AI-Data-Cleaner / app.py
reab5555's picture
Update app.py
d7ffd4d verified
raw
history blame
2.48 kB
import gradio as gr
import pandas as pd
from clean import clean_data
from report import create_full_report, REPORT_DIR
import os
import tempfile
def clean_and_visualize(file, progress=gr.Progress()):
# Load the data
df = pd.read_csv(file.name)
# Clean the data
cleaned_df = None
nonconforming_cells_before = None
process_times = None
removed_columns = None
removed_rows = None
for progress_value, status_text in clean_data(df):
if isinstance(status_text, tuple):
cleaned_df, nonconforming_cells_before, process_times, removed_columns, removed_rows = status_text
progress(progress_value, desc="Cleaning completed")
else:
progress(progress_value, desc=status_text)
# Generate full visualization report
create_full_report(
df,
cleaned_df,
nonconforming_cells_before,
process_times,
removed_columns,
removed_rows
)
# Save cleaned DataFrame to a temporary CSV file
with tempfile.NamedTemporaryFile(delete=False, suffix='.csv') as tmp_file:
cleaned_df.to_csv(tmp_file.name, index=False)
cleaned_csv_path = tmp_file.name
# Collect all generated images
image_files = [os.path.join(REPORT_DIR, f) for f in os.listdir(REPORT_DIR) if f.endswith('.png')]
return cleaned_csv_path, image_files
def launch_app():
with gr.Blocks() as app:
gr.Markdown("# AI Data Cleaner")
with gr.Row():
file_input = gr.File(label="Upload CSV File")
with gr.Row():
clean_button = gr.Button("Start Cleaning")
with gr.Row():
progress_bar = gr.Progress()
with gr.Row():
cleaned_file_output = gr.File(label="Download Cleaned CSV", visible=False)
with gr.Row():
output_gallery = gr.Gallery(label="Visualization Results", show_label=True, elem_id="gallery", columns=[2],
rows=[2], object_fit="contain", height="auto")
def process_and_show_download(file):
cleaned_csv_path, image_files = clean_and_visualize(file, progress=progress_bar)
return gr.File.update(value=cleaned_csv_path, visible=True), image_files
clean_button.click(
fn=process_and_show_download,
inputs=file_input,
outputs=[cleaned_file_output, output_gallery]
)
app.launch()
if __name__ == "__main__":
launch_app()