Update app.py

app.py

@@ -18,6 +18,10 @@ from langchain_community.llms import HuggingFaceHub
 from langchain_core.documents import Document
 from sentence_transformers import SentenceTransformer
 from llama_parse import LlamaParse
+import logging
+
+# Set up logging
+logging.basicConfig(level=logging.DEBUG, format='%(asctime)s - %(levelname)s - %(message)s')
 
 # Environment variables and configurations
 huggingface_token = os.environ.get("HUGGINGFACE_TOKEN")
@@ -92,51 +96,73 @@ def get_model(temperature, top_p, repetition_penalty):
     )
 
 def duckduckgo_search(query):
+    logging.debug(f"Performing DuckDuckGo search for query: {query}")
     with DDGS() as ddgs:
-        results = ddgs.text(query, max_results=5)
+        results = list(ddgs.text(query, max_results=5))
+    logging.debug(f"Search returned {len(results)} results")
     return results
 
 def get_response_with_search(query, temperature, top_p, repetition_penalty, use_pdf=False):
+    logging.debug(f"Getting response for query: {query}")
+    logging.debug(f"Parameters: temperature={temperature}, top_p={top_p}, repetition_penalty={repetition_penalty}, use_pdf={use_pdf}")
+    
     model = get_model(temperature, top_p, repetition_penalty)
     embed = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
 
     if use_pdf and os.path.exists("faiss_database"):
+        logging.debug("Using PDF database for context")
         database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
         retriever = database.as_retriever()
         relevant_docs = retriever.get_relevant_documents(query)
         context = "\n".join([f"Content: {doc.page_content}\nSource: {doc.metadata['source']}\n" for doc in relevant_docs])
     else:
+        logging.debug("Using web search for context")
         search_results = duckduckgo_search(query)
         context = "\n".join(f"{result['title']}\n{result['body']}\nSource: {result['href']}\n" 
                             for result in search_results if 'body' in result)
+    
+    logging.debug(f"Context generated. Length: {len(context)} characters")
 
-     prompt = f"""<s>[INST] Using the following context:
+    prompt = f"""<s>[INST] Using the following context:
 {context}
 Write a detailed and complete research document that fulfills the following user request: '{query}'
 After the main content, provide a list of sources used in your response, prefixed with 'Sources:'. 
 Do not include any part of these instructions in your response. [/INST]"""
 
+    logging.debug("Sending prompt to model")
     response = model(prompt)
+    logging.debug(f"Received response from model. Length: {len(response)} characters")
     
     main_content, sources = split_response(response)
     
+    logging.debug(f"Split response. Main content length: {len(main_content)}, Sources length: {len(sources)}")
     return main_content, sources
 
 def split_response(response):
+    logging.debug("Splitting response")
+    logging.debug(f"Original response: {response[:100]}...") # Log first 100 characters
+    
     # Remove any remaining instruction text
     response = re.sub(r'\[/?INST\]', '', response)
     response = re.sub(r'~~.*?~~', '', response, flags=re.DOTALL)
     
+    logging.debug(f"After removing instructions: {response[:100]}...") # Log first 100 characters
+    
     # Split the response into main content and sources
     parts = response.split("Sources:", 1)
     main_content = parts[0].strip()
     sources = parts[1].strip() if len(parts) > 1 else ""
     
+    logging.debug(f"Main content starts with: {main_content[:100]}...") # Log first 100 characters
+    logging.debug(f"Sources: {sources[:100]}...") # Log first 100 characters
+    
     return main_content, sources
 
 def chatbot_interface(message, history, temperature, top_p, repetition_penalty, use_pdf):
+    logging.debug(f"Chatbot interface called with message: {message}")
     main_content, sources = get_response_with_search(message, temperature, top_p, repetition_penalty, use_pdf)
     formatted_response = f"{main_content}\n\nSources:\n{sources}"
+    logging.debug(f"Formatted response: {formatted_response[:100]}...") # Log first 100 characters
     return formatted_response
 
 # Gradio interface