Update app.py

app.py

@@ -472,11 +472,11 @@ def get_response_from_excel(query, model, context, num_calls=3, temperature=0.2)
         logging.info("Finished generating response for Excel data")
 
 def get_response_from_llama(query, model, selected_docs, file_type, num_calls=1, temperature=0.2):
-    logging.info(f"Starting get_response_from_llama with query: {query}, model: {model}, file_type: {file_type}")
+    logging.info(f"Getting response from Llama using model: {model}")
     
+    # Initialize the Hugging Face client
     client = InferenceClient(model, token=huggingface_token)
-    logging.info("InferenceClient initialized")
- 
+    
     if file_type == "excel":
         # Excel functionality
         system_instruction = """You are a highly specialized Python programmer with deep expertise in data analysis and visualization using Excel spreadsheets. 
@@ -492,11 +492,8 @@ def get_response_from_llama(query, model, selected_docs, file_type, num_calls=1,
         relevant_docs = retriever.get_relevant_documents(query)
         context = "\n".join([doc.page_content for doc in relevant_docs if doc.metadata["source"] in selected_docs])
         
-        messages = [
-            {"role": "system", "content": system_instruction},
-            {"role": "user", "content": f"Based on the following data extracted from Excel spreadsheets:\n{context}\n\nPlease provide the Python code needed to execute the following task: '{query}'. Ensure that the code is derived directly from the dataset. If a chart is requested, use the matplotlib library to generate the appropriate visualization."}
-        ]
-    
+        prompt = f"{system_instruction}\n\nBased on the following data extracted from Excel spreadsheets:\n{context}\n\nPlease provide the Python code needed to execute the following task: '{query}'. Ensure that the code is derived directly from the dataset. If a chart is requested, use the matplotlib library to generate the appropriate visualization."
+        
     elif file_type == "pdf":
         # PDF functionality
         embed = get_embeddings()
@@ -511,41 +508,32 @@ def get_response_from_llama(query, model, selected_docs, file_type, num_calls=1,
         Your goal is to provide accurate, detailed, and precise summaries based on the context provided. 
         Avoid making assumptions or adding information that is not explicitly supported by the context from the PDF documents."""
 
-        messages = [
-            {"role": "system", "content": system_instruction},
-            {"role": "user", "content": f"Using the following context from the PDF documents:\n{context_str}\n\nPlease generate a step-by-step reasoning before arriving at a comprehensive and accurate summary addressing the following question: '{query}'. Ensure your response is strictly based on the provided context, highlighting key metrics, trends, and significant details relevant to the query. Avoid any speculative or unverified information."}
-        ]
+        prompt = f"{system_instruction}\n\nUsing the following context from the PDF documents:\n{context_str}\n\nPlease generate a step-by-step reasoning before arriving at a comprehensive and accurate summary addressing the following question: '{query}'. Ensure your response is strictly based on the provided context, highlighting key metrics, trends, and significant details relevant to the query. Avoid any speculative or unverified information."
 
     else:
         raise ValueError("Invalid file type. Use 'excel' or 'pdf'.")
-      
-#    logging.info(f"Prepared messages: {messages}")
- 
+
     full_response = ""
-    for i in range(num_calls):
-        logging.info(f"Starting API call {i+1}/{num_calls}")
+    for _ in range(num_calls):
         try:
-            for message in client.chat.completion(
-                messages=messages,
-                max_tokens=2048,
+            # Generate content with streaming enabled
+            for response in client.text_generation(
+                prompt=prompt,
+                max_new_tokens=2000,
                 temperature=temperature,
-                stream=True,   
+                stream=True,
             ):
-                logging.debug(f"Received message chunk: {message}")
-                if message.choices and message.choices[0].delta and message.choices[0].delta.content:
-                    chunk = message.choices[0].delta.content
+                if response.token.text:
+                    chunk = response.token.text
                     full_response += chunk
-                    logging.debug(f"Accumulated response length: {len(full_response)}")
-                    yield full_response
+                    yield full_response  # Yield the accumulated response so far
         except Exception as e:
-            logging.error(f"Error during API call {i+1}: {str(e)}")
+            logging.error(f"Error during API call: {str(e)}")
             yield f"An error occurred with the Llama model: {str(e)}. Please try again."
 
     if not full_response:
         logging.warning("No response generated from the Llama model")
         yield "No response generated from the Llama model."
-    else:
-        logging.info(f"Final response length: {len(full_response)}") 
     
 # Modify the existing respond function to handle both PDF and web search
 def respond(message, history, model, temperature, num_calls, use_web_search, selected_docs):