changes

app/main_api.py

@@ -276,17 +276,38 @@ class MultiLLMManager:
     
     async def _groq_response(self, prompt: str, max_tokens: int) -> str:
         key = next(self.groq_keys)
-        client = groq.Groq(api_key=key)
         
-        response = client.chat.completions.create(
-            model="llama-3.3-70b-versatile",
-            messages=[{"role": "user", "content": prompt}],
-            temperature=0.1,
-            max_tokens=max_tokens,
-            top_p=0.9
-        )
-        return response.choices[0].message.content.strip()
-    
+        # --- THE FINAL FIX: BYPASS GROQ PROXY BUG ---
+        # Create our own clean HTTP client and pass it to Groq
+        # This bypasses the internal bug that incorrectly handles proxies in HF Spaces
+        
+        try:
+            # Create a clean HTTP client without proxy issues
+            clean_http_client = httpx.Client(
+                timeout=30.0,
+                limits=httpx.Limits(max_keepalive_connections=5, max_connections=10)
+            )
+            
+            client = groq.Groq(
+                api_key=key,
+                http_client=clean_http_client  # <-- This bypasses the proxy bug
+            )
+            
+            response = client.chat.completions.create(
+                model="llama-3.3-70b-versatile",  # Updated to latest model
+                messages=[{"role": "user", "content": prompt}],
+                temperature=0.1,
+                max_tokens=max_tokens,
+                top_p=0.9
+            )
+            
+            return response.choices[0].message.content.strip()
+            
+        except Exception as e:
+            logger.error(f"Groq response error: {e}")
+            raise e  # Let the parent handle fallback
+
+        
     async def _openai_response(self, prompt: str, max_tokens: int) -> str:
         key = next(self.openai_keys)
         openai.api_key = key