Update app.py

app.py

@@ -279,17 +279,15 @@ def get_web_search_database():
     embed = get_embeddings()
     temp_dir = tempfile.mkdtemp()
     
-    # Create a dummy document to initialize the database
-    dummy_doc = WebSearchDocument(page_content="Dummy content", metadata={"source": "dummy"})
-    
     try:
-        database = WebSearchFAISS.from_documents([dummy_doc], embed)
-        # Remove the dummy document after initialization
-        database.delete([dummy_doc.metadata["source"]])
+        # Initialize with an empty list of documents
+        database = WebSearchFAISS.from_documents([], embed)
+        logging.info("Successfully initialized empty WebSearchFAISS database")
     except Exception as e:
         logging.error(f"Error initializing WebSearchFAISS: {str(e)}", exc_info=True)
         # If initialization fails, create an empty database
-        database = WebSearchFAISS(embed.embed_query, None, {}, {}, None)
+        database = WebSearchFAISS(embed, None, {}, {}, None)
+        logging.info("Created empty WebSearchFAISS database manually")
     
     return database, temp_dir
 
@@ -313,7 +311,14 @@ def duckduckgo_search(query):
         
         logging.info(f"Number of documents created: {len(documents)}")
         if documents:
-            database.add_documents(documents)
+            try:
+                database.add_documents(documents)
+                logging.info(f"Successfully added {len(documents)} documents to the database")
+            except Exception as e:
+                logging.error(f"Error adding documents to database: {str(e)}", exc_info=True)
+                # If adding documents fails, create a new database with these documents
+                database = WebSearchFAISS.from_documents(documents, get_embeddings())
+                logging.info("Created new WebSearchFAISS database with search results")
         return database, temp_dir, results
     except Exception as e:
         logging.error(f"Error in duckduckgo_search: {str(e)}", exc_info=True)
@@ -364,7 +369,7 @@ def retry_last_response(history, use_web_search, model, temperature, num_calls):
     
     return chatbot_interface(last_user_msg, history, use_web_search, model, temperature, num_calls)
 
-def respond(message: str, history: List[Tuple[str, str]], model: str, temperature: float, num_calls: int, use_web_search: bool, selected_docs: List[str]) -> str:
+def respond(message, history, model, temperature, num_calls, use_web_search, selected_docs):
     logging.info(f"User Query: {message}")
     logging.info(f"Model Used: {model}")
     logging.info(f"Search Type: {'Web Search' if use_web_search else 'PDF Search'}")
@@ -373,24 +378,28 @@ def respond(message: str, history: List[Tuple[str, str]], model: str, temperatur
     try:
         if use_web_search:
             database, temp_dir, search_results = duckduckgo_search(message)
-            if database is None:
-                yield "I'm sorry, but I encountered an error while searching. Please try again later."
+            if database is None or not search_results:
+                yield "I'm sorry, but I couldn't find any search results for your query. Could you please rephrase or ask a different question?"
                 return
             
-            context = retrieve_web_search_results(database, message)
-            logging.info(f"Retrieved context length: {len(context)}")
-            
-            if model == "@cf/meta/llama-3.1-8b-instruct":
-                # Use Cloudflare API
-                for partial_response in get_response_from_cloudflare(prompt="", context=context, query=message, num_calls=num_calls, temperature=temperature, search_type="web"):
-                    logging.debug(f"Partial response: {partial_response[:100]}...")  # Log first 100 chars
-                    yield partial_response
-            else:
-                # Use Hugging Face API
-                for main_content, sources in get_response_with_search(message, model, num_calls=num_calls, temperature=temperature):
-                    response = f"{main_content}\n\n{sources}"
-                    logging.debug(f"Response: {response[:100]}...")  # Log first 100 chars
-                    yield response
+            try:
+                context = retrieve_web_search_results(database, message)
+                logging.info(f"Retrieved context length: {len(context)}")
+                
+                if model == "@cf/meta/llama-3.1-8b-instruct":
+                    # Use Cloudflare API
+                    for partial_response in get_response_from_cloudflare(prompt="", context=context, query=message, num_calls=num_calls, temperature=temperature, search_type="web"):
+                        logging.debug(f"Partial response: {partial_response[:100]}...")  # Log first 100 chars
+                        yield partial_response
+                else:
+                    # Use Hugging Face API
+                    for main_content, sources in get_response_with_search(message, model, num_calls=num_calls, temperature=temperature):
+                        response = f"{main_content}\n\n{sources}"
+                        logging.debug(f"Response: {response[:100]}...")  # Log first 100 chars
+                        yield response
+            finally:
+                # Clean up the temporary database
+                cleanup_web_search_database(temp_dir)
         else:
             # PDF search logic
             embed = get_embeddings()