app.py

import pickle
import streamlit as st
import numpy as np
import pandas as pd
import re
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing.sequence import pad_sequences
from nltk.corpus import stopwords
from sklearn.preprocessing import StandardScaler
import nltk

# Download required NLTK data (for stopwords only)
nltk.download('stopwords')

# Load the trained model
model = load_model("./best_model.keras")

# Load the tokenizer
with open("./tokenizer.pkl", "rb") as f:
    tokenizer = pickle.load(f)

# Ensure the tokenizer is valid
if not hasattr(tokenizer, 'texts_to_sequences'):
    raise ValueError("Loaded tokenizer is not a valid Keras Tokenizer object.")

# Define constants
MAX_SEQUENCE_LENGTH = 100

# Simple tokenizer function
def simple_tokenizer(text):
    text = text.lower()
    text = re.sub(r'[^a-zA-Z\s]', '', text)
    tokens = text.split()  # Split by whitespace
    return tokens

# Preprocessing function for input review
def preprocess_input_review(review_text):
    tokens = simple_tokenizer(review_text)
    stop_words = set(stopwords.words('english'))
    tokens = [word for word in tokens if word not in stop_words]
    processed_review = ' '.join(tokens)
    return processed_review

# Convert simplified time strings (e.g., 2H30M or 15M) to minutes
def time_to_minutes(time_str):
    if not time_str or pd.isna(time_str):
        return 0
    hours, minutes = 0, 0
    if 'H' in time_str:
        parts = time_str.split('H')
        hours = int(parts[0]) if parts[0].isdigit() else 0
        time_str = parts[1] if len(parts) > 1 else ''
    if 'M' in time_str and time_str.split('M')[0].isdigit():
        minutes = int(time_str.split('M')[0])
    return hours * 60 + minutes

# Prediction function
def predict_menu_inclusion(food_name, review_text, cook_time, prep_time, total_time, protein, carbs, fat):
    # Preprocess the review
    processed_review = preprocess_input_review(review_text)
    
    # Tokenize and pad the review
    review_sequence = tokenizer.texts_to_sequences([processed_review])
    padded_review = pad_sequences(review_sequence, maxlen=MAX_SEQUENCE_LENGTH)
    
    # Prepare numerical features
    numerical_features = np.array([[time_to_minutes(cook_time), 
                                     time_to_minutes(prep_time), 
                                     time_to_minutes(total_time), 
                                     protein, carbs, fat]])
    
    # Normalize numerical features
    scaler = StandardScaler()
    numerical_features = scaler.fit_transform(numerical_features)
    
    # Predict the rating
    predicted_rating = model.predict([padded_review, numerical_features]).flatten()[0]
    
    # Decide whether to include the item on the menu
    include_on_menu = predicted_rating > 4  # Threshold for inclusion
    
    return {
        "FoodName": food_name,
        "PredictedRating": round(predicted_rating, 2),
        "IncludeOnMenu": include_on_menu
    }

# Streamlit app with enhanced UI
st.set_page_config(page_title="Recipe Rating & Menu Optimization", layout="wide")

st.markdown(
    """
    <style>
        .main {
            background-color: #f5f5f5;
            font-family: Arial, sans-serif;
        }
        h1 {
            color: #3c3c7b;
        }
        .sidebar .sidebar-content {
            background-color: #eef2f3;
        }
        .stButton>button {
            background-color: #3c3c7b;
            color: white;
            border-radius: 5px;
            font-size: 16px;
        }
        .stButton>button:hover {
            background-color: #2c2c5a;
        }
        .prediction-box {
            background-color: #eaf6f6;
            padding: 20px;
            border-radius: 10px;
            box-shadow: 2px 2px 10px #ccc;
        }
        .success {
            color: green;
            font-weight: bold;
        }
        .error {
            color: red;
            font-weight: bold;
        }
    </style>
    """,
    unsafe_allow_html=True,
)

st.title("🌟 Recipe Rating & Menu Optimization")

st.sidebar.header("📋 Enter Recipe Details")
food_name = st.sidebar.text_input("🍴 Food Name", value="Paneer Butter Masala")
review_text = st.sidebar.text_area("📝 Review Text", value="This dish is delicious, rich in flavor, and a favorite among customers!")
cook_time = st.sidebar.text_input("⏳ Cooking Time (e.g., 15M or 2H30M)", value="30M")
prep_time = st.sidebar.text_input("🕒 Preparation Time (e.g., 15M or 1H15M)", value="15M")
total_time = st.sidebar.text_input("📅 Total Time (e.g., 45M or 1H45M)", value="45M")
protein = st.sidebar.number_input("💪 Protein (grams)", min_value=0.0, step=1.0, value=20.0)
carbs = st.sidebar.number_input("🍞 Carbohydrates (grams)", min_value=0.0, step=1.0, value=10.0)
fat = st.sidebar.number_input("🍟 Fat (grams)", min_value=0.0, step=1.0, value=15.0)

if st.sidebar.button("🔮 Predict"):
    result = predict_menu_inclusion(food_name, review_text, cook_time, prep_time, total_time, protein, carbs, fat)
    st.markdown("<div class='prediction-box'>", unsafe_allow_html=True)
    st.subheader("Prediction Results")
    st.write(f"**🍴 Food Name:** {result['FoodName']}")
    st.write(f"**⭐ Predicted Rating:** {result['PredictedRating']}")
    st.write(
        f"**🍽️ Include on Menu:** "
        f"<span class='{'success' if result['IncludeOnMenu'] else 'error'}'>"
        f"{'Yes' if result['IncludeOnMenu'] else 'No'}</span>",
        unsafe_allow_html=True,
    )
    st.markdown("</div>", unsafe_allow_html=True)