import re
import pandas as pd

def markdown_to_csv(llm_output):
    # Find all Markdown tables
    tables = re.findall(r"(\|(?:[^\n]+\|)+\n\|(?:\s*-+\s*\|)+\n(?:\|(?:[^\n]+\|)+\n)+)", llm_output)
    
    dataframes = []

    if tables:  # Ajouté pour éviter le 'NoneType' issue
        for table in tables:
            # Split lines and extract columns
            lines = table.strip().split("\n")
            headers = [col.strip() for col in lines[0].split("|")[1:-1]]  # Headers
            data_rows = [line.split("|")[1:-1] for line in lines[2:]]  # Ignore separator row

            # Clean and validate rows
            cleaned_data = [[col.strip() for col in row] for row in data_rows if len(row) == len(headers)]

            if cleaned_data:
                df = pd.DataFrame(cleaned_data, columns=headers)
                dataframes.append(df)

    return dataframes  # <- Corrigé : en dehors de la boucle