-        "meta/llama-3.1-405b-instruct",
-    ]
-    
-    best_model = None
-    best_latency = 999.0
-    
-    # Mark stable models as permanently ONLINE in the status map
-    stable_models = [
-        "stepfun-ai/step-3.7-flash",
-        "nvidia/nemotron-3-ultra-550b-a55b",
-        "qwen/qwen2.5-coder-32b-instruct"
-    ]
-    for model in stable_models:
+    # Mark all models as statically ONLINE to save API RPM quotas
+    for model in ALL_MODELS.keys():
         MODEL_STATUSES[model] = {"status": "ONLINE (Stable)", "latency": "Fast", "raw_latency": 0.1}
-
-    log_activity("Periodic health check started: verifying unstable frontier NIM models...")
-    for model in models_to_test:
-        start_time = time.time()
-        try:
-            # Send a fast test prompt
-            async with anyio.fail_after(15.0): # 15 seconds max timeout
-                await nim_client.chat.completions.create(
-                    model=model,
-                    messages=[{"role": "user", "content": "1+1="}],
-                    max_tokens=3,
-                )
-            latency = time.time() - start_time
-            MODEL_STATUSES[model] = {"status": "ONLINE", "latency": f"{latency:.2f}s", "raw_latency": latency}
-            log_activity(f"Model checked: {model} is ONLINE ({latency:.2f}s)")
-            
-            # Choose the fastest online unstable model
-            if latency < best_latency:
-                best_latency = latency
-                best_model = model
-                
-        except Exception as e:
-            MODEL_STATUSES[model] = {"status": "OFFLINE", "latency": "N/A", "raw_latency": 999.0}
-            log_activity(f"Model checked: {model} is OFFLINE / TIMEOUT: {e}")
-            
-    if best_model:
-        RECOMMENDED_MODEL = best_model
-        log_activity(f"Best frontier model selected: {RECOMMENDED_MODEL} ({best_latency:.2f}s)")
-    else:
-        # Fallback to the stable Step 3.7 Flash if all frontier models are offline/throttled
-        RECOMMENDED_MODEL = "stepfun-ai/step-3.7-flash"
-        log_activity(f"All frontier models offline. Falling back to stable recommended model: {RECOMMENDED_MODEL}")
-
-async def periodic_health_check_loop():
-    # Wait 10 seconds after startup before the first check to let the space boot fully
-    await asyncio.sleep(10)
-    while True:
-        try:
-            await check_models_health()
-        except Exception as e:
-            log_activity(f"Health check loop error: {e}")
-        await asyncio.sleep(900) # every 15 minutes (reduce frequency to save quota)
+    log_activity("[Health Check] Zero-cost status check complete: all models marked ONLINE (No API calls made).")
 
 @app.on_event("startup")
 async def startup():