# -*- coding: utf-8 -*-
__company__ = ''
__project__ = 'Observatory News - Final Project'
__author__ = 'Strauss'
__intial__ = '2023-08-07'

import sys
import os
import pandas as pd
import numpy as np
import streamlit as st
from database.data import Data
from utils.graphs import Altair
from database.data import Data
import glob
from datasets import load_dataset
from huggingface_hub import login, interpreter_login

from bertopic import BERTopic
from sentence_transformers import SentenceTransformer



import os

sys.path.append(os.path.dirname(os.getcwd()))

## Label Metrics Size
st.markdown("""<style>[data-testid="stMetricValue"] {font-size: 20px;}</style>""", unsafe_allow_html=True,)

data = Data()


def load_ds(media):
    dir = '{0}/{1}'.format(data.pth_data, media)
    st.info(dir)
    files = glob.glob(dir + '/*.csv')
    df_list = (pd.read_csv(file) for file in files)
    st.info(files)
    df = pd.concat(df_list, ignore_index=True)
    st.info('Loaded {0} rows and {1} columns'.format(df.shape[0], df.shape[1]))
    return df


def load_hugging_face_ds(media):

    # If the dataset is gated/private, make sure you have run huggingface-cli login
    dataset = load_dataset("strauss-oak/observatory-brazilian-news")

    dir = '{0}{1}/202101.csv'.format(data.pth_data, media)
    st.info(dir)
    dataset = load_dataset('csv', data_files=[dir], delimiter=',')
    return dataset


@st.cache_resource
def load_embeddings():
    media = 'globo'
    mdl_st_nm = 'paraphrase-multilingual-mpnet-base-v2'
    pth_gd_embedding = '/home/user/app/embeddings'
    st.info(os.getcwd())
    embedding_nm = 'embedding_{0}_{1}.npy'.format(media, mdl_st_nm)
    embeddings = np.load('{0}/{1}'.format(pth_gd_embedding, embedding_nm), allow_pickle=True)
    st.info('Embedding da mídia {0} carregada com sucesso com tamanho {1}'.format(media, embeddings.shape))
    return embeddings


@st.cache_resource
def load_topic_model():
    access_token = os.getenv('EMB_TKN')
    login(token=access_token)
    mdl_bertopic_globo = BERTopic.load("strauss-oak/observatory-news-topic-model")
    st.info(mdl_bertopic_globo)
    return mdl_bertopic_globo

@st.cache_resource
def load_sbert():
    access_token = os.getenv('EMB_TKN')
    login(token=access_token)
    mdl_sbert = SentenceTransformer.load('strauss-oak/mdl_bertopic_globo')
    st.info(mdl_sbert)
    return mdl_sbert


#embeddings = load_embeddings()
mdl_bertopic_globo = load_topic_model()
mdl_sbert = load_sbert()

c1 = st.container()
with c1:
    topics_map = {-1: 'nao agrupado', 0: 'Pandemia Covid-19', 1: 'Política', 2: 'Segurança Pública', 3: 'Música',
                  4: 'Guerra - Rússia', 5: 'Clima Tempo', 6: 'Combustíveis', 7: 'Medicina', 8: 'Economia',
                  9: 'Loterias', 10: 'Aeronáutica', 11: 'Tecnologia', 12: 'Tributações', 13: 'Cotidiano',
                  14: 'Ecologia', 15: 'Guerra Palestina', 16: 'Mercado de trabalho', 17: 'Diversidade de gênero',
                  18: 'Nascimento'}
    st.info(topics_map)

    news_text = st.text_input('Texto da notícia', 'Digite o texto da notícia do g1')
    news = []
    news.append(news_text)

    if st.button('Identificar Tópico da notícia:'):
        embeddings_ = mdl_sbert.encode(news)
        topics, probs = mdl_bertopic_globo.transform(news, embeddings_)
        st.info(topics)
        st.info(topics_map.get(topics[0]))