Update app.py

app.py

@@ -21,6 +21,7 @@ import inspect
 import logging
 import shutil
 import tempfile
+from typing import List, Tuple
 
 
 # Set up basic configuration for logging
@@ -346,7 +347,7 @@ 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, history, model, temperature, num_calls, use_web_search, selected_docs):
+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:
     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'}")
@@ -355,47 +356,59 @@ def respond(message, history, model, temperature, num_calls, use_web_search, sel
     try:
         if use_web_search:
             database, temp_dir, search_results = duckduckgo_search(message)
+            logging.info(f"Number of search results: {len(search_results)}")
+            
             if 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
             
-            for main_content, sources in get_response_with_search(message, model, num_calls=num_calls, temperature=temperature):
-                response = f"{main_content}\n\n{sources}"
-                yield response
+            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
         else:
+            # PDF search logic
             embed = get_embeddings()
-            if os.path.exists("faiss_database"):
-                database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
-                retriever = database.as_retriever(search_kwargs={"k": 20})
-                
-                # Filter relevant documents based on user selection
-                all_relevant_docs = retriever.get_relevant_documents(message)
-                relevant_docs = [doc for doc in all_relevant_docs if doc.metadata["source"] in selected_docs]
-                
-                if not relevant_docs:
-                    yield "No relevant information found in the selected documents. Please try selecting different documents or rephrasing your query."
-                    return
-
-                context_str = "\n".join([doc.page_content for doc in relevant_docs])
-            else:
-                context_str = "No documents available."
+            if not os.path.exists("faiss_database"):
                 yield "No documents available. Please upload PDF documents to answer questions."
                 return
             
+            database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
+            retriever = database.as_retriever(search_kwargs={"k": 20})
+            
+            all_relevant_docs = retriever.get_relevant_documents(message)
+            relevant_docs = [doc for doc in all_relevant_docs if doc.metadata["source"] in selected_docs]
+            
+            if not relevant_docs:
+                yield "No relevant information found in the selected documents. Please try selecting different documents or rephrasing your query."
+                return
+
+            context_str = "\n".join([doc.page_content for doc in relevant_docs])
+            logging.info(f"Context length for PDF search: {len(context_str)}")
+            
             if model == "@cf/meta/llama-3.1-8b-instruct":
                 # Use Cloudflare API
                 for partial_response in get_response_from_cloudflare(prompt="", context=context_str, query=message, num_calls=num_calls, temperature=temperature, search_type="pdf"):
-                    first_line = partial_response.split('\n')[0] if partial_response else ''
-#                   logging.info(f"Generated Response (first line): {first_line}")
+                    logging.debug(f"Partial response: {partial_response[:100]}...")  # Log first 100 chars
                     yield partial_response
             else:
                 # Use Hugging Face API
                 for partial_response in get_response_from_pdf(message, model, selected_docs, num_calls=num_calls, temperature=temperature):
-                    first_line = partial_response.split('\n')[0] if partial_response else ''
-#                    logging.info(f"Generated Response (first line): {first_line}")
+                    logging.debug(f"Partial response: {partial_response[:100]}...")  # Log first 100 chars
                     yield partial_response
+    
     except Exception as e:
-        logging.error(f"Error with {model}: {str(e)}")
+        logging.error(f"Error in respond function with {model}: {str(e)}", exc_info=True)
         if "microsoft/Phi-3-mini-4k-instruct" in model:
             logging.info("Falling back to Mistral model due to Phi-3 error")
             fallback_model = "mistralai/Mistral-7B-Instruct-v0.3"