Update app.py

app.py

@@ -1,8 +1,8 @@
-import logging
 import gradio as gr
 from transformers import pipeline
 from sentence_transformers import SentenceTransformer, util
 import PyPDF2
+import datetime
 import os
 
 # Load models
@@ -19,7 +19,7 @@ def extract_text_from_pdf(file_path):
     return text
 
 # Find the most relevant section in the document
-def find_relevant_section(query, sections, section_embeddings):
+def find_relevant_section(query, sections, section_embeddings, log_messages):
     stopwords = {"and", "the", "is", "for", "to", "a", "an", "of", "in", "on", "at", "with", "by", "it", "as", "so", "what"}
     
     # Semantic search
@@ -31,39 +31,40 @@ def find_relevant_section(query, sections, section_embeddings):
     
     SIMILARITY_THRESHOLD = 0.4
     if similarity_score >= SIMILARITY_THRESHOLD:
-        log_message(f"Found relevant section using embeddings for query: {query}")
-        return best_section
+        log_messages = log_message(f"Found relevant section using embeddings for query: {query}", log_messages)
+        return best_section, log_messages
     
-    log_message(f"Low similarity ({similarity_score}). Falling back to keyword search.")
+    log_messages = log_message(f"Low similarity ({similarity_score}). Falling back to keyword search.", log_messages)
     
     # Keyword-based fallback search with stopword filtering
-    query_words = {word for word in sections:
+    query_words = {word for word in query.lower().split() if word not in stopwords}  # Corrected line
+    for section in sections:
         section_words = {word for word in section.lower().split() if word not in stopwords}
         common_words = query_words.intersection(section_words)
         if len(common_words) >= 2:
-            log_message(f"Keyword match found for query: {query} with common words: {common_words}")
-            return section
+            log_messages = log_message(f"Keyword match found for query: {query} with common words: {common_words}", log_messages)
+            return section, log_messages
     
-    log_message(f"No good keyword match found. Returning default fallback response.")
-    return "I don’t have enough information to answer that."
+    log_messages = log_message(f"No good keyword match found. Returning default fallback response.", log_messages)
+    return "I don’t have enough information to answer that.", log_messages
 
 # Process the uploaded file with detailed logging
-def process_file(file, state):
+def process_file(file, state, log_messages):
     if file is None:
-        log_message("No file uploaded.")
-        return [("Bot", "Please upload a file.")], state
+        log_messages = log_message("No file uploaded.", log_messages)
+        return [("Bot", "Please upload a file.")], state, log_messages
     
     file_path = file.name
     if file_path.lower().endswith(".pdf"):
-        log_message(f"Uploaded PDF file: {file_path}")
+        log_messages = log_message(f"Uploaded PDF file: {file_path}", log_messages)
         text = extract_text_from_pdf(file_path)
     elif file_path.lower().endswith(".txt"):
-        log_message(f"Uploaded TXT file: {file_path}")
+        log_messages = log_message(f"Uploaded TXT file: {file_path}", log_messages)
         with open(file_path, 'r', encoding='utf-8') as f:
             text = f.read()
     else:
-        log_message(f"Unsupported file format: {file_path}")
-        return [("Bot", "Unsupported file format. Please upload a PDF or TXT file.")], state
+        log_messages = log_message(f"Unsupported file format: {file_path}", log_messages)
+        return [("Bot", "Unsupported file format. Please upload a PDF or TXT file.")], state, log_messages
     
     sections = text.split('\n\n')
     section_embeddings = embedder.encode(sections, convert_to_tensor=True)
@@ -74,27 +75,27 @@ def process_file(file, state):
     state['feedback_count'] = 0
     state['mode'] = 'waiting_for_query'
     state['chat_history'] = [("Bot", "File processed. You can now ask questions.")]
-    log_message(f"Processed file: {file_path}")
-    return state['chat_history'], state
+    log_messages = log_message(f"Processed file: {file_path}", log_messages)
+    return state['chat_history'], state, log_messages
 
 # Handle user input (queries and feedback)
-def handle_input(user_input, state):
+def handle_input(user_input, state, log_messages):
     if state['mode'] == 'waiting_for_upload':
         state['chat_history'].append(("Bot", "Please upload a file first."))
-        log_message("User attempted to interact without uploading a file.")
-        return state['chat_history'], state
+        log_messages = log_message("User attempted to interact without uploading a file.", log_messages)
+        return state['chat_history'], state, log_messages
     elif state['mode'] == 'waiting_for_query':
         if user_input.lower() == "exit":
-            log_message("User entered 'exit'. Ending session.")
+            log_messages = log_message("User entered 'exit'. Ending session.", log_messages)
             state['mode'] = 'exited'
             state['chat_history'].append(("User", "exit"))
             state['chat_history'].append(("Bot", "Session ended. You can download the log file."))
-            return state['chat_history'], state
+            return state['chat_history'], state, log_messages
         
         query = user_input
         state['current_query'] = query
         state['feedback_count'] = 0
-        context = find_relevant_section(query, state['sections'], state['section_embeddings'])
+        context, log_messages = find_relevant_section(query, state['sections'], state['section_embeddings'], log_messages)
         if context == "I don’t have enough information to answer that.":
             answer = context
         else:
@@ -105,45 +106,45 @@ def handle_input(user_input, state):
         state['chat_history'].append(("User", query))
         state['chat_history'].append(("Bot", f"Answer: {answer}\nPlease provide feedback: good, too vague, not helpful."))
         # Log the query and initial answer here:
-        log_message(f"Query: {query}, Answer: {answer}")
+        log_messages = log_message(f"Query: {query}, Answer: {answer}", log_messages)
     elif state['mode'] == 'waiting_for_feedback':
         if user_input.lower() == "exit":
-            log_message("User entered 'exit'. Ending session.")
+            log_messages = log_message("User entered 'exit'. Ending session.", log_messages)
             state['mode'] = 'exited'
             state['chat_history'].append(("User", "exit"))
             state['chat_history'].append(("Bot", "Session ended. You can download the log file."))
-            return state['chat_history'], state
+            return state['chat_history'], state, log_messages
         
         feedback = user_input.lower()
         state['chat_history'].append(("User", feedback))
-        log_message(f"Feedback: {feedback}")
+        log_messages = log_message(f"Feedback: {feedback}", log_messages)
         if feedback == "good" or state['feedback_count'] >= 2:
             state['mode'] = 'waiting_for_query'
             if feedback == "good":
                 state['chat_history'].append(("Bot", "Thank you for your feedback. You can ask another question."))
-                log_message("Feedback accepted as 'good'. Waiting for next query.")
+                log_messages = log_message("Feedback accepted as 'good'. Waiting for next query.", log_messages)
             else:
                 state['chat_history'].append(("Bot", "Maximum feedback iterations reached. You can ask another question."))
-                log_message("Max feedback iterations reached. Waiting for next query.")
+                log_messages = log_message("Max feedback iterations reached. Waiting for next query.", log_messages)
         else:
             query = state['current_query']
-            context = find_relevant_section(query, state['sections'], state['section_embeddings'])
+            context, log_messages = find_relevant_section(query, state['sections'], state['section_embeddings'], log_messages)
             if feedback == "too vague":
                 adjusted_answer = f"{state['last_answer']}\n\n(More details:\n{context[:500]}...)"
             elif feedback == "not helpful":
                 adjusted_answer = qa_model(question=query + " Please provide more detailed information with examples.", context=context)['answer']
             else:
                 state['chat_history'].append(("Bot", "Please provide valid feedback: good, too vague, not helpful."))
-                log_message(f"Invalid feedback received: {feedback}")
-                return state['chat_history'], state
+                log_messages = log_message(f"Invalid feedback received: {feedback}", log_messages)
+                return state['chat_history'], state, log_messages
             state['last_answer'] = adjusted_answer
             state['feedback_count'] += 1
             state['chat_history'].append(("Bot", f"Updated answer: {adjusted_answer}\nPlease provide feedback: good, too vague, not helpful."))
-            log_message(f"Adjusted answer: {adjusted_answer}")
+            log_messages = log_message(f"Adjusted answer: {adjusted_answer}", log_messages)
     elif state['mode'] == 'exited':
         state['chat_history'].append(("Bot", "Session is over. Please download the log."))
-        log_message("User interacted after exiting.")
-    return state['chat_history'], state
+        log_messages = log_message("User interacted after exiting.", log_messages)
+    return state['chat_history'], state, log_messages
 
 # Initial state
 initial_state = {
@@ -154,20 +155,23 @@ initial_state = {
     'feedback_count': 0,
     'mode': 'waiting_for_upload',
     'chat_history': [("Bot", "Please upload a PDF or TXT file to start.")],
-    'last_answer': None,
-    'log_messages':# Initialize an empty list for log messages
+    'last_answer': None
 }
 
+# Initialize log_messages outside initial_state
+log_messages =
+
 # Logging function to store messages in memory
-def log_message(message):
+def log_message(message, log_messages):
     timestamp = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
     log_entry = f"{timestamp} - {message}"
-    initial_state['log_messages'].append(log_entry)
+    log_messages.append(log_entry)
+    return log_messages
 
 # Function to save logs to file
-def save_logs_to_file():
+def save_logs_to_file(log_messages):
     with open("support_bot_log.txt", "w") as log_file:
-        for log_message in initial_state['log_messages']:
+        for log_message in log_messages:
             log_file.write(log_message + "\n")
 
 # Gradio interface
@@ -180,21 +184,21 @@ with gr.Blocks() as demo:
     log_file = gr.File(label="Download Log File")  # Changed: No initial value
 
     # Process file upload
-    file_upload.upload(process_file, inputs=[file_upload, state], outputs=[chat, state])
+    file_upload.upload(process_file, inputs=[file_upload, state, gr.State(log_messages)], outputs=[chat, state, gr.State(log_messages)])
 
     # Handle user input and clear the textbox
-    submit_btn.click(handle_input, inputs=[user_input, state], outputs=[chat, state]).then(lambda: "", None, user_input)
+    submit_btn.click(handle_input, inputs=[user_input, state, gr.State(log_messages)], outputs=[chat, state, gr.State(log_messages)]).then(lambda: "", None, user_input)
 
     # Update the log file just before download
-    log_file.click(save_logs_to_file,, [log_file])  # Trigger save on click
+    log_file.click(save_logs_to_file, inputs=[gr.State(log_messages)], outputs=[log_file])  # Trigger save on click
     
     # Also save logs when user exits
     user_input.submit(
-        lambda user_input, state: (
-            save_logs_to_file() if user_input.lower() == "exit" else None,
+        lambda user_input, state, log_messages: (
+            save_logs_to_file(log_messages) if user_input.lower() == "exit" else None,
             state
         ),
-        [user_input, state],
+        [user_input, state, gr.State(log_messages)],
         [log_file, state]
     )