app.py

import os
import re
import ast
import json
import pickle
import secrets
import random
import datetime
import pandas as pd
from huggingface_hub import login, InferenceClient
from sklearn.metrics.pairwise import cosine_similarity
import streamlit as st
import wordcloud
import matplotlib.pyplot as plt
from pygwalker.api.streamlit import StreamlitRenderer
import plotly.express as px

st.set_page_config(layout="wide")

# Load environment token
login(token=os.getenv("TOKEN"))

# Load data from pickled files
with open('cv.pkl', 'rb') as file:
    cv = pickle.load(file)

with open('vectors.pkl', 'rb') as file:
    vectors = pickle.load(file)

with open('items_dict.pkl', 'rb') as file:
    items_dict = pd.DataFrame.from_dict(pickle.load(file))


mode = st.toggle(label="MART")

if 'ind' not in st.session_state:
    st.session_state.ind = []

def get_recommendations(user_description, count_vectorizer, count_matrix):
    user_description = preprocess_text(user_description)
    user_vector = count_vectorizer.transform([user_description])
    cosine_similarities = cosine_similarity(user_vector, count_matrix).flatten()
    similar_indices = cosine_similarities.argsort()[::-1]
    return similar_indices


def show_recipe(recipe):
    with st.spinner("HANG TIGHT, RECIPE INCOMING..."):
        name_and_dis = f'# {recipe["name"]}\n\n'
        name_and_dis += f'{recipe["description"]}\n\n'
        ingredients = '## Ingredients:\n'
        instructions = '\n## Instructions:\n'
        for instruction in recipe["instructions"]:
            instructions += f"{instruction['step_number']}. {instruction['instruction']}\n"

        st.write(name_and_dis)
        col01, col02 = st.columns(2)
        with col01:
            cont = st.container(border=True, height=500)
            cont.write(ingredients)
            for j, i in enumerate(recipe["ingredients"]):
                ind = get_recommendations(i['name'], cv, vectors)
                st.session_state.ind.append(ind[:5].tolist())
                cont.selectbox(i['name'],
                               options=items_dict.iloc[ind][
                                   "PRODUCT_NAME"].values,
                               key=f"selectbox_{j}_{i['name']}{random.random() * 100}")
        with col02:
            cont = st.container(border=True, height=500)
            cont.write(instructions)


# Initialize the inference client for the Mixtral model

if not mode:
    cook, saved = st.tabs(['COOK', 'SAVED'])
    with cook:
        client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")

        if 'recipe' not in st.session_state:
            st.session_state.recipe = None

        if 'recipe_saved' not in st.session_state:
            st.session_state.recipe_saved = None

        if 'user_direction' not in st.session_state:
            st.session_state.user_direction = None

        if 'serving_size' not in st.session_state:
            st.session_state.serving_size = 2

        if 'selected_difficulty' not in st.session_state:
            st.session_state.selected_difficulty = "Quick & Easy"

        if 'exclusions' not in st.session_state:
            st.session_state.exclusions = None


        def preprocess_text(text):
            # Remove non-alphabet characters and extra spaces
            text = re.sub(r'[^a-zA-Z\s]', '', text)
            text = re.sub(r'\s+', ' ', text).strip()
            return text.lower()


        def create_detailed_prompt(user_direction, exclusions, serving_size, difficulty):
            if difficulty == "Quick & Easy":
                prompt = (
                    f"Provide a 'Quick and Easy' recipe for {user_direction} that excludes {exclusions} and has a serving size of {serving_size}. "
                    f"It should require as few ingredients as possible and should be ready in as little time as possible. "
                    f"The steps should be simple, and the ingredients should be commonly found in a household pantry. "
                    f"Provide a detailed ingredient list and step-by-step guide that explains the instructions to prepare in detail."
                )
            elif difficulty == "Intermediate":
                prompt = (
                    f"Provide a classic recipe for {user_direction} that excludes {exclusions} and has a serving size of {serving_size}. "
                    f"The recipe should offer a bit of a cooking challenge but should not require professional skills. "
                    f"The recipe should feature traditional ingredients and techniques that are authentic to its cuisine. "
                    f"Provide a detailed ingredient list and step-by-step guide that explains the instructions to prepare in detail."
                )
            elif difficulty == "Professional":
                prompt = (
                    f"Provide an advanced recipe for {user_direction} that excludes {exclusions} and has a serving size of {serving_size}. "
                    f"The recipe should push the boundaries of culinary arts, integrating unique ingredients, advanced cooking techniques, and innovative presentations. "
                    f"The recipe should be able to be served at a high-end restaurant or would impress at a gourmet food competition. "
                    f"Provide a detailed ingredient list and step-by-step guide that explains the instructions to prepare in detail."
                )
            return prompt


        def generate_recipe(user_inputs):
            with st.spinner('Building the perfect recipe...'):
                prompt = create_detailed_prompt(user_inputs['user_direction'], user_inputs['exclusions'],
                                                user_inputs['serving_size'], user_inputs['difficulty'])

                functions = [
                    {
                        "name": "provide_recipe",
                        "description": "Provides a detailed recipe strictly adhering to the user input/specifications, especially ingredient exclusions and the recipe difficulty",
                        "parameters": {
                            "type": "object",
                            "properties": {
                                "name": {
                                    "type": "string",
                                    "description": "A creative name for the recipe"
                                },
                                "description": {
                                    "type": "string",
                                    "description": "a brief one-sentence description of the provided recipe"
                                },
                                "ingredients": {
                                    "type": "array",
                                    "items": {
                                        "type": "object",
                                        "properties": {
                                            "name": {
                                                "type": "string",
                                                "description": "Quantity and name of the ingredient"
                                            }
                                        }
                                    }
                                },
                                "instructions": {
                                    "type": "array",
                                    "items": {
                                        "type": "object",
                                        "properties": {
                                            "step_number": {
                                                "type": "number",
                                                "description": "The sequence number of this step"
                                            },
                                            "instruction": {
                                                "type": "string",
                                                "description": "Detailed description of what to do in this step"
                                            }
                                        }
                                    }
                                }
                            },
                            "required": [
                                "name",
                                "description",
                                "ingredients",
                                "instructions"
                            ],
                        },
                    }
                ]

                generate_kwargs = dict(
                    temperature=0.9,
                    max_new_tokens=10000,
                    top_p=0.9,
                    repetition_penalty=1.0,
                    do_sample=True,
                )

                prompt += f"\nPlease format the output in JSON. The JSON should include fields for 'name', 'description', 'ingredients', and 'instructions', with each field structured as described below.\n\n{json.dumps(functions)}"

                response = client.text_generation(prompt, **generate_kwargs)
                st.session_state.recipe = response
                st.session_state.recipe_saved = False


        def clear_inputs():
            st.session_state.user_direction = None
            st.session_state.exclusions = None
            st.session_state.serving_size = 2
            st.session_state.selected_difficulty = "Quick & Easy"
            st.session_state.recipe = None


        st.title("Let's get cooking")
        col1, col2 = st.columns(2)
        with col1:
            st.session_state.user_direction = st.text_area(
                "What do you want to cook? Describe anything - a dish, cuisine, event, or vibe.",
                value=st.session_state.user_direction,
                placeholder="quick snack, asian style bowl with either noodles or rice, something italian",
            )
        with col2:
            st.session_state.serving_size = st.number_input(
                "How many servings would you like to cook?",
                min_value=1,
                max_value=100,
                value=st.session_state.serving_size,
                step=1
            )

        difficulty_dictionary = {
            "Quick & Easy": {
                "description": "Easy recipes with straightforward instructions. Ideal for beginners or those seeking quick and simple cooking.",
            },
            "Intermediate": {
                "description": "Recipes with some intricate steps that invite a little challenge. Perfect for regular cooks wanting to expand their repertoire with new ingredients and techniques.",
            },
            "Professional": {
                "description": "Complex recipes that demand a high level of skill and precision. Suited for seasoned cooks aspiring to professional-level sophistication and creativity.",
            }
        }

        st.session_state.selected_difficulty = st.radio(
            "Choose a difficulty level for your recipe.",
            [
                list(difficulty_dictionary.keys())[0],
                list(difficulty_dictionary.keys())[1],
                list(difficulty_dictionary.keys())[2]
            ],
            captions=[
                difficulty_dictionary["Quick & Easy"]["description"],
                difficulty_dictionary["Intermediate"]["description"],
                difficulty_dictionary["Professional"]["description"]
            ],
            index=list(difficulty_dictionary).index(st.session_state.selected_difficulty)
        )

        st.session_state.exclusions = st.text_area(
            "Any ingredients you want to exclude?",
            value=st.session_state.exclusions,
            placeholder="gluten, dairy, nuts, cilantro",
        )

        fancy_exclusions = ""

        if st.session_state.selected_difficulty == "Professional":
            exclude_fancy = st.checkbox(
                "Exclude cliche professional ingredients? (gold leaf, truffle, edible flowers, microgreens)",
                value=True)
            if exclude_fancy:
                fancy_exclusions = "gold leaf, truffle, edible flowers, microgreens, gold dust"

        user_inputs = {
            "user_direction": st.session_state.user_direction,
            "exclusions": f"{st.session_state.exclusions}, {fancy_exclusions}".strip(", "),
            "serving_size": st.session_state.serving_size,
            "difficulty": st.session_state.selected_difficulty
        }

        button_cols_submit = st.columns([1, 1, 4])
        with button_cols_submit[0]:
            st.button(label='Submit', on_click=generate_recipe, kwargs=dict(user_inputs=user_inputs), type="primary",
                      use_container_width=True)
        with button_cols_submit[1]:
            st.button(label='Reset', on_click=clear_inputs, type="secondary", use_container_width=True)
        with button_cols_submit[2]:
            st.empty()


        def create_safe_filename(recipe_name):
            # format and generate random URL-safe text string
            safe_name = recipe_name.lower()
            safe_name = safe_name.replace(" ", "_")
            safe_name = re.sub(r"[^a-zA-Z0-9_]", "", safe_name)
            safe_name = (safe_name[:50]) if len(safe_name) > 50 else safe_name
            unique_token = secrets.token_hex(8)
            safe_filename = f"{unique_token}_{safe_name}"
            return safe_filename


        def save_recipe():
            with st.spinner('WAIT SAVING YOUR DISH...'):
                filename = create_safe_filename(recipe["name"])
                os.makedirs('data', exist_ok=True)
                with open(f'./data/{filename}.pkl', 'wb') as f:
                    pickle.dump(recipe, f)
                st.session_state.recipe_saved = True


        if st.session_state.recipe is not None:
            st.divider()
            print(st.session_state.recipe)
            recipe = json.loads(st.session_state.recipe)
            if not st.session_state.recipe_saved:
                show_recipe(recipe)
            recipe['timestamp'] = str(datetime.datetime.now())
            if st.session_state.recipe_saved == True:
                disable_button = True
            else:
                disable_button = False
            button_cols_save = st.columns([1, 1, 4])
            with button_cols_save[0]:
                st.button("Save Recipe", on_click=save_recipe, disabled=disable_button, type="primary")
            with button_cols_save[1]:
                st.empty()
            with button_cols_save[2]:
                st.empty()
            if st.session_state.recipe_saved == True:
                st.success("Recipe Saved!")
    with saved:
        st.title("Saved Recipes")
        with st.spinner('LOADING YOUR RECIPE...'):
            def load_saved_recipes_from_pickle(directory_path):
                os.makedirs('data', exist_ok=True)
                recipes = []
                # Iterate through all files in the directory
                for filename in os.listdir(directory_path):
                    if filename.endswith('.pkl'):
                        file_path = os.path.join(directory_path, filename)
                        with open(file_path, 'rb') as file:
                            recipe = pickle.load(file)
                            recipes.append(recipe)
                return recipes


            # get all saved files
            directory_path = 'data'
            recipes = load_saved_recipes_from_pickle(directory_path)
            print(recipes)

            cols = st.columns([4, 1])
            with cols[1]:
                user_sort = st.selectbox("Sort", ('Recent', 'Oldest', 'A-Z', 'Z-A', 'Random'))
                if user_sort == 'Recent':
                    recipes.sort(key=lambda x: x['timestamp'], reverse=True)
                elif user_sort == 'Oldest':
                    recipes.sort(key=lambda x: x['timestamp'])
                elif user_sort == 'A-Z':
                    recipes.sort(key=lambda x: x['name'])
                elif user_sort == 'Z-A':
                    recipes.sort(key=lambda x: x['name'], reverse=True)
                elif user_sort == 'Random':
                    recipes.sort(key=lambda x: x['file'])
            with cols[0]:
                user_search = st.selectbox("Search Recipes", [""] + [recipe['name'] for recipe in recipes])
            st.write("")  # just some space

            if user_search != "":
                st.divider()
                filtered_recipes = [recipe for recipe in recipes if recipe['name'] == user_search]
                if filtered_recipes:
                    show_recipe(filtered_recipes[0])
                else:
                    st.write("No recipe found.")
else:
    st.markdown("# :eyes: PEOPLE SEARCHING FOR...")
    print(st.session_state.ind)
    flattened_indices = list(set(index for sublist in st.session_state.ind for index in sublist))
    df = items_dict.iloc[flattened_indices]
    if st.session_state.ind != []:
        with st.spinner("LET'S SEE, WHAT PEOPLE WANT..."):
            cont1 = st.container(border=True, height=500)
            with cont1:
                wc = wordcloud.WordCloud(width=1000, height=320, background_color='white').generate(
                    ' '.join(df['CATEGORY']))

                # Display word cloud
                fig, ax = plt.subplots()
                ax.imshow(wc, interpolation='bilinear')
                ax.axis('off')

                st.pyplot(fig, use_container_width=True)

            with st.container(border=True, height=500):
                col1, col2 = st.columns(2)
                with col1:
                    st.markdown('## DEMAND AS PER CATEGORY')
                    with plt.style.context('Solarize_Light2'):
                        fig0 = px.pie(df, names='CATEGORY', hole=0.5)
                        fig0.update_traces(text=df['CATEGORY'], textposition='outside')
                        st.plotly_chart(fig0, use_container_width=True)

                with col2:
                    st.markdown('## DEMAND AS PER AVAILABILITY')
                    with plt.style.context('Solarize_Light2'):
                        fig1 = px.pie(df, names='AVAILABILITY', hole=0.5)
                        fig1.update_traces(text=df['AVAILABILITY'], textposition='outside')
                        st.plotly_chart(fig1, use_container_width=True)

            with st.container(border=True, height=620):
                st.markdown('## :deciduous_tree: Hierarchical view of Sales using TreeMap')
                fig4 = px.treemap(df, path=['CATEGORY', 'AVAILABILITY','PRODUCT'],color='AVAILABILITY')
                fig4.update_layout(height=550)
                st.plotly_chart(fig4, use_container_width=True,scrolling=False)

            with st.container(border=True, height=500):
                fig = px.bar(df,
                             y='CATEGORY',
                             x='BREADCRUMBS',
                             color='BRAND',
                             # title='Gross Sales of Every Segment Country-wise',
                             barmode='relative',
                             title="Segment Vs. Gross Sales"
                             )
                st.plotly_chart(fig, use_container_width=True)

            @st.cache_resource
            def pygwalerapp(df):
                return StreamlitRenderer(df)
            # st.markdown("## :desktop_computer: DO YOUR OWN RESEARCH...")
            with st.expander("## :desktop_computer: LOOKING FOR SOMETHING ELSE..."):
                pygwalerapp(df).explorer()