import streamlit as st
from transformers import pipeline
import warnings

warnings.filterwarnings('ignore')

st.set_page_config(
    page_title="AI Solutions for Customer Complaints",
    layout="wide",
    initial_sidebar_state="expanded"
)

# --- CSS for centering content and enlarging tabs ---
st.markdown("""
<style>
/* Center all content inside the page */
.centered-div {
    text-align: center !important;
    max-width: 800px;
    margin: auto;
}

/* Center tabs */
div[role="tablist"] {
    justify-content: center;
}

/* Increase tab size */
button[role="tab"] {
    font-size: 20px !important;
    padding: 12px 30px !important;
    height: 60px !important;
}

/* Center all Streamlit headers and text elements inside the div */
.centered-div h1, 
.centered-div h2, 
.centered-div h3, 
.centered-div h4, 
.centered-div h5, 
.centered-div h6, 
.centered-div p {
    text-align: center !important;
}
</style>
""", unsafe_allow_html=True)

# --- Page Title ---
# Wrapped the title and introductory text in a centered-div
st.markdown('<div class="centered-div">', unsafe_allow_html=True)
st.title("🤖 AI Customer Support Solutions")
st.markdown("""
Welcome! Explore AI-powered classification for customer complaints in **E-commerce** and **Banking**.  
Select a tab below to get started.
""")
st.markdown('</div>', unsafe_allow_html=True)

# --- Tabs ---
tab1, tab2 = st.tabs(["🛍️ E-commerce", "🏦 Bank"])

# ================= E-COMMERCE TAB =================
with tab1:
    st.markdown('<div class="centered-div">', unsafe_allow_html=True)
    
    st.header("E-commerce Customer Support Classification")
    
    st.subheader("📂 Categories You Can Explore:")
    st.markdown("""
    - 📦 **Delivery Problem** – Delays, lost packages, or damaged goods.
    - 💸 **Returns and Refunds** – Queries about refunds or return policies.
    - 🛒 **Product Inquiry** – Asking about product details, specifications, or availability.
    - 💳 **Payment Issue** – Failed transactions, double charges, or payment errors.
    - 🗂️ **...and 11 more categories fine-tuned in the model!**
    """)
    
    # --- Try It Yourself Section ---
    st.subheader("🔎 Try It Yourself!")
    try:
        ecommerce_classifier = pipeline(
            'text-classification',
            model='E-commerce-customer-query-classifier',
            trust_remote_code=True
        )
    except Exception:
        st.error("⚠️ Could not load the E-commerce model.")
        st.stop()
    
    text = st.text_area("✍️ Enter the customer query:", height=150, placeholder="I haven’t received my order yet.")
    if st.button("Classify E-commerce Query", use_container_width=True, type="primary"):
        if text.strip():
            with st.spinner('🤔 Analyzing the query...'):
                try:
                    result = ecommerce_classifier(text)
                    st.success(f"✅ The query has been classified as: **{result[0]['label']}**")
                    st.info(f"🔒 Confidence Level: **{result[0]['score']:.4f}**")
                except Exception as e:
                    st.error(f"❌ Classification error: {e}")
        else:
            st.warning("⚠️ Please enter a query to classify.")
    
    st.subheader("🔧 Technical Overview & Features")
    st.markdown("""
    **Model & Dataset:**
    - Fine-tuned `distilbert-base-uncased`
    - Dataset: [`Ataur77/ecommerce-customer-support`](https://huggingface.co/datasets/Ataur77/ecommerce-customer-support)

    **Framework & Libraries:**
    - Streamlit for interactive UI
    - Hugging Face Transformers

    **Features:**
    - 🚀 Quick Classification – Queries are categorized instantly.
    - 🔌 Offline Deployment – Works with locally stored models.
    - 🧩 Scalable Design – Extendable to more categories or data.
    - 😀 User-Friendly – Minimal clicks, maximum insight.
    """)
    
    st.image("e-comm-accuracy.png", caption="📊 Model Accuracy", width=700)
    
    st.subheader("📩 Contact")
    st.markdown("If you’d like to collaborate or learn more: [chiraagpv2000@gmail.com](mailto:chiraagpv2000@gmail.com)")
    
    st.markdown('</div>', unsafe_allow_html=True)

# ================= BANK TAB =================
with tab2:
    st.markdown('<div class="centered-div">', unsafe_allow_html=True)
    
    st.header("Bank Customer Complaint Classification")
    
    st.subheader("📂 Example Categories:")
    st.markdown("""
    - 💳 **Credit Card or Prepaid Card** – Fraud, charges, or transaction disputes.
    - 🏦 **Checking or Savings Account** – Account balance, overdrafts, or account access.
    - 🏡 **Mortgage** – Loan approvals, EMI, foreclosure, or rate queries.
    """)
    
    # --- Try It Yourself Section ---
    st.subheader("🔎 Try It Yourself!")
    bank_classifier = pipeline('text-classification', model='bank_customer_ticket_category_classifier')
    
    text = st.text_area("✍️ Please describe your issue:", placeholder="My credit card was charged twice for one purchase.")
    if st.button("Classify Bank Complaint", use_container_width=True, type="primary"):
        if text.strip():
            with st.spinner('🔍 Analyzing complaint...'):
                result = bank_classifier(text)
                st.success(f"✅ Your issue is classified under: **{result[0]['label']}**")
                st.info(f"🔒 Confidence Level: **{result[0]['score']:.4f}**")
        else:
            st.warning("⚠️ Please enter your complaint.")
    
    st.subheader("🔧 Technical Overview & Features")
    st.markdown("""
    **Model & Framework:**
    - Fine-tuned Transformer
    - Streamlit for rapid prototyping

    **Libraries:**
    - Hugging Face Transformers

    **Features:**
    - ⏱️ Faster Processing – No waiting for manual triage.
    - 📊 High Accuracy – Reliable complaint classification.
    - 📡 Scalability – Handles thousands of inputs seamlessly.
    """)
    
    st.image("bank-accuracy.png", caption="📊 Model Accuracy", width=700)
    
    st.subheader("📩 Contact")
    st.markdown("Reach out?: [chiraagpv2000@gmail.com](mailto:chiraagpv2000@gmail.com)")
    
    st.markdown('</div>', unsafe_allow_html=True)