app.py

import gradio as gr
import numpy as np
import pandas as pd
import tensorflow as tf
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import load_model

# Step 1: Load and recreate tokenizer
df = pd.read_csv("dataset.csv", encoding='latin-1')[['text_type', 'text']]
df.columns = ['label', 'message']

tokenizer = Tokenizer(num_words=5000, oov_token="<OOV>")
tokenizer.fit_on_texts(df['message'])

# Step 2: Load trained model
model = load_model("spam_detection_model.h5")
max_length = 100

# Step 3: Prediction function with confidence, color, and audio trigger
history_list = []
audio_path = "audio.mpeg"

def predict_message(msg):
    seq = tokenizer.texts_to_sequences([msg])
    padded = pad_sequences(seq, maxlen=max_length, padding='post')
    pred = model.predict(padded)[0][0]
    
    label = "🚫 SPAM" if pred > 0.5 else "✅ HAM"
    color = "red" if pred > 0.5 else "green"
    advice = "⚠️ Don't click unknown links!" if pred > 0.5 else "📨 Safe to read."

    history_list.append((msg, f"{label} ({pred*100:.2f}%)"))

    html_output = f"""
    <div style="color:{color}; font-size: 22px; font-weight: bold;">{label}</div>
    <div style="font-size: 16px; margin-top: 5px;">Confidence: {pred*100:.2f}%</div>
    <div style="margin-top: 10px;">{advice}</div>
    """

    # Return audio only if SPAM
    return html_output, history_list[-5:], audio_path if pred > 0.5 else None

# Step 4: Gradio UI
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("""
    # 📩 SMS Spam Classifier
    _Check if your message is spam or not using a powerful LSTM neural network._
    - 💡 Type your message below
    - 🧠 Trained on real-world SMS dataset
    - 🚨 Hear a warning if it's spam!
    """)

    msg_input = gr.Textbox(label="Enter SMS Text", placeholder="E.g. You've won ₹10,000 cash prize!", lines=2)
    submit_btn = gr.Button("🔍 Analyze")
    clear_btn = gr.Button("🧹 Clear History")

    result = gr.HTML()
    history = gr.Dataframe(headers=["Message", "Prediction"], label="📋 Prediction History", interactive=False)
    audio_out = gr.Audio(label="🔊 Spam Alert", interactive=False, autoplay=True)

    def clear_history():
        history_list.clear()
        return None, [], None

    submit_btn.click(predict_message, inputs=msg_input, outputs=[result, history, audio_out])
    clear_btn.click(clear_history, inputs=[], outputs=[result, history, audio_out])

demo.launch()