AIToyBot / app.py
masadonline's picture
Update app.py
44b1add verified
raw
history blame
33.5 kB
import streamlit as st
import os
import time
from datetime import datetime, timezone
import json
import PyPDF2
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np
from twilio.rest import Client
from groq import Groq
import re # Import re module
# --- Page Configuration ---
st.set_page_config(page_title="RAG Customer Support Chatbot", layout="wide")
# --- Default Configurations & File Paths ---
DEFAULT_TWILIO_ACCOUNT_SID_FALLBACK = "" # Fallback if secret "TWILIO_SID" is not found
DEFAULT_TWILIO_AUTH_TOKEN_FALLBACK = "" # Fallback if secret "TWILIO_TOKEN" is not found
DEFAULT_GROQ_API_KEY_FALLBACK = "" # Fallback if secret "GROQ_API_KEY" is not found
DEFAULT_TWILIO_CONVERSATION_SERVICE_SID = "" 
DEFAULT_TWILIO_BOT_WHATSAPP_IDENTITY = st.secrets.get("TWILIO_PHONE_NUMBER")#"whatsapp:+14155238886" # Twilio Sandbox default
DEFAULT_EMBEDDING_MODEL_NAME = "sentence-transformers/all-MiniLM-L6-v2"
DEFAULT_POLLING_INTERVAL_S = 30
DOCS_FOLDER = "docs/"
CUSTOMER_ORDERS_FILE = os.path.join(DOCS_FOLDER, "CustomerOrders.json")
PRODUCTS_FILE = os.path.join(DOCS_FOLDER, "Products.json")
POLICY_PDF_FILE = os.path.join(DOCS_FOLDER, "ProductReturnPolicy.pdf")
FAQ_PDF_FILE = os.path.join(DOCS_FOLDER, "FAQ.pdf")
# --- Application Secrets Configuration ---
# These are the primary keys fetched from st.secrets as per user request
APP_TWILIO_ACCOUNT_SID = st.secrets.get("TWILIO_ACCOUNT_SID")
APP_TWILIO_AUTH_TOKEN = st.secrets.get("TWILIO_AUTH_TOKEN")
APP_GROQ_API_KEY = st.secrets.get("GROQ_API_KEY")
# Other secrets with fallback to defaults/sidebar input (if secrets not found)
APP_TWILIO_CONVERSATION_SERVICE_SID_SECRET = st.secrets.get("TWILIO_CONVERSATION_SERVICE_SID")
APP_TWILIO_BOT_WHATSAPP_IDENTITY_SECRET = st.secrets.get("TWILIO_BOT_WHATSAPP_IDENTITY")
# --- RAG Processing Utilities ---
def load_json_data(file_path):
    """Loads data from a JSON file."""
    try:
        with open(file_path, 'r', encoding='utf-8') as f:
            data = json.load(f)
        return data
    except FileNotFoundError:
        st.error(f"Error: JSON file not found at {file_path}")
        return None
    except json.JSONDecodeError:
        st.error(f"Error: Could not decode JSON from {file_path}")
        return None
    except Exception as e:
        st.error(f"An unexpected error occurred while loading {file_path}: {e}")
        return None
def load_pdf_data(file_path):
    """Extracts text from a PDF file, page by page."""
    try:
        with open(file_path, 'rb') as f:
            reader = PyPDF2.PdfReader(f)
            text_pages = []
            for page_num in range(len(reader.pages)):
                page = reader.pages[page_num]
                text_pages.append(page.extract_text() or "")
        return text_pages
    except FileNotFoundError:
        st.error(f"Error: PDF file not found at {file_path}")
        return []
    except Exception as e:
        st.error(f"An error occurred while processing PDF {file_path}: {e}")
        return []
def chunk_text(text_pages, chunk_size=1000, chunk_overlap=200):
    """Chunks text from PDF pages into smaller, overlapping pieces."""
    full_text = "\n".join(text_pages)
    if not full_text.strip():
        return []
    chunks = []
    start = 0
    while start < len(full_text):
        end = start + chunk_size
        chunks.append(full_text[start:end])
        if end >= len(full_text):
            break
        start += (chunk_size - chunk_overlap)
        if start >= len(full_text):
            break
    return [chunk for chunk in chunks if chunk.strip()]
@st.cache_resource(show_spinner="Initializing embedding model...")
def initialize_embedding_model(model_name=DEFAULT_EMBEDDING_MODEL_NAME):
    """Initializes and returns a SentenceTransformer model."""
    try:
        model = SentenceTransformer(model_name)
        return model
    except Exception as e:
        st.error(f"Error initializing embedding model '{model_name}': {e}")
        return None
@st.cache_resource(show_spinner="Building FAISS index for PDF documents...")
def create_faiss_index(_text_chunks, _embedding_model):
    """Creates a FAISS index from text chunks and an embedding model."""
    if not _text_chunks or _embedding_model is None:
        st.warning("Cannot create FAISS index: No text chunks or embedding model available.")
        return None, []
    try:
        valid_chunks = [str(chunk) for chunk in _text_chunks if chunk and isinstance(chunk, str) and chunk.strip()]
        if not valid_chunks:
            st.warning("No valid text chunks to embed for FAISS index.")
            return None, []
        embeddings = _embedding_model.encode(valid_chunks, convert_to_tensor=False)
        if embeddings.ndim == 1:
            embeddings = embeddings.reshape(1, -1)
        if embeddings.shape[0] == 0:
             st.warning("No embeddings were generated for FAISS index.")
             return None, []
        dimension = embeddings.shape[1]
        index = faiss.IndexFlatL2(dimension)
        index.add(np.array(embeddings, dtype=np.float32))
        return index, valid_chunks
    except Exception as e:
        st.error(f"Error creating FAISS index: {e}")
        return None, []
def search_faiss_index(index, query_text, embedding_model, indexed_chunks, k=3):
    """Searches the FAISS index and returns top_k relevant chunk texts."""
    if index is None or embedding_model is None or not query_text:
        return []
    try:
        query_embedding = embedding_model.encode([query_text], convert_to_tensor=False)
        if query_embedding.ndim == 1:
            query_embedding = query_embedding.reshape(1, -1)
        distances, indices = index.search(np.array(query_embedding, dtype=np.float32), k)
        results = []
        for i in range(len(indices[0])):
            idx = indices[0][i]
            if 0 <= idx < len(indexed_chunks):
                results.append(indexed_chunks[idx])
        return results
    except Exception as e:
        st.error(f"Error searching FAISS index: {e}")
        return []
def get_order_details(order_id, customer_orders_data):
    """Retrieves order details for a given order_id."""
    if not customer_orders_data:
        return "Customer order data is not loaded."
    for order in customer_orders_data:
        if order.get("order_id") == order_id:
            return json.dumps(order, indent=2)
    return f"No order found with ID: {order_id}."
def get_product_info(query, products_data):
    """Retrieves product information based on a query."""
    if not products_data:
        return "Product data is not loaded."
    query_lower = query.lower()
    found_products = []
    for product in products_data:
        if query_lower in (product.get("name", "").lower()) or \
           query_lower in (product.get("description", "").lower()) or \
           query_lower == (product.get("product_id", "").lower()):
            found_products.append(product)
    if found_products:
        return json.dumps(found_products, indent=2)
    return f"No product information found matching your query: '{query}'."
# --- LLM Operations ---
@st.cache_data(show_spinner="Generating response with LLaMA3...")
def generate_response_groq(_groq_client, query, context, model="llama3-8b-8192"):
    """Generates a response using GROQ LLaMA3 API."""
    if not _groq_client:
        return "GROQ client not initialized. Please check API key."
    if not query:
        return "Query is empty."
    prompt = f"""You are a helpful customer support assistant.
Use the following context to answer the user's question.
If the context doesn't contain the answer, state that you don't have enough information.
Do not make up information. Be concise and polite.
Context:
{context}
User Question: {query}
Assistant Answer:
"""
    try:
        chat_completion = _groq_client.chat.completions.create(
            messages=[
                {"role": "system", "content": "You are a helpful customer support assistant."},
                {"role": "user", "content": prompt}
            ],
            model=model, temperature=0.7, max_tokens=1024, top_p=1
        )
        response = chat_completion.choices[0].message.content
        return response
    except Exception as e:
        st.error(f"Error calling GROQ API: {e}")
        return "Sorry, I encountered an error while trying to generate a response."
def initialize_groq_client(api_key_val):
    """Initializes the GROQ client."""
    if not api_key_val: # Changed parameter name to avoid conflict
        st.warning("GROQ API Key is missing.")
        return None
    try:
        client = Groq(api_key=api_key_val)
        return client
    except Exception as e:
        st.error(f"Failed to initialize GROQ client: {e}")
        return None
# --- Twilio Operations ---
def initialize_twilio_client(acc_sid, auth_tkn): # Changed parameter names
    """Initializes the Twilio client."""
    if not acc_sid or not auth_tkn:
        st.warning("Twilio Account SID or Auth Token is missing.")
        return None
    try:
        client = Client(acc_sid, auth_tkn)
        return client
    except Exception as e:
        st.error(f"Failed to initialize Twilio client: {e}")
        return None
def get_new_whatsapp_messages(twilio_client, conversation_service_sid_val, bot_start_time_utc, # Renamed
                              processed_message_sids, bot_whatsapp_identity_val): # Renamed
    """Fetches new, unanswered WhatsApp messages from Twilio Conversations."""
    if not twilio_client:
        st.warning("Twilio client not initialized.")
        return []
    if not conversation_service_sid_val:
        st.warning("Twilio Conversation Service SID not provided.")
        return []
    new_messages_to_process = []
    try:
        conversations = twilio_client.conversations.v1 \
            .services(conversation_service_sid_val) \
            .conversations \
            .list(limit=50)
        for conv in conversations:
            if conv.date_updated and conv.date_updated > bot_start_time_utc:
                messages = twilio_client.conversations.v1 \
                    .services(conversation_service_sid_val) \
                    .conversations(conv.sid) \
                    .messages \
                    .list(order='desc', limit=10)
                for msg in messages:
                    if msg.sid in processed_message_sids:
                        continue
                    if msg.author and msg.author.lower() != bot_whatsapp_identity_val.lower() and \
                       msg.date_created and msg.date_created > bot_start_time_utc:
                        new_messages_to_process.append({
                            "conversation_sid": conv.sid, "message_sid": msg.sid,
                            "author_identity": msg.author, "message_body": msg.body,
                            "timestamp_utc": msg.date_created 
                        })
                        break 
    except Exception as e:
        st.error(f"Error fetching Twilio messages: {e}")
    return sorted(new_messages_to_process, key=lambda m: m['timestamp_utc'])
def send_whatsapp_message(twilio_client, conversation_service_sid_val, conversation_sid, message_body, bot_identity_val): # Renamed
    """Sends a message to a Twilio Conversation from the bot's identity."""
    if not twilio_client:
        st.error("Twilio client not initialized for sending message.")
        return False
    if not conversation_service_sid_val:
        st.error("Twilio Conversation Service SID not provided for sending message.")
        return False
    if not bot_identity_val:
        st.error("Bot identity not provided for sending message.")
        return False
    try:
        twilio_client.conversations.v1 \
            .services(conversation_service_sid_val) \
            .conversations(conversation_sid) \
            .messages \
            .create(author=bot_identity_val, body=message_body)
        st.success(f"Sent reply to conversation {conversation_sid}")
        return True
    except Exception as e:
        st.error(f"Error sending Twilio message to {conversation_sid}: {e}")
        return False
# --- Main Application Logic & UI ---
st.title("🤖 RAG-Based Customer Support Chatbot")
st.markdown("Powered by Streamlit, Twilio, GROQ LLaMA3, and FAISS.")
# --- Sidebar for Configurations ---
st.sidebar.title("⚙️ Configurations")
# Use APP_ prefixed variables for values from secrets, then allow manual input if not found
if APP_TWILIO_ACCOUNT_SID:
    st.sidebar.text_input("Twilio Account SID (from Secrets)", value="********" + APP_TWILIO_ACCOUNT_SID[-4:] if len(APP_TWILIO_ACCOUNT_SID) > 4 else "********", disabled=True)
    twilio_account_sid_to_use = APP_TWILIO_ACCOUNT_SID
else:
    st.sidebar.warning("Secret 'TWILIO_SID' not found.")
    twilio_account_sid_to_use = st.sidebar.text_input("Twilio Account SID (Enter Manually)", value=DEFAULT_TWILIO_ACCOUNT_SID_FALLBACK, type="password")
if APP_TWILIO_AUTH_TOKEN:
    st.sidebar.text_input("Twilio Auth Token (from Secrets)", value="********", disabled=True)
    twilio_auth_token_to_use = APP_TWILIO_AUTH_TOKEN
else:
    st.sidebar.warning("Secret 'TWILIO_TOKEN' not found.")
    twilio_auth_token_to_use = st.sidebar.text_input("Twilio Auth Token (Enter Manually)", value=DEFAULT_TWILIO_AUTH_TOKEN_FALLBACK, type="password")
if APP_GROQ_API_KEY:
    st.sidebar.text_input("GROQ API Key (from Secrets)", value="gsk_********" + APP_GROQ_API_KEY[-4:] if len(APP_GROQ_API_KEY) > 8 else "********", disabled=True)
    groq_api_key_to_use = APP_GROQ_API_KEY
else:
    st.sidebar.warning("Secret 'GROQ_API_KEY' not found.")
    groq_api_key_to_use = st.sidebar.text_input("GROQ API Key (Enter Manually)", value=DEFAULT_GROQ_API_KEY_FALLBACK, type="password")
# For other configurations that can be overridden if secrets not found or for user preference
twilio_conversation_service_sid_to_use = st.sidebar.text_input(
    "Twilio Conversation Service SID (IS...)"
    value=APP_TWILIO_CONVERSATION_SERVICE_SID_SECRET or DEFAULT_TWILIO_CONVERSATION_SERVICE_SID, 
    type="password"
    help="The SID of your Twilio Conversations Service. Can be set by 'TWILIO_CONVERSATION_SERVICE_SID' secret."
)
twilio_bot_whatsapp_identity_to_use = st.sidebar.text_input(
    "Twilio Bot WhatsApp Identity"
    value=APP_TWILIO_BOT_WHATSAPP_IDENTITY_SECRET or DEFAULT_TWILIO_BOT_WHATSAPP_IDENTITY,
    help="e.g., 'whatsapp:+1234567890'. Can be set by 'TWILIO_BOT_WHATSAPP_IDENTITY' secret."
)
embedding_model_name_to_use = st.sidebar.text_input( # Renamed
    "Embedding Model Name"
    value=DEFAULT_EMBEDDING_MODEL_NAME
)
polling_interval_to_use = st.sidebar.number_input( # Renamed
    "Twilio Polling Interval (seconds)"
    min_value=10, max_value=300
    value=DEFAULT_POLLING_INTERVAL_S, 
    step=5
)
# --- Initialize Session State ---
if "app_started" not in st.session_state: st.session_state.app_started = False
if "bot_started" not in st.session_state: st.session_state.bot_started = False
if "rag_pipeline_ready" not in st.session_state: st.session_state.rag_pipeline_ready = False
if "last_twilio_poll_time" not in st.session_state: st.session_state.last_twilio_poll_time = time.time()
if "bot_start_time_utc" not in st.session_state: st.session_state.bot_start_time_utc = None
if "processed_message_sids" not in st.session_state: st.session_state.processed_message_sids = set()
if "manual_chat_history" not in st.session_state: st.session_state.manual_chat_history = []
# --- Helper: Simple Intent Classifier ---
def simple_intent_classifier(query):
    query_lower = query.lower()
    if any(k in query_lower for k in ["order", "status", "track", "delivery"]):
        # More specific regex to find 'ORD' followed by digits (assuming order IDs are like ORD1001)
        match = re.search(r'\b(ord\d{3,})\b', query_lower) # Matches 'ord' followed by at least 3 digits, as a whole word
        if match:
            return "ORDER_STATUS", match.group(1).upper() # Return intent and extracted ID
        # Fallback if specific order ID not found but still an order-related query
        return "ORDER_STATUS", None # Indicate order status intent but no specific ID found yet
    if any(k in query_lower for k in ["product", "item", "buy", "price", "feature", "stock"]): return "PRODUCT_INFO", None
    if any(k in query_lower for k in ["return", "policy", "refund", "exchange", "faq", "question", "how to", "support"]): return "GENERAL_POLICY_FAQ", None
    return "UNKNOWN", None # Return intent and None for ID if unknown
# --- Main Application Controls ---
col1, col2, col3, col4 = st.columns(4)
with col1:
    if st.button("🚀 Start App", disabled=st.session_state.app_started, use_container_width=True):
        if not groq_api_key_to_use: # Use the correct variable
            st.error("GROQ API Key is required.")
        else:
            with st.spinner("Initializing RAG pipeline..."):
                st.session_state.embedding_model = initialize_embedding_model(embedding_model_name_to_use) # Use correct var
                st.session_state.customer_orders_data = load_json_data(CUSTOMER_ORDERS_FILE)
                st.session_state.products_data = load_json_data(PRODUCTS_FILE)
                policy_pdf_pages = load_pdf_data(POLICY_PDF_FILE)
                faq_pdf_pages = load_pdf_data(FAQ_PDF_FILE)
                all_pdf_text_pages = policy_pdf_pages + faq_pdf_pages
                st.session_state.pdf_text_chunks_raw = chunk_text(all_pdf_text_pages)
                if st.session_state.embedding_model and st.session_state.pdf_text_chunks_raw:
                    st.session_state.faiss_index_pdfs, st.session_state.indexed_pdf_chunks = \
                        create_faiss_index(st.session_state.pdf_text_chunks_raw, st.session_state.embedding_model)
                else:
                    st.session_state.faiss_index_pdfs, st.session_state.indexed_pdf_chunks = None, []
                    st.warning("FAISS index for PDFs could not be created.")
                
                st.session_state.groq_client = initialize_groq_client(groq_api_key_to_use) # Use correct var
                if st.session_state.embedding_model and st.session_state.groq_client and \
                   st.session_state.customer_orders_data and st.session_state.products_data:
                    st.session_state.rag_pipeline_ready = True
                    st.session_state.app_started = True
                    st.success("RAG Application Started!")
                    st.rerun()
                else:
                    st.error("Failed to initialize RAG pipeline. Check configurations and ensure all data files are present in 'docs/'.")
                    st.session_state.app_started = False
with col2:
    if st.button("🛑 Stop App", disabled=not st.session_state.app_started, use_container_width=True):
        keys_to_reset = ["app_started", "bot_started", "rag_pipeline_ready", "embedding_model"
                         "customer_orders_data", "products_data", "pdf_text_chunks_raw"
                         "faiss_index_pdfs", "indexed_pdf_chunks", "groq_client", "twilio_client"
                         "bot_start_time_utc", "processed_message_sids", "manual_chat_history"]
        for key in keys_to_reset:
            if key in st.session_state: del st.session_state[key]
        st.session_state.app_started = False
        st.session_state.bot_started = False
        st.session_state.rag_pipeline_ready = False
        st.session_state.processed_message_sids = set()
        st.session_state.manual_chat_history = []
        st.success("Application Stopped.")
        st.rerun()
with col3:
    if st.button("💬 Start WhatsApp Bot", disabled=not st.session_state.app_started or st.session_state.bot_started, use_container_width=True):
        if not all([twilio_account_sid_to_use, twilio_auth_token_to_use, twilio_conversation_service_sid_to_use, twilio_bot_whatsapp_identity_to_use]): # Use correct vars
            st.error("Twilio credentials, Service SID, and Bot Identity are required.")
        else:
            st.session_state.twilio_client = initialize_twilio_client(twilio_account_sid_to_use, twilio_auth_token_to_use) # Use correct vars
            if st.session_state.twilio_client:
                st.session_state.bot_started = True
                st.session_state.bot_start_time_utc = datetime.now(timezone.utc)
                st.session_state.processed_message_sids = set()
                st.session_state.last_twilio_poll_time = time.time() - polling_interval_to_use -1 # Use correct var
                st.success("WhatsApp Bot Started!")
                st.rerun()
            else:
                st.error("Failed to initialize Twilio client.")
with col4:
    if st.button("🔕 Stop WhatsApp Bot", disabled=not st.session_state.bot_started, use_container_width=True):
        st.session_state.bot_started = False
        st.info("WhatsApp Bot Stopped.")
        st.rerun()
st.divider()
# --- Manual Query Interface ---
if st.session_state.get("app_started") and st.session_state.get("rag_pipeline_ready"):
    st.subheader("💬 Manual Query")
    for chat_entry in st.session_state.manual_chat_history:
        with st.chat_message(chat_entry["role"]):
            st.markdown(chat_entry["content"])
            if "context" in chat_entry and chat_entry["context"]:
                with st.expander("Retrieved Context"):
                    try:
                        # Attempt to parse as JSON only if it looks like a JSON string
                        if isinstance(chat_entry["context"], str) and (chat_entry["context"].strip().startswith('{') or chat_entry["context"].strip().startswith('[')):
                            st.json(json.loads(chat_entry["context"]))
                        else:
                            # Otherwise, display as plain text
                            st.text(str(chat_entry["context"]))
                    except (json.JSONDecodeError, TypeError):
                        # Fallback for any other parsing errors
                        st.text(str(chat_entry["context"]))
    user_query_manual = st.chat_input("Ask a question:")
    if user_query_manual:
        st.session_state.manual_chat_history.append({"role": "user", "content": user_query_manual})
        with st.chat_message("user"): st.markdown(user_query_manual)
        with st.spinner("Thinking..."):
            intent_result = simple_intent_classifier(user_query_manual) # Get both intent and potential_id
            intent = intent_result[0]
            potential_oid_from_intent = intent_result[1] # This is the extracted ID if any
            context_for_llm, raw_context_data = "No specific context.", None
            if intent == "ORDER_STATUS":
                order_id_to_check = None
                if potential_oid_from_intent:
                    order_id_to_check = potential_oid_from_intent
                else:
                    # Fallback for edge cases, though the regex should catch most
                    words = user_query_manual.upper().split()
                    # This regex specifically looks for 'ORD' followed by digits
                    possible_match = next((w for w in words if re.match(r'ORD\d+', w)), None)
                    if possible_match:
                        order_id_to_check = possible_match
                if order_id_to_check:
                    raw_context_data = get_order_details(order_id_to_check.upper(), st.session_state.customer_orders_data)
                    context_for_llm = f"Order Details: {raw_context_data}"
                else:
                    context_for_llm = "Please provide a valid Order ID (e.g., ORD1234)."
                    raw_context_data = {"message": "Order ID needed."}
            elif intent == "PRODUCT_INFO":
                raw_context_data = get_product_info(user_query_manual, st.session_state.products_data)
                context_for_llm = f"Product Information: {raw_context_data}"
            elif intent == "GENERAL_POLICY_FAQ" or intent == "UNKNOWN":
                # ... (rest of your existing logic for these intents) ...
                if st.session_state.faiss_index_pdfs and st.session_state.embedding_model:
                    k_val = 2 if intent == "GENERAL_POLICY_FAQ" else 1
                    retrieved_chunks = search_faiss_index(st.session_state.faiss_index_pdfs, user_query_manual,
                                                          st.session_state.embedding_model, st.session_state.indexed_pdf_chunks, k=k_val)
                    if retrieved_chunks:
                        context_for_llm = "\n\n".join(retrieved_chunks)
                        raw_context_data = retrieved_chunks
                    else:
                        context_for_llm = "No specific policy/FAQ info found." if intent == "GENERAL_POLICY_FAQ" else "Could not find relevant info."
                        raw_context_data = {"message": "No relevant PDF chunks found."}
                else:
                    context_for_llm = "Policy/FAQ documents unavailable."
                    raw_context_data = {"message": "PDF index not ready."}
            llm_response = generate_response_groq(st.session_state.groq_client, user_query_manual, context_for_llm)
            with st.chat_message("assistant"):
                st.markdown(llm_response)
                if raw_context_data:
                    with st.expander("Retrieved Context"):
                        try:
                            if isinstance(raw_context_data, str) and (raw_context_data.strip().startswith('{') or raw_context_data.strip().startswith('[')):
                                st.json(json.loads(raw_context_data))
                            else:
                                st.text(str(raw_context_data))
                        except (json.JSONDecodeError, TypeError):
                            st.text(str(raw_context_data))
            st.session_state.manual_chat_history.append({"role": "assistant", "content": llm_response, "context": raw_context_data})
# --- Twilio Bot Polling Logic ---
if st.session_state.get("bot_started") and st.session_state.get("rag_pipeline_ready"):
    current_time = time.time()
    if (current_time - st.session_state.get("last_twilio_poll_time", 0)) > polling_interval_to_use: # Use correct var
        st.session_state.last_twilio_poll_time = current_time
        with st.spinner("Checking WhatsApp messages..."):
            if not st.session_state.get("twilio_client") or not twilio_conversation_service_sid_to_use or not twilio_bot_whatsapp_identity_to_use: # Use correct vars
                st.warning("Twilio client/config missing for polling.")
            else:
                new_messages = get_new_whatsapp_messages(st.session_state.twilio_client, twilio_conversation_service_sid_to_use, 
                                                         st.session_state.bot_start_time_utc, st.session_state.processed_message_sids,
                                                         twilio_bot_whatsapp_identity_to_use) # Use correct vars
                if new_messages:
                    st.info(f"Found {len(new_messages)} new WhatsApp message(s).")
                    for msg_data in new_messages:
                        user_query_whatsapp, conv_sid, msg_sid, author_id = msg_data["message_body"], msg_data["conversation_sid"], msg_data["message_sid"], msg_data["author_identity"]
                        st.write(f"Processing from {author_id} in {conv_sid}: '{user_query_whatsapp}'")
                        intent_result_whatsapp = simple_intent_classifier(user_query_whatsapp) # Use the updated classifier
                        intent_whatsapp = intent_result_whatsapp[0]
                        potential_oid_whatsapp = intent_result_whatsapp[1] # Extracted ID from intent classifier
                        context_whatsapp = "No specific context."
                        if intent_whatsapp == "ORDER_STATUS":
order_id_to_check_whatsapp = None
if potential_oid_whatsapp:
order_id_to_check_whatsapp = potential_oid_whatsapp
else:
words_whatsapp = user_query_whatsapp.upper().split()
possible_match_whatsapp = next((w for w in words_whatsapp if re.match(r'ORD\d+', w)), None)
if possible_match_whatsapp:
order_id_to_check_whatsapp = possible_match_whatsapp
if order_id_to_check_whatsapp:
                            context_whatsapp = f"Order Details: {get_order_details(order_id_to_check_whatsapp.upper(), st.session_state.customer_orders_data)}"
else:
context_whatsapp = "Please provide a valid Order ID."
                        elif intent_whatsapp == "PRODUCT_INFO":
                            context_whatsapp = f"Product Info: {get_product_info(user_query_whatsapp, st.session_state.products_data)}"
                        elif intent_whatsapp == "GENERAL_POLICY_FAQ" or intent_whatsapp == "UNKNOWN":
                            if st.session_state.faiss_index_pdfs and st.session_state.embedding_model:
                                k_val = 2 if intent_whatsapp == "GENERAL_POLICY_FAQ" else 1
                                chunks = search_faiss_index(st.session_state.faiss_index_pdfs, user_query_whatsapp, st.session_state.embedding_model, st.session_state.indexed_pdf_chunks, k=k_val)
                                context_whatsapp = "\n\n".join(chunks) if chunks else ("No policy/FAQ info." if intent_whatsapp == "GENERAL_POLICY_FAQ" else "No relevant info.")
                            else: context_whatsapp = "Policy/FAQ docs unavailable."
                        
                        response_whatsapp = generate_response_groq(st.session_state.groq_client, user_query_whatsapp, context_whatsapp)
                        if send_whatsapp_message(st.session_state.twilio_client, twilio_conversation_service_sid_to_use, conv_sid, response_whatsapp, twilio_bot_whatsapp_identity_to_use): # Use correct vars
                            st.session_state.processed_message_sids.add(msg_sid)
                            st.success(f"Responded to {msg_sid} from {author_id}")
                        else: st.error(f"Failed to send response for {msg_sid}")
                    st.experimental_rerun()
# --- Footer & Status ---
st.sidebar.markdown("---")
st.sidebar.info("Ensure all keys and SIDs are correctly configured. Primary API keys (Twilio SID/Token, GROQ Key) are loaded from secrets if available.")
if st.session_state.get("app_started"):
    st.sidebar.success(f"App RUNNING. WhatsApp Bot {'RUNNING' if st.session_state.get('bot_started') else 'STOPPED'}.")
else:
    st.sidebar.warning("App is STOPPED.")