import streamlit as st
import pandas as pd
from st_aggrid import AgGrid, GridOptionsBuilder, GridUpdateMode
from test import predict_hoax, evaluate_model_performance
from load_model import load_model
from styles import COMMON_CSS
from google.cloud import storage
from io import StringIO
import os
from datetime import datetime
import pytz
def save_corrections_to_gcs(bucket_name, file_name, correction_data):
client = storage.Client()
bucket = client.bucket(bucket_name)
blob = bucket.blob(file_name)
# Check if the blob (file) exists
if blob.exists():
existing_data = blob.download_as_string().decode('utf-8')
existing_df = pd.read_csv(StringIO(existing_data))
else:
# Create a new DataFrame if the file does not exist
existing_df = pd.DataFrame(columns=['Timestamp', 'Title', 'Content', 'Prediction', 'Correction'])
new_data_df = pd.DataFrame(correction_data)
updated_df = pd.concat([existing_df, new_data_df], ignore_index=True)
updated_csv_data = updated_df.to_csv(index=False)
blob.upload_from_string(updated_csv_data, content_type='text/csv')
def load_data(file):
return pd.read_csv(file)
def show_deteksi_upload():
st.markdown(COMMON_CSS, unsafe_allow_html=True)
st.markdown("Pilih Model
", unsafe_allow_html=True)
selected_model = st.selectbox(
"",
[
"cahya/bert-base-indonesian-522M",
"indobenchmark/indobert-base-p2",
"indolem/indobert-base-uncased",
"mdhugol/indonesia-bert-sentiment-classification"
],
key="model_selector_upload"
)
tokenizer, model = load_model(selected_model)
st.markdown("Unggah File Disini (format: Title; Content)
", unsafe_allow_html=True)
uploaded_file = st.file_uploader("", type="csv")
if 'df' not in st.session_state:
st.session_state.df = None
if uploaded_file is not None:
df = load_data(uploaded_file)
df.index = df.index + 1
st.markdown("Data yang Diunggah
", unsafe_allow_html=True)
grid_options = GridOptionsBuilder.from_dataframe(df)
grid_options.configure_pagination(paginationAutoPageSize=False, paginationPageSize=10)
grid_options.configure_default_column(cellStyle={'fontSize': '12px'})
gridOptions = grid_options.build()
AgGrid(
df,
gridOptions=gridOptions,
update_mode=GridUpdateMode.VALUE_CHANGED,
use_container_width=True
)
if st.button("Deteksi", key="detect_upload"):
try:
df['Detection'] = df.apply(lambda row: predict_hoax(row['Title'], row['Content']), axis=1)
df['Correction'] = False
st.session_state.df = df.copy()
except Exception as e:
st.error(f"Terjadi kesalahan saat deteksi: {e}")
if st.session_state.df is not None:
st.markdown("Hasil Deteksi
", unsafe_allow_html=True)
cols = ['Correction', 'Detection'] + [col for col in st.session_state.df.columns if col not in ['Correction', 'Detection', 'Label_id']]
st.session_state.df = st.session_state.df[cols]
if 'checked_corrections' not in st.session_state:
st.session_state.checked_corrections = []
st.session_state.df['Correction'] = st.session_state.df.index.isin(st.session_state.checked_corrections)
grid_options = GridOptionsBuilder.from_dataframe(st.session_state.df)
grid_options.configure_pagination(paginationAutoPageSize=False, paginationPageSize=10)
grid_options.configure_default_column(editable=True, groupable=True)
grid_options.configure_column('Correction', editable=True, type='boolean', cellRenderer='agCheckboxCellRenderer', cellEditor='agCheckboxCellEditor', singleClickEdit=True)
gridOptions = grid_options.build()
grid_response = AgGrid(
st.session_state.df,
gridOptions=gridOptions,
update_mode=GridUpdateMode.VALUE_CHANGED,
key="detection_grid"
)
if grid_response['data'] is not None:
edited_df = pd.DataFrame(grid_response['data'])
edited_df.index = st.session_state.df.index
st.session_state.checked_corrections = edited_df[edited_df['Correction']].index.tolist()
st.session_state.df = edited_df.copy()
corrected_indices = edited_df[edited_df['Correction']].index
edited_df['Correction'] = edited_df.apply(lambda row:
'HOAX' if (row['Detection'] == 'NON-HOAX' and row['Correction']) else
('NON-HOAX' if (row['Detection'] == 'HOAX' and row['Correction']) else row['Detection']),
axis=1
)
st.session_state.df = edited_df.copy()
st.session_state.corrected_indices = corrected_indices
corrected_df = st.session_state.df.loc[st.session_state.corrected_indices].copy()
if not corrected_df.empty:
corrected_df['Correction'] = corrected_df.apply(lambda row:
'HOAX' if (row['Detection'] == 'NON-HOAX' and row['Correction']) else
('NON-HOAX' if (row['Detection'] == 'HOAX' and row['Correction']) else row['Detection']),
axis=1
)
# Tambahkan Timestamp hanya untuk penyimpanan
wib = pytz.timezone('Asia/Jakarta')
corrected_df['Timestamp'] = datetime.now(wib).strftime('%Y-%m-%d %H:%M:%S')
cols = ['Timestamp', 'Correction', 'Detection', 'Label_id', 'Label', 'Title', 'Content', 'Fact', 'References', 'Classification', 'Datasource']
cols = [col for col in cols if col in corrected_df.columns]
corrected_df_to_display = corrected_df[cols]
st.markdown("Data yang Dikoreksi
", unsafe_allow_html=True)
st.dataframe(corrected_df_to_display, use_container_width=True, hide_index=True)
else:
st.write("Tidak ada data yang dikoreksi.")
if st.button("Simpan", key="corrected_data"):
if 'df' in st.session_state and 'corrected_indices' in st.session_state:
corrected_df = st.session_state.df.loc[st.session_state.corrected_indices].copy()
wib = pytz.timezone('Asia/Jakarta')
corrected_df['Timestamp'] = datetime.now(wib).strftime('%Y-%m-%d %H:%M:%S')
corrected_df = corrected_df.rename(columns={'Detection': 'Prediction'})
cols = ['Timestamp', 'Title', 'Content', 'Prediction', 'Correction']
corrected_df = corrected_df[cols]
if not corrected_df.empty:
# Define GCS bucket and file name
bucket_name = "dashboardindohoax-bucket"
file_name = "koreksi_pengguna.csv"
# Convert DataFrame to list of dicts for GCS
correction_data = corrected_df.to_dict(orient='records')
# Save corrected data to GCS
save_corrections_to_gcs(bucket_name, file_name, correction_data)
st.success("Data telah disimpan.")
st.session_state.corrected_df = corrected_df
else:
st.warning("Tidak ada data yang dikoreksi untuk disimpan.")
else:
st.warning("Data deteksi tidak ditemukan.")