Update app.py

app.py

@@ -55,18 +55,29 @@ Regole:
 """
 
 # Funzione per inviare la richiesta al modello Hugging Face
-def generate_response(message, max_tokens, temperature):
+async def generate_response(message, max_tokens, temperature):
     system_message = create_system_message(rdf_context)
-    full_prompt = f"{system_message}\nUtente: {message}\nAssistente:"
+    messages = [
+        {"role": "system", "content": system_message},
+        {"role": "user", "content": message}
+    ]
 
     try:
-        # Usa il metodo text_generation per invocare il modello Hugging Face
-        response = client.text_generation(
+        # Usa il metodo chat.completions.create per lo streaming dei risultati
+        stream = client.chat.completions.create(
             model="Qwen/Qwen2.5-72B-Instruct",
-            inputs=full_prompt,
-            max_new_tokens=max_tokens,
-            temperature=temperature
+            messages=messages,
+            temperature=temperature,
+            max_tokens=max_tokens,
+            top_p=0.7,
+            stream=True
         )
+
+        response = ""
+        for chunk in stream:
+            if "choices" in chunk and len(chunk["choices"]) > 0:
+                response += chunk["choices"][0]["delta"]["content"]
+
         return response.strip()
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"Errore nell'elaborazione: {str(e)}")
@@ -75,7 +86,7 @@ def generate_response(message, max_tokens, temperature):
 @app.post("/generate-query/")
 async def generate_query(request: QueryRequest):
     # Genera risposta
-    response = generate_response(request.message, request.max_tokens, request.temperature)
+    response = await generate_response(request.message, request.max_tokens, request.temperature)
 
     # Valida la query se possibile
     if response.startswith("SELECT") or response.startswith("ASK"):