Update engine.py

engine.py

@@ -33,41 +33,34 @@ from reportlab.lib.units import inch
 # ── Hugging Face LLM call ────────────────────────────────────────────────────
 import requests
 
+from huggingface_hub import InferenceClient
+import os
+
+# Initialize the official Hugging Face production inference engine
+client = InferenceClient(model=HF_MODEL, token=os.environ.get("HF_TOKEN"))
+
 def ask_llm(prompt: str, temperature: float = 0.7, max_tokens: int = 512) -> str:
     """
-    Hugging Face Serverless Inference API call using raw requests.
-    Uses HF_TOKEN from environment variables.
+    Hugging Face Serverless Inference API call using the official InferenceClient.
+    This automatically manages API cold-starts and internal retry flags.
     """
     try:
-        token = os.environ.get("HF_TOKEN")
-        headers = {"Authorization": f"Bearer {token}"}
-        url = f"https://api-inference.huggingface.co/models/{HF_MODEL}"
+        # Hugging Face API requirements demand a temperature reading strictly > 0
+        safe_temp = temperature if temperature > 0 else 0.01
         
-        payload = {
-            "inputs": prompt,
-            "parameters": {
-                "max_new_tokens": max_tokens,
-                "temperature": temperature,
-                "return_full_text": False
-            }
-        }
-        
-        response = requests.post(url, headers=headers, json=payload, timeout=60)
+        response = client.text_generation(
+            prompt,
+            max_new_tokens=max_tokens,
+            temperature=safe_temp,
+            return_full_text=False
+        )
         
-        if response.status_code != 200:
-            print(f"\n[API ERROR] Status: {response.status_code}")
-            print(f"[API ERROR] Body: {response.text}\n")
-            
-        response.raise_for_status()
-        data = response.json()
+        # Strip structural infrastructure tokens to keep ValidatorAgent clean
+        return response.replace("<s>", "").replace("</s>", "").strip()
         
-        if isinstance(data, list) and len(data) > 0 and "generated_text" in data[0]:
-            return data[0]["generated_text"]
-        return str(data)
     except Exception as e:
-        print(f"\n[EXCEPTION] {str(e)}\n")
-        return f"❌ HF API Error: {str(e)}. Please check your HF_TOKEN."
-
+        print(f"\n[API EXCEPTION] {str(e)}\n")
+        return "" # Safe fallback triggers programmatic error routing rather than crashing UI
 
 
 # ── Agent instances (singletons, created once at startup) ─────────────────────