app.py

import streamlit as st
import pandas as pd
from transformers import BartForConditionalGeneration, TapexTokenizer, T5ForConditionalGeneration, T5Tokenizer
from prophet import Prophet
import datetime
import sentencepiece as spm

st.markdown("""
<style>
@import url('https://fonts.googleapis.com/css2?family=Kanit:wght@700&display=swap');
</style>
<div style='display: flex; flex-direction: column; align-items: flex-start;'>
    <div style='display: flex; align-items: center;'>
        <div style='width: 20px; height: 40px; background-color: green; margin-right: 1px;'></div>
        <div style='width: 20px; height: 40px; background-color: red; margin-right: 1px;'></div>
        <div style='width: 20px; height: 40px; background-color: yellow; margin-right: 20px;'></div>
        <span style='font-size: 40px; font-weight: normal; font-family: "Kanit", sans-serif;'>PROTAX</span>
    </div>
    <div style='text-align: left; width: 100%;'>
        <span style='font-size: 20px; font-weight: normal; color: #333; font-family: "Kanit", sans-serif'>
        <strong>PRO</strong>phet & <strong>TA</strong>pex E<strong>X</strong>plorer</span>
    </div>
</div>
""", unsafe_allow_html=True)

# File upload interface
uploaded_file = st.file_uploader("Carregue um arquivo CSV ou XLSX", type=['csv', 'xlsx'])

if uploaded_file:
    if 'all_anomalies' not in st.session_state:
        with st.spinner('Aplicando modelo de série temporal...'):
            # Load the file into a DataFrame
            if uploaded_file.name.endswith('.csv'):
                df = pd.read_csv(uploaded_file, quotechar='"', encoding='utf-8')
            elif uploaded_file.name.endswith('.xlsx'):
                df = pd.read_excel(uploaded_file)

            # Data preprocessing for Prophet
            new_df = df.iloc[2:, 9:-1].fillna(0)
            new_df.columns = df.iloc[1, 9:-1]
            new_df.columns = new_df.columns.str.replace(r" \(\d+\)", "", regex=True)

            month_dict = {
                'Jan': '01', 'Fev': '02', 'Mar': '03', 'Abr': '04',
                'Mai': '05', 'Jun': '06', 'Jul': '07', 'Ago': '08',
                'Set': '09', 'Out': '10', 'Nov': '11', 'Dez': '12'
            }

            def convert_column_name(column_name):
                if column_name == 'Rótulos de Linha':
                    return column_name
                parts = column_name.split('/')
                month = parts[0].strip()
                year = parts[1].strip()
                year = ''.join(filter(str.isdigit, year))
                month_number = month_dict.get(month, '00')
                return f"{month_number}/{year}"

            new_df.columns = [convert_column_name(col) for col in new_df.columns]
            new_df.columns = pd.to_datetime(new_df.columns, errors='coerce')
            new_df.rename(columns={new_df.columns[0]: 'Rotulo'}, inplace=True)
            df_clean = new_df.copy()

            # Create an empty DataFrame to store all anomalies
            all_anomalies = pd.DataFrame()

            # Process each row in the DataFrame
            for index, row in df_clean.iterrows():
                data = pd.DataFrame({
                    'ds': [col for col in df_clean.columns if isinstance(col, pd.Timestamp)],
                    'y': row[[isinstance(col, pd.Timestamp) for col in df_clean.columns]].values
                })

                data = data[data['y'] > 0].reset_index(drop=True)
                if data.empty or len(data) < 2:
                    print(f"Skipping group {row['Rotulo']} because there are less than 2 non-zero observations.")
                    continue

                try:
                    model = Prophet(interval_width=0.95)
                    model.fit(data)
                except ValueError as e:
                    print(f"Skipping group {row['Rotulo']} due to error: {e}")
                    continue

                future = model.make_future_dataframe(periods=12, freq='M')
                forecast = model.predict(future)

                num_real = len(data)
                num_forecast = len(forecast)
                real_values = list(data['y']) + [None] * (num_forecast - num_real)
                forecast['real'] = real_values
                anomalies = forecast[(forecast['real'] < forecast['yhat_lower']) | (forecast['real'] > forecast['yhat_upper'])]

                anomalies['Group'] = row['Rotulo']
                all_anomalies = pd.concat([all_anomalies, anomalies[['ds', 'real', 'Group']]], ignore_index=True)

            # Store the result in session state
            all_anomalies.rename(columns={"ds": "datetime", "real": "monetary value", "Group": "group"}, inplace=True)
            all_anomalies = all_anomalies[all_anomalies['monetary value'].astype('float') >= 10,000,000.00]
            all_anomalies['monetary value'] = all_anomalies['monetary value'].apply(lambda x: f"{x:.2f}")
            all_anomalies.sort_values(by=['monetary value'], ascending=False, inplace=True)
            all_anomalies = all_anomalies.fillna('').astype(str)
            st.session_state['all_anomalies'] = all_anomalies

    # Load translation models
    pt_en_translator = T5ForConditionalGeneration.from_pretrained("unicamp-dl/translation-pt-en-t5")
    en_pt_translator = T5ForConditionalGeneration.from_pretrained("unicamp-dl/translation-en-pt-t5")
    tokenizer = T5Tokenizer.from_pretrained("unicamp-dl/translation-pt-en-t5")

    # Load TAPEX model
    tapex_model = BartForConditionalGeneration.from_pretrained("microsoft/tapex-large-finetuned-wtq")
    tapex_tokenizer = TapexTokenizer.from_pretrained("microsoft/tapex-large-finetuned-wtq")

    def translate(text, model, tokenizer, source_lang="pt", target_lang="en"):
        input_ids = tokenizer.encode(text, return_tensors="pt", add_special_tokens=True)
        outputs = model.generate(input_ids)
        translated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
        return translated_text

    def response(user_question, table_data):
        question_en = translate(user_question, pt_en_translator, tokenizer, source_lang="pt", target_lang="en")
        encoding = tapex_tokenizer(table=table_data, query=[question_en], padding=True, return_tensors="pt", truncation=True)
        outputs = tapex_model.generate(**encoding)
        response_en = tapex_tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
        response_pt = translate(response_en, en_pt_translator, tokenizer, source_lang="en", target_lang="pt")
        return response_pt

    # Streamlit interface
    st.dataframe(st.session_state['all_anomalies'].head())

    # Chat history
    if 'history' not in st.session_state:
        st.session_state['history'] = []

    user_question = st.text_input("Escreva sua questão aqui:", "")

    if user_question:
        st.session_state['history'].append(('👤', user_question))
        st.markdown(f"**👤 {user_question}**")
        
        bot_response = response(user_question, st.session_state['all_anomalies'])
        
        st.session_state['history'].append(('🤖', bot_response))
        st.markdown(f"<div style='text-align: right'>**🤖 {bot_response}**</div>", unsafe_allow_html=True)

    if st.button("Limpar"):
        st.session_state['history'] = []

    for sender, message in st.session_state['history']:
        if sender == '👤':
            st.markdown(f"**👤 {message}**")
        elif sender == '🤖':
            st.markdown(f"<div style='text-align: right'>**🤖 {message}**</div>", unsafe_allow_html=True)
else:
    st.warning("Por favor, carregue um arquivo CSV ou XLSX para começar.")