changes v10

question.py

@@ -62,28 +62,30 @@ async def ask_question(input: QuestionInput):
 
     # Call HF Inference API manually
     hf_api_url = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.3"
+    prompt = f"<s>[INST] Use the context below to answer the question. If not enough information is available, say 'Cannot answer'.\n\nContext:\n{context}\n\nQuestion: {question} [/INST]"
     headers = {
         "Authorization": f"Bearer {HF_TOKEN}",
         "Content-Type": "application/json"
     }
 
     payload = {
-        "inputs": {
-            "past_user_inputs": [],
-            "generated_responses": [],
-            "text": f"You are an assistant that answers questions based on recent news.\n\nContext:\n{context}\n\nQuestion: {question}"
-        },
+        "inputs": prompt,
         "parameters": {
             "max_new_tokens": 256,
             "temperature": 0.7
         }
     }
 
+
     try:
         response = requests.post(hf_api_url, headers=headers, data=json.dumps(payload), timeout=30)
         response.raise_for_status()
         hf_response = response.json()
-        answer = hf_response.get("generated_text", "Cannot answer.")
+        if isinstance(hf_response, list) and len(hf_response) > 0:
+            answer = hf_response[0].get("generated_text", "").strip()
+        else:
+            answer = "Cannot answer – model did not return a valid response."
+
     except Exception as e:
         return {"error": f"Hugging Face API error: {str(e)}"}