app.py

import streamlit as st
import numpy as np
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
import pandas as pd
import re

# Load the trained model
model = load_model('final_poetry_model.h5')

# Load and preprocess dataset (for tokenizer)
df = pd.read_csv('Roman-Urdu-Poetry.csv')

def clean_text(text):
    text = text.lower()  # Convert to lowercase
    text = re.sub(r"[^a-zA-Zñḳḍāī\s]", "", text)  # Keep letters, diacritics, apostrophes
    text = re.sub(r'(\n)(\S)', r'\1 \2', text)
    return text

df['Poetry'] = df['Poetry'].apply(clean_text)

# Initialize and fit tokenizer
tokenizer = Tokenizer(num_words=5000, filters='')
tokenizer.fit_on_texts(df['Poetry'])


total_words = len(tokenizer.word_index) + 1

# Function to generate poetry
def generate_poem(seed_text, next_words, max_sequence_len):
    for _ in range(next_words):
        token_list = tokenizer.texts_to_sequences([seed_text])[0]
        token_list = pad_sequences([token_list], maxlen=max_sequence_len-1, padding='pre')
        predicted = model.predict(token_list, verbose=0)
        predicted_word_index = np.argmax(predicted, axis=1)[0]
        predicted_word = tokenizer.index_word.get(predicted_word_index, '')
        seed_text += " " + predicted_word
    return seed_text

# Streamlit UI
st.title("Poetry Generator")
st.write("Enter a seed phrase to generate poetry in Roman Urdu!")

seed_text = st.text_input("Enter seed text:")
next_words = st.slider("Number of words to generate:", min_value=5, max_value=100, value=50)

if st.button("Generate Poetry"):
    max_sequence_len = 225  # Ensure this matches your training setup
    generated_poem = generate_poem(clean_text(seed_text), next_words, max_sequence_len)
    st.subheader("Generated Poetry:")
    st.text(generated_poem)