Skip to main content

​
Overview

The Extrator de Tarefas Auvo uses keyword-based filtering to identify actionable tasks from CSV and Excel reports. The filtering mechanism uses Python’s pandas library with case-insensitive regex pattern matching.

​
How It Works

The filtering process is handled by the processar_arquivo() function in app.py:32: 
def processar_arquivo(file, palavras_chave):
    """Processa o arquivo CSV ou Excel e retorna os dados filtrados"""
    filename = file.filename.lower()
    
    # Decide como ler o arquivo com base na extensão
    if filename.endswith('.csv'):
        df = pd.read_csv(file, skiprows=5)
    elif filename.endswith(('.xls', '.xlsx')):
        df = pd.read_excel(file, skiprows=5, engine='openpyxl')
    else:
        raise ValueError("Formato de arquivo não suportado")

    regex_busca = '|'.join(palavras_chave)
    
    coluna_descricao = 'Relato'
    necessidades = df[df[coluna_descricao].astype(str).str.contains(
        regex_busca, case=False, na=False
    )].copy()
    
    colunas_resultado = ['Data', 'Cliente', 'Endereco', 'OS Digital', 'Relato']
    return df, necessidades[colunas_resultado]

​
Key Features

Multi-format Support

Supports CSV, XLS, and XLSX files with automatic format detection

Regex Patterns

Uses pipe-separated regex for efficient multi-keyword matching

Case-insensitive

Matches keywords regardless of capitalization

Null-safe

Handles missing values with na=False parameter

​
Filtering Logic

1

File Format Detection

The system automatically detects file format based on extension (.csv, .xls, .xlsx) and uses the appropriate pandas reader.
2

Skip Header Rows

Both CSV and Excel files skip the first 5 rows (skiprows=5) to handle Auvo report formatting.
3

Regex Pattern Construction

Keywords are joined with the pipe operator (|) to create a single regex pattern: solicitar peça|quebrado|trocar cabo
4

Column Filtering

The ‘Relato’ (Report) column is searched using str.contains() with case=False for case-insensitive matching.
5

Result Extraction

Only relevant columns are returned: Data, Cliente, Endereco, OS Digital, Relato

​
Default Keywords

The application comes with pre-configured keywords optimized for identifying maintenance and repair tasks: 
default_keywords = [
    'solicitar peça',
    'quebrado',
    'quebrada',
    'quebrados',
    'orçamento',
    'danificada',
    'danificado',
    'danificados',
    'danificadas',
    'trocar cabo',
    'soldar',
    'trocar',
    'instalar',
    'orçamento'
]
These keywords are defined in app.py:115 and stored in the Flask session as custom_keywords.

​
Customizing Keywords

Users can customize keywords through the /config route, which accepts comma-separated values and stores them in the session.

​
Configuration Route

From app.py:108: 
@app.route('/config', methods=['GET', 'POST'])
def config():
    if request.method == 'POST':
        keywords = request.form.get('keywords', '').split(',')
        session['custom_keywords'] = [k.strip() for k in keywords if k.strip()]
        return redirect(url_for('index'))
    
    current_keywords = session.get('custom_keywords', [...])
    return render_template('config.html', keywords=', '.join(current_keywords))

​
Technical Details

​
Pattern Matching Behavior

  • Substring matching: The regex matches keywords anywhere in the text
  • No word boundaries: trocar will match “trocar”, “trocando”, “retrocado”
  • OR logic: Any keyword match includes the row in results

​
Performance Considerations

For very large files (10,000+ rows), the regex matching is performed in-memory using pandas vectorized operations, which is efficient but memory-intensive.

​
Column Output

Filtered results include exactly 5 columns:
ColumnDescription
DataTask date
ClienteClient name
EnderecoService address
OS DigitalDigital work order (with clickable links)
RelatoTask report/description

​
Error Handling

The function raises ValueError for unsupported file formats: 
raise ValueError("Formato de arquivo não suportado. Use .csv, .xls ou .xlsx")
This error is caught in the upload route and displayed to the user.

Build docs developers (and LLMs) love

Get started for freeTalk to us