model.py

import numpy as np
import re
from sklearn.preprocessing import StandardScaler
from sklearn.neighbors import NearestNeighbors
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import FunctionTransformer

def scaling(dataframe):
    scaler=StandardScaler()
    prep_data=scaler.fit_transform(dataframe.iloc[:,6:15].to_numpy())
    return prep_data,scaler

def nn_predictor(prep_data):
    neigh = NearestNeighbors(metric='cosine',algorithm='brute')
    neigh.fit(prep_data)
    return neigh

def build_pipeline(neigh,scaler,params):
    transformer = FunctionTransformer(neigh.kneighbors,kw_args=params)
    pipeline=Pipeline([('std_scaler',scaler),('NN',transformer)])
    return pipeline

def extract_data(dataframe,ingredients):
    extracted_data=dataframe.copy()
    extracted_data=extract_ingredient_filtered_data(extracted_data,ingredients)
    return extracted_data
    
def extract_ingredient_filtered_data(dataframe,ingredients):
    extracted_data=dataframe.copy()
    regex_string=''.join(map(lambda x:f'(?=.*{x})',ingredients))
    extracted_data=extracted_data[extracted_data['RecipeIngredientParts'].str.contains(regex_string,regex=True,flags=re.IGNORECASE)]
    return extracted_data

def apply_pipeline(pipeline,_input,extracted_data):
    _input=np.array(_input).reshape(1,-1)
    return extracted_data.iloc[pipeline.transform(_input)[0]]

import pandas as pd
def recommend(dataset,_input,ingredients=[],params={'n_neighbors':5,'return_distance':False})-> pd.DataFrame:
    extracted_data=extract_data(dataset,ingredients)
    if extracted_data.shape[0]>=params['n_neighbors']:
        prep_data,scaler=scaling(extracted_data)
        neigh=nn_predictor(prep_data)
        pipeline=build_pipeline(neigh,scaler,params)
        return apply_pipeline(pipeline,_input,extracted_data)
    else:
       raise Exception("an error occured") 

def extract_quoted_strings(s):
    # Find all the strings inside double quotes
    strings = re.findall(r'"([^"]*)"', s)
    # Join the strings with 'and'
    return strings

def output_recommended_recipes(dataframe):
    if dataframe is not None:
        output=dataframe.copy()
        output=output.to_dict("records")
        for recipe in output:
            recipe['RecipeIngredientParts']=extract_quoted_strings(recipe['RecipeIngredientParts'])
            recipe['RecipeInstructions']=extract_quoted_strings(recipe['RecipeInstructions'])
    else:
        output=None
    return output