Update app.py

app.py

@@ -337,16 +337,16 @@ def get_response_from_pdf(query, model, num_calls=3, temperature=0.2):
     relevant_docs = retriever.get_relevant_documents(query)
     context_str = "\n".join([doc.page_content for doc in relevant_docs])
 
-    prompt = f"""Using the following context from the PDF documents:
-{context_str}
-Write a detailed and complete response that answers the following user question: '{query}'"""
-
     if model == "@cf/meta/llama-3.1-8b-instruct":
         # Use Cloudflare API with the retrieved context
-        for response in get_response_from_cloudflare(prompt, context_str, query, num_calls, temperature):
+        for response in get_response_from_cloudflare(prompt="", context=context_str, query=query, num_calls=num_calls, temperature=temperature):
             yield response
     else:
         # Use Hugging Face API
+        prompt = f"""Using the following context from the PDF documents:
+{context_str}
+Write a detailed and complete response that answers the following user question: '{query}'"""
+        
         client = InferenceClient(model, token=huggingface_token)
         
         response = ""