diff --git a/app/services/ai/cloudflare_ai.py b/app/services/ai/cloudflare_ai.py
index d58a34c..3c523e8 100644
--- a/app/services/ai/cloudflare_ai.py
+++ b/app/services/ai/cloudflare_ai.py
@@ -10,7 +10,7 @@
 
 50K credits available for inference at the edge.
 
-Model Selection Strategy (consistent with workers/src/config/models.ts):
+Model Selection Strategy (aligned with the Workers model tier selection in workers/src/lib/workers-ai.ts):
 - SPEED: llama-3.2-3b-instruct (ultra-fast, <100ms, simple queries)
 - BALANCED: llama-3.3-70b-instruct-fp8-fast (best quality/speed balance)
 - QUALITY: gpt-oss-120b (highest quality, production use)
@@ -110,7 +110,8 @@ class ModelTier(str, Enum):
     """
     Model tiers for automatic selection based on query complexity.
     
-    Consistent with workers/src/config/models.ts TIER_MODELS mapping.
+    Keep in sync with Workers model tier selection logic in 
+    workers/src/lib/workers-ai.ts.
     """
     SPEED = "speed"  # Ultra-fast, simple queries (<100ms)
     BALANCED = "balanced"  # Best quality/speed balance (default)
@@ -119,7 +120,7 @@ class ModelTier(str, Enum):
 
 
 # Tier to model mapping
-# Consistent with workers/src/config/models.ts TIER_MODELS
+# consistent with Workers model tier selection logic in workers/src/lib/workers-ai.ts
 TIER_MODELS = {
     ModelTier.SPEED: CFModel.LLAMA_3_2_3B,  # Ultra-fast for simple tasks
     ModelTier.BALANCED: CFModel.LLAMA_3_3_70B_FAST,  # Best quality/speed balance
diff --git a/app/services/search/multi_search.py b/app/services/search/multi_search.py
index 745f476..b149199 100644
--- a/app/services/search/multi_search.py
+++ b/app/services/search/multi_search.py
@@ -68,7 +68,7 @@ def __init__(self):
         """Initialize multi-provider search service."""
         self.client = httpx.AsyncClient(timeout=30)
         self.providers = self._configure_providers()
-        self.provider_stats = {provider: {"requests": 0, "errors": 0, "avg_latency": 0.0} 
+        self.provider_stats = {provider: {"requests": 0, "errors": 0,"total_latency":0.0, "avg_latency": 0.0} 
                               for provider in SearchProvider}
         
     def _configure_providers(self) -> Dict[SearchProvider, SearchProviderConfig]:
@@ -128,19 +128,21 @@ async def search(self, options: SearchOptions) -> List[SearchResult]:
                 start_time = asyncio.get_event_loop().time()
                 results = await self._search_with_provider(provider_type, options)
                 
-                if results:
-                    # Update stats
-                    latency = asyncio.get_event_loop().time() - start_time
-                    self.provider_stats[provider_type]["requests"] += 1
-                    self.provider_stats[provider_type]["avg_latency"] = (
-                        (self.provider_stats[provider_type]["avg_latency"] + latency) / 2
-                    )
-                    
-                    logger.info("search_successful",
-                               provider=config.name,
-                               results=len(results),
-                               latency=latency)
-                    return results
+                if results:    
+                        # Update stats
+                        latency = asyncio.get_event_loop().time() - start_time
+                        stats = self.provider_stats[provider_type]
+                        stats["requests"] += 1
+                        stats["total_latency"] += latency
+                        stats["avg_latency"] = (
+                            stats["total_latency"] / stats["requests"]
+                        )
+                        
+                        logger.info("search_successful",
+                                provider=config.name,
+                                results=len(results),
+                                latency=latency)
+                        return results
                 
             except Exception as e:
                 self.provider_stats[provider_type]["errors"] += 1