Update app.py

app.py

@@ -472,10 +472,10 @@ 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"Getting response from Llama using model: {model}")
+    logging.info(f"Starting get_response_from_llama with query: {query}, model: {model}, file_type: {file_type}")
     
-    # Initialize the Hugging Face client
     client = InferenceClient(model, token=huggingface_token)
+    logging.info("InferenceClient initialized")
     
     if file_type == "excel":
         # Excel functionality
@@ -518,27 +518,35 @@ def get_response_from_llama(query, model, selected_docs, file_type, num_calls=1,
 
     else:
         raise ValueError("Invalid file type. Use 'excel' or 'pdf'.")
-
+      
+    logging.info(f"Prepared messages: {messages}")
+    
     full_response = ""
-    for _ in range(num_calls):
+    for i in range(num_calls):
+        logging.info(f"Starting API call {i+1}/{num_calls}")
         try:
-            # Generate content with streaming enabled
             for message in client.chat.completions.create(
                 messages=messages,
                 max_tokens=2000,
                 temperature=temperature,
                 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
                     full_response += chunk
-                    yield full_response  # Yield the accumulated response so far
+                    logging.debug(f"Accumulated response length: {len(full_response)}")
+                    yield full_response
         except Exception as e:
+            logging.error(f"Error during API call {i+1}: {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):
     logging.info(f"User Query: {message}")