app.py

import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestClassifier
from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import train_test_split, GridSearchCV
import gradio as gr
import os
import warnings
import logging

warnings.filterwarnings('ignore')
logging.basicConfig(level=logging.INFO)

# Dataset generation
np.random.seed(42)
moods = ['happy', 'stressed', 'bored', 'sad', 'excited', 'tired', 'anxious', 'content', 'nostalgic', 'hungry']
snacks = [
    'fruit', 'chocolate', 'chips', 'popcorn', 'ice cream', 'pretzels', 'cookies', 'candy', 
    'yogurt', 'granola bar', 'crackers', 'veggies', 'cheese',
    'chin chin', 'kuli kuli', 'plantain chips', 'puff puff', 'akara', 'coconut candy', 
    'kokoro', 'dodo ikire', 'roasted groundnuts', 'suya', 'boli', 'kilishi',
    'buns', 'doughnuts', 'meat pie', 'egg rolls'
]
times_of_day = ['morning', 'afternoon', 'evening', 'midnight']

snack_groups = {
    'nigerian_fried': ['chin chin', 'puff puff', 'akara', 'buns', 'doughnuts', 'meat pie', 'egg rolls'],
    'nigerian_savory': ['suya', 'kuli kuli', 'plantain chips', 'boli', 'kilishi', 'roasted groundnuts'],
    'nigerian_sweet': ['coconut candy', 'dodo ikire', 'chocolate', 'candy', 'cookies', 'ice cream'],
    'savory_snacks': ['chips', 'popcorn', 'pretzels', 'crackers', 'kokoro'],
    'healthy_light': ['fruit', 'yogurt', 'veggies', 'granola bar', 'cheese']
}
snack_to_group = {snack: group for group, snacks in snack_groups.items() for snack in snacks}
group_list = list(snack_groups.keys())

mood_time_group_probs = {
    'happy': {
        'morning': {'nigerian_fried': 0.75, 'nigerian_sweet': 0.2, 'healthy_light': 0.05},
        'afternoon': {'nigerian_fried': 0.75, 'nigerian_sweet': 0.2, 'healthy_light': 0.05},
        'evening': {'nigerian_sweet': 0.75, 'healthy_light': 0.2, 'savory_snacks': 0.05},
        'midnight': {'nigerian_sweet': 0.75, 'nigerian_savory': 0.2, 'savory_snacks': 0.05}
    },
    'stressed': {
        'morning': {'nigerian_sweet': 0.75, 'nigerian_fried': 0.2, 'healthy_light': 0.05},
        'afternoon': {'nigerian_sweet': 0.75, 'nigerian_fried': 0.2, 'savory_snacks': 0.05},
        'evening': {'nigerian_sweet': 0.75, 'savory_snacks': 0.2, 'nigerian_savory': 0.05},
        'midnight': {'nigerian_sweet': 0.75, 'nigerian_savory': 0.2, 'savory_snacks': 0.05}
    },
    'bored': {
        'morning': {'savory_snacks': 0.75, 'nigerian_fried': 0.2, 'healthy_light': 0.05},
        'afternoon': {'savory_snacks': 0.75, 'nigerian_savory': 0.2, 'healthy_light': 0.05},
        'evening': {'savory_snacks': 0.75, 'nigerian_savory': 0.2, 'healthy_light': 0.05},
        'midnight': {'savory_snacks': 0.75, 'nigerian_savory': 0.2, 'healthy_light': 0.05}
    },
    'sad': {
        'morning': {'nigerian_sweet': 0.75, 'nigerian_fried': 0.2, 'healthy_light': 0.05},
        'afternoon': {'nigerian_sweet': 0.75, 'healthy_light': 0.2, 'nigerian_fried': 0.05},
        'evening': {'nigerian_sweet': 0.75, 'healthy_light': 0.2, 'savory_snacks': 0.05},
        'midnight': {'nigerian_sweet': 0.75, 'healthy_light': 0.2, 'nigerian_savory': 0.05}
    },
    'excited': {
        'morning': {'nigerian_fried': 0.75, 'nigerian_sweet': 0.2, 'healthy_light': 0.05},
        'afternoon': {'nigerian_fried': 0.75, 'nigerian_savory': 0.2, 'nigerian_sweet': 0.05},
        'evening': {'nigerian_sweet': 0.75, 'nigerian_savory': 0.2, 'savory_snacks': 0.05},
        'midnight': {'nigerian_savory': 0.75, 'nigerian_sweet': 0.2, 'savory_snacks': 0.05}
    },
    'tired': {
        'morning': {'healthy_light': 0.75, 'nigerian_fried': 0.2, 'nigerian_sweet': 0.05},
        'afternoon': {'healthy_light': 0.75, 'nigerian_fried': 0.2, 'savory_snacks': 0.05},
        'evening': {'healthy_light': 0.75, 'nigerian_sweet': 0.2, 'savory_snacks': 0.05},
        'midnight': {'healthy_light': 0.75, 'nigerian_savory': 0.2, 'nigerian_sweet': 0.05}
    },
    'anxious': {
        'morning': {'savory_snacks': 0.75, 'nigerian_fried': 0.2, 'healthy_light': 0.05},
        'afternoon': {'savory_snacks': 0.75, 'nigerian_savory': 0.2, 'healthy_light': 0.05},
        'evening': {'savory_snacks': 0.75, 'nigerian_savory': 0.2, 'healthy_light': 0.05},
        'midnight': {'savory_snacks': 0.75, 'nigerian_savory': 0.2, 'healthy_light': 0.05}
    },
    'content': {
        'morning': {'healthy_light': 0.75, 'nigerian_fried': 0.2, 'nigerian_sweet': 0.05},
        'afternoon': {'nigerian_savory': 0.75, 'healthy_light': 0.2, 'nigerian_fried': 0.05},
        'evening': {'healthy_light': 0.75, 'nigerian_sweet': 0.2, 'savory_snacks': 0.05},
        'midnight': {'healthy_light': 0.75, 'nigerian_savory': 0.2, 'nigerian_sweet': 0.05}
    },
    'nostalgic': {
        'morning': {'nigerian_sweet': 0.75, 'nigerian_fried': 0.2, 'healthy_light': 0.05},
        'afternoon': {'nigerian_sweet': 0.75, 'nigerian_fried': 0.2, 'healthy_light': 0.05},
        'evening': {'nigerian_sweet': 0.75, 'healthy_light': 0.2, 'savory_snacks': 0.05},
        'midnight': {'nigerian_sweet': 0.75, 'nigerian_savory': 0.2, 'healthy_light': 0.05}
    },
    'hungry': {
        'morning': {'nigerian_fried': 0.75, 'savory_snacks': 0.2, 'healthy_light': 0.05},
        'afternoon': {'nigerian_savory': 0.75, 'nigerian_fried': 0.2, 'savory_snacks': 0.05},
        'evening': {'nigerian_savory': 0.75, 'savory_snacks': 0.2, 'nigerian_sweet': 0.05},
        'midnight': {'nigerian_savory': 0.75, 'savory_snacks': 0.2, 'nigerian_sweet': 0.05}
    }
}

n_samples = 1800
data = {'mood': [], 'time_of_day': [], 'hunger_level': [], 'sentiment': [], 'snack': [], 'snack_group': []}

for _ in range(n_samples):
    mood = np.random.choice(moods, p=[0.15, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.05])
    time = np.random.choice(times_of_day)
    hunger_level = 1.0 if mood == 'hungry' else np.random.uniform(0, 0.8)
    sentiment = round(np.random.uniform(-1, 1), 2)
    
    group_probs = [mood_time_group_probs[str(mood)][time].get(g, 0.01) for g in group_list]
    group = np.random.choice(group_list, p=group_probs / np.sum(group_probs))
    group_snacks = snack_groups[group]
    snack_probs = [
        0.6 if (snack == 'suya' and time in ['evening', 'midnight']) or 
               (snack == 'boli' and time == 'afternoon') or
               (snack in ['puff puff', 'buns', 'doughnuts', 'meat pie', 'egg rolls'] and time in ['morning', 'afternoon']) or
               (snack == 'akara' and time in ['morning', 'midnight']) or
               (snack == 'chin chin' and time in ['morning', 'afternoon', 'midnight'])
        else 0.35 if snack in ['kuli kuli', 'plantain chips', 'popcorn', 'kokoro', 'roasted groundnuts', 'kilishi'] 
        else 0.2 for snack in group_snacks
    ]
    if time not in ['evening', 'midnight'] and 'suya' in group_snacks:
        snack_probs[group_snacks.index('suya')] = 0
    if time != 'afternoon' and 'boli' in group_snacks:
        snack_probs[group_snacks.index('boli')] = 0
    if time not in ['morning', 'afternoon']:
        for snack in ['puff puff', 'buns', 'doughnuts', 'meat pie', 'egg rolls']:
            if snack in group_snacks:
                snack_probs[group_snacks.index(snack)] = 0
    if time not in ['morning', 'midnight'] and 'akara' in group_snacks:
        snack_probs[group_snacks.index('akara')] = 0
    snack_probs = [p / sum(snack_probs) if sum(snack_probs) > 0 else 0.2 for p in snack_probs]
    snack = np.random.choice(group_snacks, p=snack_probs)
    
    data['mood'].append(mood)
    data['time_of_day'].append(time)
    data['hunger_level'].append(hunger_level)
    data['sentiment'].append(sentiment)
    data['snack'].append(snack)
    data['snack_group'].append(group)

df = pd.DataFrame(data)

# Adjust sentiment
df.loc[df['mood'].isin(['happy', 'excited', 'content', 'nostalgic']), 'sentiment'] = df.loc[
    df['mood'].isin(['happy', 'excited', 'content', 'nostalgic']), 'sentiment'].clip(lower=0.2)
df.loc[df['mood'].isin(['stressed', 'sad', 'anxious', 'tired']), 'sentiment'] = df.loc[
    df['mood'].isin(['stressed', 'sad', 'anxious', 'tired']), 'sentiment'].clip(upper=-0.1)
df.loc[df['mood'].isin(['bored', 'hungry']), 'sentiment'] = df.loc[
    df['mood'].isin(['bored', 'hungry']), 'sentiment'].clip(-0.3, 0.3)

# Add snack_type and snack_texture
snack_types = {
    'chin chin': 'sweet', 'puff puff': 'sweet', 'akara': 'savory', 'suya': 'spicy', 
    'kuli kuli': 'spicy', 'plantain chips': 'savory', 'coconut candy': 'sweet', 
    'dodo ikire': 'sweet', 'roasted groundnuts': 'savory', 'fruit': 'light', 
    'yogurt': 'light', 'veggies': 'light', 'granola bar': 'light', 'cheese': 'light',
    'chocolate': 'sweet', 'candy': 'sweet', 'cookies': 'sweet', 'ice cream': 'sweet',
    'chips': 'savory', 'popcorn': 'savory', 'pretzels': 'savory', 'crackers': 'savory',
    'kokoro': 'savory', 'boli': 'savory', 'kilishi': 'spicy',
    'buns': 'sweet', 'doughnuts': 'sweet', 'meat pie': 'savory', 'egg rolls': 'savory'
}
snack_textures = {
    'chin chin': 'crisp', 'puff puff': 'soft', 'akara': 'soft', 'suya': 'chewy', 
    'kuli kuli': 'crisp', 'plantain chips': 'crisp', 'coconut candy': 'chewy', 
    'dodo ikire': 'soft', 'roasted groundnuts': 'crisp', 'fruit': 'soft', 
    'yogurt': 'soft', 'veggies': 'crisp', 'granola bar': 'crisp', 'cheese': 'soft',
    'chocolate': 'soft', 'candy': 'chewy', 'cookies': 'crisp', 'ice cream': 'soft',
    'chips': 'crisp', 'popcorn': 'crisp', 'pretzels': 'crisp', 'crackers': 'crisp',
    'kokoro': 'crisp', 'boli': 'soft', 'kilishi': 'chewy',
    'buns': 'soft', 'doughnuts': 'soft', 'meat pie': 'soft', 'egg rolls': 'soft'
}
df['snack_type'] = df['snack'].map(snack_types)
df['snack_texture'] = df['snack'].map(snack_textures)

# Encode features
le_mood = LabelEncoder()
le_time = LabelEncoder()
le_type = LabelEncoder()
le_texture = LabelEncoder()
le_group = LabelEncoder()

df['mood_encoded'] = le_mood.fit_transform(df['mood'])
df['time_encoded'] = le_time.fit_transform(df['time_of_day'])
df['type_encoded'] = le_type.fit_transform(df['snack_type'])
df['texture_encoded'] = le_texture.fit_transform(df['snack_texture'])
df['group_encoded'] = le_group.fit_transform(df['snack_group'])

X = df[['mood_encoded', 'time_encoded', 'hunger_level', 'sentiment', 'type_encoded', 'texture_encoded']]
y = df['group_encoded']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42, stratify=y)

# Train model
param_grid = {
    'n_estimators': [300, 400],
    'max_depth': [12, 15],
    'min_samples_split': [5, 10]
}
model = RandomForestClassifier(class_weight='balanced', random_state=42)
grid_search = GridSearchCV(model, param_grid, cv=5, scoring='accuracy', n_jobs=-1)
grid_search.fit(X_train, y_train)
best_model = grid_search.best_estimator_

# Prediction function
def predict_snack(mood, time_of_day, hunger_level, sentiment, snack_type):
    mood_enc = le_mood.transform([mood])[0]
    time_enc = le_time.transform([time_of_day])[0]
    type_enc = le_type.transform([snack_type])[0]
    type_to_texture = {'sweet': 'soft', 'savory': 'crisp', 'spicy': 'chewy', 'light': 'soft'}
    texture_enc = le_texture.transform([type_to_texture[snack_type]])[0]
    input_data = np.array([[mood_enc, time_enc, hunger_level, sentiment, type_enc, texture_enc]])
    pred = best_model.predict(input_data)
    group = le_group.inverse_transform(pred)[0]
    group_snacks = snack_groups[group]
    snack_probs = [
        0.6 if (snack == 'suya' and time_of_day in ['evening', 'midnight']) or 
               (snack == 'boli' and time_of_day == 'afternoon') or
               (snack in ['puff puff', 'buns', 'doughnuts', 'meat pie', 'egg rolls'] and time_of_day in ['morning', 'afternoon']) or
               (snack == 'akara' and time_of_day in ['morning', 'midnight']) or
               (snack == 'chin chin' and time_of_day in ['morning', 'afternoon', 'midnight'])
        else 0.35 if snack in ['kuli kuli', 'plantain chips', 'popcorn', 'kokoro', 'roasted groundnuts', 'kilishi'] 
        else 0.2 for snack in group_snacks
    ]
    if time_of_day not in ['evening', 'midnight'] and 'suya' in group_snacks:
        snack_probs[group_snacks.index('suya')] = 0
    if time_of_day != 'afternoon' and 'boli' in group_snacks:
        snack_probs[group_snacks.index('boli')] = 0
    if time_of_day not in ['morning', 'afternoon']:
        for snack in ['puff puff', 'buns', 'doughnuts', 'meat pie', 'egg rolls']:
            if snack in group_snacks:
                snack_probs[group_snacks.index(snack)] = 0
    if time_of_day not in ['morning', 'midnight'] and 'akara' in group_snacks:
        snack_probs[group_snacks.index('akara')] = 0
    snack_probs = [p / sum(snack_probs) if sum(snack_probs) > 0 else 0.2 for p in snack_probs]
    snack = np.random.choice(group_snacks, p=snack_probs)
    return f"You should try {snack}!", snack

# Gradio interface
with gr.Blocks(css="""
    body {background-color: #FFF8E7; font-family: 'Poppins', sans-serif;}
    .gradio-container {max-width: 800px; margin: auto; padding: 20px;}
    h1 {color: #4A2C2A; text-align: center; font-size: 2.5em; margin-bottom: 10px;}
    p {color: #4A2C2A; text-align: center; font-size: 1.2em;}
    .gr-button {background-color: #FF4500 !important; color: white !important; border-radius: 25px !important; padding: 10px 20px !important; font-weight: bold !important;}
    .gr-button:hover {background-color: #E03C00 !important;}
    .gr-textbox, .gr-dropdown, .gr-slider {border: 2px solid #D4A373 !important; border-radius: 10px !important; padding: 10px !important;}
    .gr-image {border-radius: 15px; margin: auto; max-width: 200px;}
    .footer {text-align: center; color: #808080; font-size: 0.9em; margin-top: 20px;}
    .explanations {text-align: center; color: #4A2C2A; font-size: 0.9em; margin-top: 20px;}
    .explanations p {margin: 5px 0; text-align: left; display: inline-block;}
""") as demo:
    gr.HTML("""
        <div style='background: linear-gradient(to right, #D4A373, #FEE440); padding: 20px; border-radius: 15px; text-align: center;'>
            <h1>Snack Predictor 🍪</h1>
            <p>Tell us your vibe, and we'll find your perfect snack! Powered by ML (~97% accurate)</p>
            <p style='font-size: 1em; margin-top: 10px;'>Select your mood, time, and preferences below, then hit "Find My Snack!" to get a tasty recommendation with a pic! 😋</p>
        </div>
    """)
    
    with gr.Row():
        with gr.Column(scale=1):
            mood = gr.Dropdown(
                choices=moods, 
                label="Mood", 
                value="happy", 
                elem_classes="gr-dropdown"
            )
            time_of_day = gr.Dropdown(
                choices=times_of_day, 
                label="Time of Day", 
                value="morning", 
                elem_classes="gr-dropdown"
            )
            hunger_level = gr.Slider(
                minimum=0, 
                maximum=1, 
                step=0.1, 
                label="Hunger Level (0 to 1)", 
                value=0.5, 
                elem_classes="gr-slider"
            )
            sentiment = gr.Slider(
                minimum=-1, 
                maximum=1, 
                step=0.1, 
                label="Sentiment (-1 to 1)", 
                value=0.0, 
                elem_classes="gr-slider"
            )
            snack_type = gr.Dropdown(
                choices=['sweet', 'savory', 'spicy', 'light'], 
                label="Snack Type", 
                value="sweet", 
                elem_classes="gr-dropdown"
            )
            predict_btn = gr.Button("Find My Snack!", variant="primary", elem_classes="gr-button")
        
        with gr.Column(scale=1):
            output_text = gr.Textbox(label="Your Snack Recommendation", elem_classes="gr-textbox")
            output_image = gr.Image(label="Snack Preview", elem_classes="gr-image")
    
    def predict_and_show(mood, time_of_day, hunger_level, sentiment, snack_type):
        text, snack = predict_snack(mood, time_of_day, hunger_level, sentiment, snack_type)
        image_path = f"assets/{snack.replace(' ', '_')}.jpeg"
        if not os.path.exists(image_path):
            logging.info(f"Image not found: {image_path}")
            image_path = f"assets/{snack.replace(' ', '_')}.png"  # Check for .png as fallback
            if not os.path.exists(image_path):
                logging.info(f"PNG fallback not found: {image_path}, using placeholder")
                image_path = "assets/placeholder.jpeg"  # Final fallback
                if not os.path.exists(image_path):
                    logging.error(f"Placeholder not found: {image_path}")
        return text, image_path
    
    predict_btn.click(
        fn=predict_and_show,
        inputs=[mood, time_of_day, hunger_level, sentiment, snack_type],
        outputs=[output_text, output_image]
    )
    
    gr.HTML("""
        <div class='explanations'>
            <h3 style='color: #FF4500; font-size: 1.1em; margin-bottom: 10px;'>What do the inputs mean?</h3>
            <p>Mood: How are you feeling right now? Pick a mood that matches your vibe.</p>
            <p>Time of Day: What time is it? This helps pick snacks that suit the moment.</p>
            <p>Hunger Level: How hungry are you? 0 = not hungry, 1 = starving!</p>
            <p>Sentiment: What's your emotional vibe? -1 = feeling down, 0 = neutral, +1 = super upbeat.</p>
            <p>Snack Type: What kind of snack do you crave? Sweet, savory, spicy, or light.</p>
        </div>
        <div class='footer'>
            <p>Built with ❤️ by @teganmosi</p>
            <p>Follow my #WeeklyMLProjects for more! 🍟</p>
        </div>
    """)

demo.launch()