feat: Implementar PWA, Analytics, Reportes PDF y mejoras OCR

FASE 1 - PWA y Frontend:
- Crear templates/base.html, dashboard.html, analytics.html, executive.html
- Crear static/css/main.css con diseño responsivo
- Agregar static/js/app.js, pwa.js, camera.js, charts.js
- Implementar manifest.json y service-worker.js para PWA
- Soporte para captura de tickets desde cámara móvil

FASE 2 - Analytics:
- Crear módulo analytics/ con predictions.py, trends.py, comparisons.py
- Implementar predicción básica con promedio móvil + tendencia lineal
- Agregar endpoints /api/analytics/trends, predictions, comparisons
- Integrar Chart.js para gráficas interactivas

FASE 3 - Reportes PDF:
- Crear módulo reports/ con pdf_generator.py
- Implementar SalesReportPDF con generar_reporte_diario y ejecutivo
- Agregar comando /reporte [diario|semanal|ejecutivo]
- Agregar endpoints /api/reports/generate y /api/reports/download

FASE 4 - Mejoras OCR:
- Crear módulo ocr/ con processor.py, preprocessor.py, patterns.py
- Implementar AmountDetector con patrones múltiples de montos
- Agregar preprocesador adaptativo con pipelines para diferentes condiciones
- Soporte para corrección de rotación (deskew) y threshold Otsu

Dependencias agregadas:
- reportlab, matplotlib (PDF)
- scipy, pandas (analytics)
- imutils, deskew, cachetools (OCR)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
2026-01-19 03:26:16 +00:00
parent ed1658eb2b
commit 9936deaa90
25 changed files with 5501 additions and 282 deletions

View File

@@ -0,0 +1,258 @@
"""
Amount detection for Sales Bot OCR
Improved detection of total amounts from ticket text
"""
import re
import logging
from typing import Dict, List, Optional, Tuple
logger = logging.getLogger(__name__)
# Amount patterns in priority order
PATTERNS = [
# Explicit total patterns (highest priority)
(r'total\s*a\s*pagar\s*:?\s*\$?\s*([\d,]+[\s\.]?\d{0,2})', 'total_a_pagar', 1),
(r'gran\s*total\s*:?\s*\$?\s*([\d,]+[\s\.]?\d{0,2})', 'gran_total', 2),
(r'total\s+final\s*:?\s*\$?\s*([\d,]+[\s\.]?\d{0,2})', 'total_final', 3),
(r'(?:^|\n)\s*total\s*:?\s*\$?\s*([\d,]+[\s\.]?\d{0,2})', 'total', 4),
# Payment related
(r'a\s*cobrar\s*:?\s*\$?\s*([\d,]+[\s\.]?\d{0,2})', 'a_cobrar', 5),
(r'importe\s*(?:total)?\s*:?\s*\$?\s*([\d,]+[\s\.]?\d{0,2})', 'importe', 6),
(r'monto\s*(?:total)?\s*:?\s*\$?\s*([\d,]+[\s\.]?\d{0,2})', 'monto', 7),
(r'suma\s*:?\s*\$?\s*([\d,]+[\s\.]?\d{0,2})', 'suma', 8),
(r'pago\s*:?\s*\$?\s*([\d,]+[\s\.]?\d{0,2})', 'pago', 9),
# Subtotal (lower priority - may need to add tax)
(r'subtotal\s*:?\s*\$?\s*([\d,]+[\s\.]?\d{0,2})', 'subtotal', 10),
# Generic currency patterns (lowest priority)
(r'\$\s*([\d,]+\.\d{2})\s*(?:\n|$)', 'monto_linea', 11),
]
# Words that indicate a line is NOT a total (negative patterns)
EXCLUSION_WORDS = [
'cambio', 'efectivo', 'pago con', 'tarjeta', 'recibido',
'iva', 'impuesto', 'descuento', 'ahorro', 'puntos'
]
class AmountDetector:
"""
Detects and extracts monetary amounts from ticket text.
Uses multiple patterns and heuristics to find the most likely total.
"""
def __init__(self):
self.patterns = PATTERNS
self.min_amount = 1 # Minimum valid amount
self.max_amount = 1000000 # Maximum valid amount
def detectar_monto(self, texto: str) -> Optional[Dict]:
"""
Detecta el monto total del ticket.
Args:
texto: Texto extraído del ticket
Returns:
dict con monto, tipo, patron, y confianza, o None si no se encuentra
"""
texto_lower = texto.lower()
resultados = []
for patron, tipo, prioridad in self.patterns:
matches = re.findall(patron, texto_lower, re.IGNORECASE | re.MULTILINE)
for match in matches:
# Skip if match is in an exclusion context
if self._is_excluded(texto_lower, match):
continue
monto = self._normalizar_monto(match)
if self.min_amount <= monto <= self.max_amount:
# Calculate confidence based on pattern type and context
confianza = self._calcular_confianza(texto_lower, match, tipo)
resultados.append({
'monto': monto,
'tipo': tipo,
'patron': patron,
'prioridad': prioridad,
'confianza': confianza
})
if not resultados:
# Try to find the largest amount as fallback
return self._fallback_detection(texto)
# Sort by priority (lower is better) then by confidence (higher is better)
resultados.sort(key=lambda x: (x['prioridad'], -x['confianza']))
# Return the best match
best = resultados[0]
return {
'monto': best['monto'],
'tipo': best['tipo'],
'patron': best['patron'],
'confianza': best['confianza']
}
def _normalizar_monto(self, monto_str: str) -> float:
"""
Normaliza string de monto a float.
Handles various formats:
- 1,234.56 (US/Mexico format)
- 1234.56
- 1 234.56 (space separator)
- 1234,56 (European format)
"""
if not monto_str:
return 0.0
# Remove currency symbols and whitespace
monto = monto_str.strip().replace('$', '').replace(' ', '')
# Handle different decimal separators
# If there's both comma and dot, determine which is decimal
if ',' in monto and '.' in monto:
# US/Mexico format: 1,234.56
monto = monto.replace(',', '')
elif ',' in monto:
# Could be European (1234,56) or thousand separator (1,234)
parts = monto.split(',')
if len(parts) == 2 and len(parts[1]) == 2:
# European format
monto = monto.replace(',', '.')
else:
# Thousand separator
monto = monto.replace(',', '')
try:
return float(monto)
except ValueError:
return 0.0
def _is_excluded(self, texto: str, match: str) -> bool:
"""
Checks if the match appears in an exclusion context.
"""
# Find the line containing this match
for linea in texto.split('\n'):
if match in linea:
linea_lower = linea.lower()
for exclusion in EXCLUSION_WORDS:
if exclusion in linea_lower:
return True
return False
def _calcular_confianza(self, texto: str, match: str, tipo: str) -> float:
"""
Calculates confidence score for a match.
Returns value between 0.0 and 1.0
"""
confianza = 0.5 # Base confidence
# Higher confidence for explicit total patterns
if tipo in ['total_a_pagar', 'gran_total', 'total_final']:
confianza += 0.3
elif tipo == 'total':
confianza += 0.2
# Higher confidence if near end of text
position = texto.find(match)
text_length = len(texto)
if position > text_length * 0.6: # In last 40% of text
confianza += 0.1
# Higher confidence if followed by payment info
after_match = texto[texto.find(match) + len(match):texto.find(match) + len(match) + 50]
if any(word in after_match.lower() for word in ['efectivo', 'tarjeta', 'cambio', 'gracias']):
confianza += 0.1
return min(confianza, 1.0)
def _fallback_detection(self, texto: str) -> Optional[Dict]:
"""
Fallback detection when standard patterns fail.
Looks for the largest reasonable amount in the text.
"""
# Find all currency-like numbers
all_amounts = re.findall(r'\$?\s*([\d,]+\.?\d{0,2})', texto)
valid_amounts = []
for amount_str in all_amounts:
amount = self._normalizar_monto(amount_str)
if self.min_amount <= amount <= self.max_amount:
valid_amounts.append(amount)
if valid_amounts:
# Return the largest amount (likely the total)
max_amount = max(valid_amounts)
return {
'monto': max_amount,
'tipo': 'fallback_max',
'patron': 'heuristic',
'confianza': 0.3
}
return None
def detectar_multiples_montos(self, texto: str) -> List[Dict]:
"""
Detecta todos los montos en el texto.
Useful for itemized receipts.
Returns:
Lista de diccionarios con monto y contexto
"""
texto_lower = texto.lower()
resultados = []
# Find all lines with amounts
lineas = texto.split('\n')
for linea in lineas:
matches = re.findall(r'\$?\s*([\d,]+\.?\d{0,2})', linea)
for match in matches:
monto = self._normalizar_monto(match)
if self.min_amount <= monto <= self.max_amount:
resultados.append({
'monto': monto,
'contexto': linea.strip(),
'es_total': 'total' in linea.lower()
})
return resultados
def detectar_monto(texto: str) -> Optional[Dict]:
"""
Convenience function to detect amount from text.
Args:
texto: Ticket text
Returns:
Dict with monto, tipo, patron, confianza or None
"""
detector = AmountDetector()
return detector.detectar_monto(texto)
def normalizar_monto(monto_str: str) -> float:
"""
Convenience function to normalize amount string.
Args:
monto_str: Amount as string
Returns:
Amount as float
"""
detector = AmountDetector()
return detector._normalizar_monto(monto_str)