import numpy as np
import pandas as pd
import subprocess
import gradio as gr



subprocess.call(["pip", "install", "scikit-learn"])

from sklearn.feature_extraction.text import CountVectorizer
from sklearn.metrics.pairwise import cosine_similarity


def p_data(df):
    """Clean and preprocess the recipe dataset."""
    df = df.drop_duplicates('name')
    df = df.dropna()
    df['name'] = df['name'].str.split('-').str[0]  # Extract full recipe name instead of ingredients
    return df

def feature_extraction(df):
    """Use scikit-learn's CountVectorizer to convert the ingredient text into numerical feature vectors."""
    vectorizer = CountVectorizer()
    ingredient_vectors = vectorizer.fit_transform(df['ingredients'])
    return ingredient_vectors, vectorizer

def calculate_ingredient_similarity(ingredient_vectors):
    """Calculate the cosine similarity between all the recipe ingredient vectors."""
    similarity_matrix = cosine_similarity(ingredient_vectors)
    return similarity_matrix

def recommend_recipes(similarity_matrix, input_ingredients, vectorizer, df):
    """Given a set of input ingredients, calculate the similarity between these ingredients and all the recipe ingredient vectors. Return the top-rated recipe as a suggestion."""
    ingredient_vector = vectorizer.transform([input_ingredients])
    similarity_scores = cosine_similarity(ingredient_vectors, ingredient_vector)
    similarity_scores = np.squeeze(similarity_scores)  # Convert to 1-dimensional array
    top_recipe_index = np.argmax(similarity_scores)
    recommended_recipe = df.iloc[top_recipe_index]
    if similarity_scores[top_recipe_index] == 0:
        return "No recipes found."
    else:
        recipe_name = recommended_recipe['name']
        recipe_ingredients = recommended_recipe['ingredients']
        
        return f"Name: {recipe_name}\nIngredients: {recipe_ingredients}"

if __name__ == '__main__':
    df = pd.read_csv('indian_food2.csv')
    df = p_data(df)
    ingredient_vectors, vectorizer = feature_extraction(df)
    similarity_matrix = calculate_ingredient_similarity(ingredient_vectors)
    def recipe_recommender(ingredients):
        recommended_recipe = recommend_recipes(similarity_matrix, ingredients, vectorizer, df)
        return recommended_recipe

    iface = gr.Interface(
        fn=recipe_recommender,
        inputs="text",
        outputs="text",
        live=True,
        examples=[
            ["rice, chicken, tomato"]
        ]
    )
    iface.launch()