879 lines
34 KiB
Python
879 lines
34 KiB
Python
from fastapi import FastAPI, UploadFile, HTTPException, Body
|
||
from PIL import Image
|
||
import pytesseract
|
||
from doctr.models import ocr_predictor
|
||
from doctr.io import DocumentFile
|
||
from PyPDF2 import PdfReader
|
||
import camelot
|
||
import spacy
|
||
import logging
|
||
import io
|
||
from logging.handlers import RotatingFileHandler
|
||
import re
|
||
|
||
LOG_PATH = "/var/log/automation-service.log"
|
||
|
||
file_handler = RotatingFileHandler(
|
||
LOG_PATH,
|
||
maxBytes=10*1024*1024,
|
||
backupCount=5,
|
||
encoding="utf-8"
|
||
)
|
||
file_handler.setFormatter(logging.Formatter(
|
||
"%(asctime)s - %(levelname)s - %(name)s - %(message)s"
|
||
))
|
||
|
||
# class AHKParser:
|
||
# lab="AHK"
|
||
# def parse(self,text):
|
||
# r=empty_weight_report("AHK")
|
||
# inv=section(text,"INVOICE WEIGHTS","Bales Weighed")
|
||
# land=section(text,"Bales Weighed","Outturn")
|
||
# loss=section(text,"LOSS","Invoice average")
|
||
|
||
# r["report"]["reference"]=safe_search(r"(AHK\s*/\S+)",text)
|
||
# r["report"]["date"]=extract("Produced On",text)
|
||
|
||
# r["contract"]["invoice_no"]=extract("Client Reference",text)
|
||
# r["contract"]["origin"]=extract("Growth",text)
|
||
# r["contract"]["commodity"]="Raw Cotton"
|
||
|
||
# r["parties"]["seller"]=extract("Client",text)
|
||
# r["parties"]["buyer"]=extract("Buyer",text)
|
||
|
||
# r["shipment"]["vessel"]=extract("Vessel",text)
|
||
# r["shipment"]["bl_no"]=extract("B/L No",text)
|
||
# r["shipment"]["port_destination"]=extract("Destination",text)
|
||
# r["shipment"]["arrival_date"]=extract("Arrival Date",text)
|
||
# r["shipment"]["weighing_method"]=extract("Weighing method",text)
|
||
# r["shipment"]["bales"]=to_float(extract("Total Bales",text))
|
||
|
||
# r["weights"]["gross_landed_kg"]=to_float(extract("Gross",land))
|
||
# r["weights"]["tare_kg"]=to_float(extract("Tare",land))
|
||
# r["weights"]["net_landed_kg"]=to_float(extract("Net",land))
|
||
# r["weights"]["invoice_net_kg"]=to_float(extract("Net",inv))
|
||
# r["weights"]["gain_loss_kg"]=to_float(extract("kg",loss))
|
||
# r["weights"]["gain_loss_percent"]=to_float(extract("Percentage",loss))
|
||
# return r
|
||
# class AHKParser:
|
||
# lab = "AHK"
|
||
|
||
# def extract_table(self, text, headers):
|
||
# lines = [l.strip() for l in text.splitlines() if l.strip()]
|
||
# out = {}
|
||
# for h in headers:
|
||
# for i,l in enumerate(lines):
|
||
# if l == h:
|
||
# for j in range(i+1, i+8):
|
||
# if j < len(lines) and lines[j].startswith(":"):
|
||
# out[h] = lines[j][1:].strip()
|
||
# break
|
||
# return out
|
||
|
||
# def extract_weights(self, text):
|
||
# lines = [l.strip() for l in text.splitlines() if l.strip()]
|
||
# res = {}
|
||
# for i,l in enumerate(lines):
|
||
# if l == "Bales Weighed":
|
||
# headers = ["Bales","Gross","Tare","Net"]
|
||
# for h in headers:
|
||
# for j in range(i, i+20):
|
||
# if j < len(lines) and lines[j].startswith(":"):
|
||
# res[h] = lines[j][1:].replace("kg","").strip()
|
||
# break
|
||
# return res
|
||
|
||
# def parse(self, text):
|
||
# r = empty_weight_report("AHK")
|
||
|
||
# # report
|
||
# r["report"]["reference"] = safe_search(r"(AHK\s*/[A-Z0-9/]+)", text)
|
||
# r["report"]["date"] = safe_search(r"Produced On\s*([0-9A-Za-z ]+)", text)
|
||
|
||
# # contract
|
||
# r["contract"]["invoice_no"] = safe_search(r"Client Reference:\s*([A-Z0-9\- /]+)", text)
|
||
# r["contract"]["commodity"] = "Raw Cotton"
|
||
|
||
# # buyer
|
||
# r["parties"]["buyer"] = safe_search(r"Buyer:\s*([A-Z0-9 ().,-]+)", text)
|
||
|
||
# # shipment tables
|
||
# ship = self.extract_table(text, [
|
||
# "Total Bales","Vessel","Voy. No.","B/L No.","B/L Date","Destination"
|
||
# ])
|
||
# ship2 = self.extract_table(text, [
|
||
# "Growth","Arrival Date","First date of weighing",
|
||
# "Last Date of Weighing","Weighing method","Tare"
|
||
# ])
|
||
|
||
# r["shipment"]["bales"] = to_float(ship.get("Total Bales"))
|
||
# r["shipment"]["vessel"] = ship.get("Vessel")
|
||
# r["shipment"]["bl_no"] = ship.get("B/L No.")
|
||
# r["shipment"]["port_destination"] = ship.get("Destination")
|
||
# r["shipment"]["arrival_date"] = ship2.get("Arrival Date")
|
||
# r["shipment"]["weighing_method"] = ship2.get("Weighing method")
|
||
# r["contract"]["origin"] = ship2.get("Growth")
|
||
|
||
# # weights
|
||
# inv = self.extract_table(text, ["Bales","Gross","Tare","Net"])
|
||
# land = self.extract_weights(text)
|
||
|
||
# r["weights"]["invoice_net_kg"] = to_float(inv.get("Net"))
|
||
# r["weights"]["gross_landed_kg"] = to_float(land.get("Gross"))
|
||
# r["weights"]["tare_kg"] = to_float(land.get("Tare"))
|
||
# r["weights"]["net_landed_kg"] = to_float(land.get("Net"))
|
||
|
||
# # loss
|
||
# loss = section(text,"LOSS","Invoice average")
|
||
# r["weights"]["gain_loss_kg"] = to_float(safe_search(r"(-?\d+\.?\d*)\s*kg", loss))
|
||
# r["weights"]["gain_loss_percent"] = to_float(safe_search(r"Percentage\s*:\s*(-?\d+\.?\d*)", loss))
|
||
|
||
# return r
|
||
|
||
class AHKParser:
|
||
lab = "AHK"
|
||
|
||
def _lines(self, text):
|
||
return [l.strip() for l in text.splitlines() if l.strip()]
|
||
|
||
def _col_block(self, lines, labels, max_scan=30):
|
||
idx = [i for i,l in enumerate(lines) if l in labels]
|
||
if not idx:
|
||
return {} # << empêche le crash
|
||
start = max(idx) + 1
|
||
vals = []
|
||
for l in lines[start:start+max_scan]:
|
||
if l.startswith(":"):
|
||
v = l[1:].replace("kg","").strip()
|
||
vals.append(v)
|
||
if len(vals) == len(labels):
|
||
break
|
||
return dict(zip(labels, vals))
|
||
|
||
def parse(self, text):
|
||
L = self._lines(text)
|
||
r = empty_weight_report("AHK")
|
||
|
||
# report
|
||
r["report"]["reference"] = safe_search(r"(AHK\s*/[A-Z0-9/]+)", text)
|
||
r["report"]["date"] = safe_search(r"Produced On\s*([0-9A-Za-z ]+)", text)
|
||
|
||
# contract
|
||
r["contract"]["invoice_no"] = safe_search(r"Client Reference:\s*([A-Z0-9\- /]+)", text)
|
||
r["contract"]["commodity"] = "Raw Cotton"
|
||
|
||
# buyer
|
||
r["parties"]["buyer"] = safe_search(r"Buyer:\s*([A-Z0-9 ().,-]+)", text)
|
||
|
||
# shipment block 1
|
||
ship1 = self._col_block(L, [
|
||
"Total Bales","Vessel","Voy. No.","B/L No.","B/L Date","Destination"
|
||
])
|
||
|
||
# shipment block 2
|
||
ship2 = self._col_block(L, [
|
||
"Growth","Arrival Date","First date of weighing",
|
||
"Last Date of Weighing","Weighing method","Tare"
|
||
])
|
||
|
||
r["shipment"]["bales"] = to_float(ship1.get("Total Bales"))
|
||
r["shipment"]["vessel"] = ship1.get("Vessel")
|
||
r["shipment"]["bl_no"] = ship1.get("B/L No.")
|
||
r["shipment"]["port_destination"] = ship1.get("Destination")
|
||
r["shipment"]["arrival_date"] = ship2.get("Arrival Date")
|
||
r["shipment"]["weighing_method"] = ship2.get("Weighing method")
|
||
r["contract"]["origin"] = ship2.get("Growth")
|
||
|
||
# invoice weights
|
||
inv = self._col_block(L, ["Bales","Gross","Tare","Net"])
|
||
r["weights"]["invoice_net_kg"] = to_float(inv.get("Net"))
|
||
|
||
# landed weights
|
||
land = self._col_block(
|
||
self._lines(section(text,"Bales Weighed","Outturn")),
|
||
["Bales","Gross","Tare","Net"]
|
||
)
|
||
|
||
r["weights"]["gross_landed_kg"] = to_float(land.get("Gross"))
|
||
r["weights"]["tare_kg"] = to_float(land.get("Tare"))
|
||
r["weights"]["net_landed_kg"] = to_float(land.get("Net"))
|
||
|
||
# loss
|
||
loss = section(text,"LOSS","Invoice average")
|
||
r["weights"]["gain_loss_kg"] = to_float(safe_search(r"(-?\d+\.?\d*)\s*kg", loss))
|
||
r["weights"]["gain_loss_percent"] = to_float(safe_search(r"Percentage\s*:\s*(-?\d+\.?\d*)", loss))
|
||
|
||
return r
|
||
|
||
|
||
class IntertekParser:
|
||
lab="INTERTEK"
|
||
def parse(self,text):
|
||
r=empty_weight_report("INTERTEK")
|
||
pct=safe_search(r"([0-9.]+)\s*%",text)
|
||
|
||
r["report"]["reference"]=extract("Global Ref",text)
|
||
r["report"]["file_no"]=extract("Report / File No",text)
|
||
r["report"]["date"]=extract("Dated",text)
|
||
|
||
r["contract"]["contract_no"]=extract("Contract No",text)
|
||
r["contract"]["invoice_no"]=extract("Invoice No",text)
|
||
r["contract"]["origin"]=extract("Growth",text)
|
||
r["contract"]["commodity"]="Raw Cotton"
|
||
|
||
r["parties"]["buyer"]=extract("Buyer",text)
|
||
|
||
r["shipment"]["vessel"]=extract("Vessel",text)
|
||
r["shipment"]["bl_no"]=extract("B/L No",text)
|
||
r["shipment"]["arrival_date"]=extract("Arrival Date",text)
|
||
r["shipment"]["weighing_place"]=extract("Weighed at",text)
|
||
r["shipment"]["bales"]=to_float(extract("Invoice Quantity",text))
|
||
|
||
r["weights"]["gross_landed_kg"]=to_float(extract("Gross",text))
|
||
r["weights"]["tare_kg"]=to_float(extract("Invoice Tare",text))
|
||
r["weights"]["net_landed_kg"]=to_float(extract("Landed Weight",text))
|
||
r["weights"]["invoice_net_kg"]=to_float(extract("Invoice Weight",text))
|
||
r["weights"]["gain_loss_kg"]=to_float(extract("Gain",text))
|
||
r["weights"]["gain_loss_percent"]=to_float(pct)
|
||
return r
|
||
|
||
class RobertsonParser:
|
||
lab="ROBERTSON"
|
||
def parse(self,text):
|
||
r=empty_weight_report("ROBERTSON")
|
||
pct=safe_search(r"([0-9.]+)\s*%",text)
|
||
|
||
r["report"]["reference"]=extract("OUR REF",text)
|
||
r["report"]["date"]=extract("DATE",text)
|
||
|
||
r["contract"]["contract_no"]=extract("CONTRACT NO",text)
|
||
r["contract"]["invoice_no"]=extract("INVOICE NO",text)
|
||
r["contract"]["lc_no"]=extract("LIC NO",text)
|
||
r["contract"]["commodity"]="Raw Cotton"
|
||
|
||
r["parties"]["seller"]=extract("SELLER",text)
|
||
r["parties"]["buyer"]=extract("BUYER",text)
|
||
|
||
r["shipment"]["vessel"]=extract("NAME OF VESSEL",text)
|
||
r["shipment"]["port_loading"]=extract("SAILED FROM",text)
|
||
r["shipment"]["port_destination"]=extract("ARRIVED AT",text)
|
||
r["shipment"]["arrival_date"]=extract("DATE OF ARRIVAL",text)
|
||
r["shipment"]["weighing_place"]=extract("PLACE OF CONTROL",text)
|
||
r["shipment"]["bales"]=to_float(extract("CONSIGNMENT",text))
|
||
|
||
r["weights"]["gross_landed_kg"]=to_float(extract("GROSS",text))
|
||
r["weights"]["tare_kg"]=to_float(extract("TARE",text))
|
||
r["weights"]["net_landed_kg"]=to_float(extract("LANDED NET",text))
|
||
r["weights"]["invoice_net_kg"]=to_float(extract("INVOICE NET",text))
|
||
r["weights"]["gain_loss_kg"]=to_float(extract("GAIN",text))
|
||
r["weights"]["gain_loss_percent"]=to_float(pct)
|
||
return r
|
||
|
||
class SGSParser:
|
||
lab="SGS"
|
||
def parse(self,text):
|
||
r=empty_weight_report("SGS")
|
||
r["report"]["reference"]=extract("LANDING REPORT No",text)
|
||
r["report"]["file_no"]=extract("FILE NO.",text)
|
||
r["report"]["date"]=extract("DATE",text)
|
||
|
||
r["contract"]["contract_no"]=extract("CONTRACT NO.",text)
|
||
r["contract"]["invoice_no"]=extract("INVOICE NO.",text)
|
||
r["contract"]["origin"]=extract("ORIGIN",text)
|
||
r["contract"]["commodity"]=extract("PRODUCT",text)
|
||
|
||
r["parties"]["seller"]=extract("Seller",text)
|
||
r["parties"]["buyer"]=extract("Buyer",text)
|
||
r["parties"]["carrier"]=extract("Carrier",text)
|
||
|
||
r["shipment"]["bl_no"]=extract("B/L no.",text)
|
||
r["shipment"]["port_loading"]=extract("Port of loading",text)
|
||
r["shipment"]["port_destination"]=extract("Port of destination",text)
|
||
r["shipment"]["arrival_date"]=extract("Vessel arrival date",text)
|
||
r["shipment"]["weighing_place"]=extract("Place of weighing",text)
|
||
r["shipment"]["weighing_method"]=extract("Weighing mode",text)
|
||
r["shipment"]["bales"]=to_float(extract("Quantity arrived",text))
|
||
|
||
r["weights"]["gross_landed_kg"]=to_float(extract("Gross landed",text))
|
||
r["weights"]["tare_kg"]=to_float(extract("Tare",text))
|
||
r["weights"]["net_landed_kg"]=to_float(extract("Net landed",text))
|
||
r["weights"]["invoice_net_kg"]=to_float(extract("Net invoiced",text))
|
||
r["weights"]["gain_loss_kg"]=to_float(safe_search(r"Gain.*?([0-9.,]+)\s*kgs",text))
|
||
r["weights"]["gain_loss_percent"]=to_float(safe_search(r"Gain\s*\+?\s*([0-9.,]+)\s*%",text))
|
||
return r
|
||
|
||
class PICLParser:
|
||
lab="PICL"
|
||
def parse(self,text):
|
||
r=empty_weight_report("PICL")
|
||
|
||
r["report"]["reference"]=safe_search(r"No[:\s]+([A-Z0-9\-]+)",text)
|
||
r["report"]["date"]=safe_search(r"(Monday|Tuesday|Wednesday|Thursday|Friday|Saturday|Sunday),?\s*([A-Za-z]+\s+[0-9]{1,2},\s*[0-9]{4})",text,group_index=2)
|
||
|
||
r["contract"]["contract_no"]=extract("Contract/Pl No & Date",text)
|
||
r["contract"]["invoice_no"]=extract("Invoice ilo & Date",text)
|
||
r["contract"]["lc_no"]=extract("L/C No & Date",text)
|
||
r["contract"]["origin"]=extract("Country of Origin",text)
|
||
r["contract"]["commodity"]=extract("Commodity",text)
|
||
|
||
r["parties"]["seller"]=extract("FAIRCOT SA",text)
|
||
r["parties"]["buyer"]=extract("M/S.",text)
|
||
r["parties"]["carrier"]=extract("Shipping Agent",text)
|
||
|
||
r["shipment"]["vessel"]=extract("Shipped Per Vessel",text)
|
||
r["shipment"]["bl_no"]=extract("B/L No & Date",text)
|
||
r["shipment"]["port_loading"]=extract("Port of Loading",text)
|
||
r["shipment"]["port_destination"]=extract("Port of Discharge",text)
|
||
r["shipment"]["arrival_date"]=extract("Date of Anival & LDL",text)
|
||
r["shipment"]["weighing_place"]=extract("Place & Date of Weighment",text)
|
||
r["shipment"]["weighing_method"]=extract("Method of Weighment",text)
|
||
r["shipment"]["bales"]=to_float(extract("Grand Total",text))
|
||
|
||
r["weights"]["gross_landed_kg"]=to_float(extract("Total;",text))
|
||
r["weights"]["tare_kg"]=to_float(extract("Tare Weight",text))
|
||
r["weights"]["net_landed_kg"]=to_float(extract("Grand Total",text))
|
||
r["weights"]["invoice_net_kg"]=to_float(extract("Invoice weight",text))
|
||
r["weights"]["gain_loss_kg"]=to_float(safe_search(r"(-[0-9.,]+)\s*KGS",text))
|
||
r["weights"]["gain_loss_percent"]=to_float(safe_search(r"\(\s*([0-9.,]+)\s*o/o\s*\)",text))
|
||
return r
|
||
|
||
# Configure root logger explicitly
|
||
root = logging.getLogger()
|
||
root.setLevel(logging.INFO)
|
||
root.addHandler(file_handler)
|
||
root.addHandler(logging.StreamHandler())
|
||
|
||
# Use root logger for your app
|
||
logger = logging.getLogger(__name__)
|
||
|
||
app = FastAPI()
|
||
logger.info("Loading models...")
|
||
|
||
nlp = spacy.load("en_core_web_sm")
|
||
predictor = ocr_predictor(pretrained=True)
|
||
|
||
logger.info("Models loaded successfully.")
|
||
|
||
import io
|
||
import re
|
||
from datetime import datetime
|
||
from typing import Dict, Any
|
||
import pytesseract
|
||
from pdf2image import convert_from_bytes
|
||
from PIL import Image
|
||
from PyPDF2 import PdfReader
|
||
import json
|
||
|
||
def parse_cotton_report(ocr_text: str) -> Dict[str, Any]:
|
||
"""
|
||
Parse structured data from cotton landing report OCR text
|
||
"""
|
||
result = {
|
||
"lab": "ALFRED H KNIGHT",
|
||
"report": {"reference": None, "file_no": None, "date": None},
|
||
"contract": {"contract_no": None, "invoice_no": None, "lc_no": None,
|
||
"origin": None, "commodity": None},
|
||
"parties": {"seller": None, "buyer": None, "carrier": None},
|
||
"shipment": {
|
||
"vessel": None, "bl_no": None, "port_loading": None,
|
||
"port_destination": None, "arrival_date": None,
|
||
"weighing_place": None, "weighing_method": None,
|
||
"bales": None
|
||
},
|
||
"weights": {
|
||
"gross_landed_kg": None, "tare_kg": None,
|
||
"net_landed_kg": None, "invoice_net_kg": None,
|
||
"gain_loss_kg": None, "gain_loss_percent": None
|
||
}
|
||
}
|
||
|
||
# Clean the text
|
||
lines = ocr_text.split('\n')
|
||
clean_lines = [line.strip() for line in lines if line.strip()]
|
||
|
||
# Extract using patterns
|
||
text = ocr_text.lower()
|
||
|
||
# 1. Extract report reference and file number
|
||
ref_match = re.search(r'client reference:\s*([^\n]+)', ocr_text, re.IGNORECASE)
|
||
if ref_match:
|
||
result["report"]["reference"] = ref_match.group(1).strip()
|
||
|
||
# Try to get file number from AHK reference
|
||
ahk_match = re.search(r'ahk s/(\d+)/', ocr_text, re.IGNORECASE)
|
||
if ahk_match:
|
||
result["report"]["file_no"] = ahk_match.group(1)
|
||
|
||
# 2. Extract dates
|
||
date_match = re.search(r'printed date:\s*(\d{1,2}-[a-z]+-\d{4})', text, re.IGNORECASE)
|
||
if date_match:
|
||
result["report"]["date"] = date_match.group(1).title()
|
||
|
||
# 3. Extract contract information
|
||
# Origin/Growth
|
||
growth_match = re.search(r'growth\s*:?\s*([^\n:]+(?:raw cotton)?)', ocr_text, re.IGNORECASE)
|
||
if growth_match:
|
||
origin = growth_match.group(1).strip()
|
||
result["contract"]["origin"] = origin
|
||
result["contract"]["commodity"] = "COTTON"
|
||
|
||
# Invoice number from reference
|
||
if result["report"]["reference"]:
|
||
inv_match = re.search(r'inv\s*(\d+)', result["report"]["reference"], re.IGNORECASE)
|
||
if inv_match:
|
||
result["contract"]["invoice_no"] = inv_match.group(1)
|
||
|
||
# 4. Extract parties
|
||
# Seller
|
||
seller_match = re.search(r'client\s*:?\s*([^\n]+)', ocr_text, re.IGNORECASE)
|
||
if seller_match:
|
||
# Skip the "Client" label if present
|
||
seller_text = seller_match.group(1).strip()
|
||
if not seller_text.lower().startswith('client'):
|
||
result["parties"]["seller"] = seller_text
|
||
|
||
# Buyer
|
||
buyer_match = re.search(r'buyer\s*:?\s*([^\n]+)', ocr_text, re.IGNORECASE)
|
||
if buyer_match:
|
||
buyer_text = buyer_match.group(1).strip()
|
||
if not buyer_text.lower().startswith('buyer'):
|
||
result["parties"]["buyer"] = buyer_text
|
||
|
||
# 5. Extract shipment details
|
||
# Vessel
|
||
vessel_match = re.search(r'vessel\s*:?\s*([^\n]+)', ocr_text, re.IGNORECASE)
|
||
if vessel_match:
|
||
vessel_text = vessel_match.group(1).strip()
|
||
if not vessel_text.lower().startswith('vessel'):
|
||
result["shipment"]["vessel"] = vessel_text
|
||
|
||
# B/L Number
|
||
bl_match = re.search(r'b/l no\.?\s*:?\s*([^\n]+)', ocr_text, re.IGNORECASE)
|
||
if bl_match:
|
||
bl_text = bl_match.group(1).strip()
|
||
result["shipment"]["bl_no"] = bl_text
|
||
|
||
# Destination
|
||
dest_match = re.search(r'destination\s*:?\s*([^\n]+)', ocr_text, re.IGNORECASE)
|
||
if dest_match:
|
||
dest_text = dest_match.group(1).strip()
|
||
if not dest_text.lower().startswith('destination'):
|
||
result["shipment"]["port_destination"] = dest_text
|
||
|
||
# Arrival Date
|
||
arrival_match = re.search(r'arrival date\s*:?\s*(\d{1,2}-[a-z]+-\d{4})', text, re.IGNORECASE)
|
||
if arrival_match:
|
||
result["shipment"]["arrival_date"] = arrival_match.group(1).title()
|
||
|
||
# Weighing method
|
||
weigh_match = re.search(r'weighing method\s*:?\s*([^\n]+)', ocr_text, re.IGNORECASE)
|
||
if weigh_match:
|
||
method_text = weigh_match.group(1).strip()
|
||
if not method_text.lower().startswith('weighing'):
|
||
result["shipment"]["weighing_method"] = method_text
|
||
|
||
# Bales count
|
||
bales_match = re.search(r'total bales\s*:?\s*(\d+)', ocr_text, re.IGNORECASE)
|
||
if bales_match:
|
||
result["shipment"]["bales"] = int(bales_match.group(1))
|
||
|
||
# 6. Extract weights (critical section)
|
||
# Gross Landed Weight
|
||
gross_match = re.search(r'gross\s*:?\s*([\d,\.]+)\s*kg', ocr_text)
|
||
if gross_match:
|
||
# We need the second occurrence (landed weight)
|
||
all_gross = re.findall(r'gross\s*:?\s*([\d,\.]+)\s*kg', ocr_text)
|
||
if len(all_gross) >= 2:
|
||
result["weights"]["gross_landed_kg"] = float(all_gross[1].replace(',', ''))
|
||
|
||
# Tare weight (should be same in both)
|
||
tare_match = re.search(r'tare\s*:?\s*([\d,\.]+)\s*kg', ocr_text)
|
||
if tare_match:
|
||
result["weights"]["tare_kg"] = float(tare_match.group(1).replace(',', ''))
|
||
|
||
# Net weights
|
||
net_matches = re.findall(r'net\s*:?\s*([\d,\.]+)\s*kg', ocr_text)
|
||
if len(net_matches) >= 2:
|
||
result["weights"]["invoice_net_kg"] = float(net_matches[0].replace(',', ''))
|
||
result["weights"]["net_landed_kg"] = float(net_matches[1].replace(',', ''))
|
||
|
||
# Loss/Gain
|
||
loss_match = re.search(r'loss\s*:?\s*[-–]?\s*([\d,\.]+)\s*kg', ocr_text, re.IGNORECASE)
|
||
if loss_match:
|
||
loss_value = float(loss_match.group(1).replace(',', ''))
|
||
# Make it negative if not already indicated
|
||
if '-' not in loss_match.group(0) and '–' not in loss_match.group(0):
|
||
loss_value = -loss_value
|
||
result["weights"]["gain_loss_kg"] = loss_value
|
||
|
||
# Percentage
|
||
percent_match = re.search(r'percentage\s*:?\s*[-–]?\s*([\d,\.]+)%', ocr_text, re.IGNORECASE)
|
||
if percent_match:
|
||
percent_value = float(percent_match.group(1).replace(',', ''))
|
||
if '-' not in percent_match.group(0) and '–' not in percent_match.group(0):
|
||
percent_value = -percent_value
|
||
result["weights"]["gain_loss_percent"] = percent_value
|
||
|
||
return result
|
||
|
||
@app.post("/ocr")
|
||
async def ocr(file: UploadFile):
|
||
"""
|
||
Enhanced OCR endpoint that returns structured data
|
||
"""
|
||
logger.info(f"Received structured OCR request: {file.filename}")
|
||
|
||
try:
|
||
file_data = await file.read()
|
||
ext = file.filename.lower()
|
||
|
||
ocr_text = ""
|
||
|
||
# Process PDF
|
||
if ext.endswith(".pdf"):
|
||
# Try native text extraction first
|
||
reader = PdfReader(io.BytesIO(file_data))
|
||
direct_text = "".join(page.extract_text() or "" for page in reader.pages)
|
||
|
||
if direct_text.strip():
|
||
logger.info("Using native PDF text")
|
||
ocr_text = direct_text
|
||
else:
|
||
# Fallback to OCR
|
||
logger.info("Using OCR for scanned PDF")
|
||
images = convert_from_bytes(file_data)
|
||
for i, img in enumerate(images):
|
||
logger.info(f"OCR page {i+1}/{len(images)}")
|
||
ocr_text += pytesseract.image_to_string(img) + "\n"
|
||
else:
|
||
# Process image
|
||
img = Image.open(io.BytesIO(file_data))
|
||
ocr_text = pytesseract.image_to_string(img)
|
||
|
||
# Parse structured data
|
||
structured_data = parse_cotton_report(ocr_text)
|
||
|
||
return {
|
||
"success": True,
|
||
"ocr_text": ocr_text[:1000] + "..." if len(ocr_text) > 1000 else ocr_text,
|
||
"structured_data": structured_data,
|
||
"json": json.dumps(structured_data, indent=2, ensure_ascii=False)
|
||
}
|
||
|
||
except Exception as e:
|
||
logger.error(f"Structured OCR failed: {e}", exc_info=True)
|
||
return {
|
||
"success": False,
|
||
"error": str(e),
|
||
"raw_text": "",
|
||
"structured_data": {}
|
||
}
|
||
|
||
# =============================
|
||
# 🧠 Smart OCR
|
||
# =============================
|
||
# @app.post("/ocr")
|
||
# async def ocr(file: UploadFile):
|
||
# logger.info(f"Received OCR request: {file.filename}")
|
||
# try:
|
||
# file_data = await file.read()
|
||
# ext = file.filename.lower()
|
||
|
||
# # --------- PDF with native text ---------
|
||
# if ext.endswith(".pdf"):
|
||
# logger.info("PDF detected → Extracting native text first")
|
||
# reader = PdfReader(io.BytesIO(file_data))
|
||
# direct_text = "".join(
|
||
# page.extract_text() or "" for page in reader.pages
|
||
# )
|
||
|
||
# if direct_text.strip():
|
||
# logger.info("Native PDF text found → No OCR needed")
|
||
# return {"ocr_text": direct_text}
|
||
|
||
# # -------- Fallback: scanned PDF OCR --------
|
||
# logger.info("No native text → PDF treated as scanned → OCR")
|
||
# from pdf2image import convert_from_bytes
|
||
# images = convert_from_bytes(file_data)
|
||
# text = ""
|
||
# for i, img in enumerate(images):
|
||
# logger.info(f"OCR page {i+1}/{len(images)}")
|
||
# text += pytesseract.image_to_string(img) + "\n"
|
||
|
||
# return {"ocr_text": text}
|
||
|
||
# # --------- Image file OCR ---------
|
||
# logger.info("Image detected → Running OCR")
|
||
# img = Image.open(io.BytesIO(file_data))
|
||
# text = pytesseract.image_to_string(img)
|
||
# return {"ocr_text": text}
|
||
|
||
# except Exception as e:
|
||
# logger.error(f"OCR failed: {e}", exc_info=True)
|
||
# raise HTTPException(status_code=500, detail=str(e))
|
||
|
||
# =============================
|
||
# 🧱 Structure / Layout
|
||
# =============================
|
||
@app.post("/structure")
|
||
async def structure(file: UploadFile):
|
||
logger.info(f"Received structure request: {file.filename}")
|
||
try:
|
||
file_data = await file.read()
|
||
ext = file.filename.lower()
|
||
|
||
if ext.endswith(".pdf"):
|
||
doc = DocumentFile.from_pdf(file_data)
|
||
logger.info(f"Structure prediction on PDF ({len(doc)} pages)")
|
||
else:
|
||
img = Image.open(io.BytesIO(file_data)).convert("RGB")
|
||
doc = DocumentFile.from_images([img])
|
||
logger.info("Structure prediction on image")
|
||
|
||
res = predictor(doc)
|
||
return {"structure": str(res)}
|
||
|
||
except Exception as e:
|
||
logger.error(f"Structure extraction failed: {e}", exc_info=True)
|
||
raise HTTPException(status_code=500, detail=str(e))
|
||
|
||
# =============================
|
||
# 📊 Tables extraction (PDF only)
|
||
# =============================
|
||
@app.post("/tables")
|
||
async def tables(file: UploadFile):
|
||
logger.info(f"Received table extraction request: {file.filename}")
|
||
try:
|
||
file_data = await file.read()
|
||
buffer = io.BytesIO(file_data)
|
||
|
||
tables = camelot.read_pdf(buffer)
|
||
logger.info(f"Found {len(tables)} tables")
|
||
return {"tables": [t.df.to_dict() for t in tables]}
|
||
|
||
except Exception as e:
|
||
logger.error(f"Table extraction failed: {e}", exc_info=True)
|
||
raise HTTPException(status_code=500, detail=str(e))
|
||
|
||
def safe_search(pattern, text, default=None, group_index=1, context=""):
|
||
"""Recherche sécurisée avec logging en cas d'absence de correspondance."""
|
||
m = re.search(pattern, text, re.I | re.S)
|
||
if not m:
|
||
logger.warning("Pattern not found for %s: %s", context, pattern)
|
||
return default
|
||
try:
|
||
return m.group(group_index).strip()
|
||
except IndexError:
|
||
logger.warning("Group index %d not found for %s: %s", group_index, context, pattern)
|
||
return default
|
||
|
||
def to_float(s):
|
||
if not s:
|
||
return None
|
||
s = s.replace(",", "").replace("Kgs", "").replace("kg", "").replace("%", "")
|
||
s = s.replace("lbs", "").replace("LBS", "")
|
||
s = s.strip()
|
||
try:
|
||
return float(s)
|
||
except:
|
||
return None
|
||
|
||
def section(text, start, end=None):
|
||
"""Extract a block of text between two headings, safely."""
|
||
pattern_start = re.escape(start)
|
||
if end:
|
||
pattern_end = re.escape(end)
|
||
reg = re.compile(pattern_start + r"(.*?)" + pattern_end, re.S | re.I)
|
||
else:
|
||
reg = re.compile(pattern_start + r"(.*)", re.S | re.I)
|
||
m = reg.search(text)
|
||
if not m:
|
||
logger.warning("Section not found: start='%s', end='%s'", start, end)
|
||
return ""
|
||
return m.group(1).strip()
|
||
|
||
def extract_field(text, label, default=None):
|
||
"""Extract a line of the form 'Label: value', safely."""
|
||
pattern = rf"{re.escape(label)}\s*:?[\s]+([^\n]+)"
|
||
return safe_search(pattern, text, default=default, context=f"field '{label}'")
|
||
|
||
def extract(label, text, default=None):
|
||
"""
|
||
Robust extraction for OCR/PDF text.
|
||
Works with:
|
||
Label: Value
|
||
Label Value
|
||
Label .... Value
|
||
"""
|
||
if not text:
|
||
return default
|
||
|
||
patterns = [
|
||
rf"{re.escape(label)}\s*[:\-]?\s*([^\n\r]+)",
|
||
rf"{re.escape(label)}\s+([^\n\r]+)"
|
||
]
|
||
|
||
for p in patterns:
|
||
m = re.search(p, text, re.I)
|
||
if m:
|
||
return m.group(1).strip()
|
||
|
||
return default
|
||
|
||
def extract_report_metadata(text):
|
||
logger.info("Starting metadata extraction, text length=%d", len(text))
|
||
|
||
try:
|
||
# ----------- SECTIONS -----------
|
||
order_details = section(text, "Order details", "Weights")
|
||
invoice_section = section(text, "INVOICE WEIGHTS", "Bales Weighed")
|
||
landed_section = section(text, "Bales Weighed", "Outturn")
|
||
loss_section = section(text, "LOSS", "Invoice average")
|
||
avg_section = section(text, "Invoice average", "Comments")
|
||
signature_block = section(text, "Signed on")
|
||
|
||
# ----------- TOP INFO -----------
|
||
top_info = {
|
||
"produced_on": extract_field(text, "Produced On"),
|
||
"printed_date": extract_field(text, "Printed Date"),
|
||
"client_reference": extract_field(text, "Client Reference"),
|
||
"report_number": safe_search(r"(AHK\S+)", text, default="", context="report_number", group_index=1),
|
||
}
|
||
|
||
# ----------- ORDER DETAILS -----------
|
||
parties = {
|
||
"client": extract_field(order_details, "Client"),
|
||
"client_ref_no": extract_field(order_details, "Client Ref No"),
|
||
"buyer": extract_field(order_details, "Buyer"),
|
||
"destination": extract_field(order_details, "Destination"),
|
||
}
|
||
|
||
shipment = {
|
||
"total_bales": extract_field(order_details, "Total Bales"),
|
||
"vessel": extract_field(order_details, "Vessel"),
|
||
"voyage_no": extract_field(order_details, "Voy. No"),
|
||
"bl_no": extract_field(order_details, "B/L No"),
|
||
"bl_date": extract_field(order_details, "B/L Date"),
|
||
"growth": extract_field(order_details, "Growth"),
|
||
"arrival_date": extract_field(order_details, "Arrival Date"),
|
||
"first_weighing_date": extract_field(order_details, "First date of weighing"),
|
||
"last_weighing_date": extract_field(order_details, "Last Date of Weighing"),
|
||
"weighing_method": extract_field(order_details, "Weighing method"),
|
||
"tare_basis": extract_field(order_details, "Tare"),
|
||
}
|
||
|
||
# ----------- INVOICE SECTION -----------
|
||
invoice = {
|
||
"bales": extract_field(invoice_section, "Bales"),
|
||
"gross": extract_field(invoice_section, "Gross"),
|
||
"tare": extract_field(invoice_section, "Tare"),
|
||
"net": extract_field(invoice_section, "Net"),
|
||
}
|
||
|
||
# ----------- LANDED SECTION -----------
|
||
landed = {
|
||
"bales": extract_field(landed_section, "Bales"),
|
||
"gross": extract_field(landed_section, "Gross"),
|
||
"tare": extract_field(landed_section, "Tare"),
|
||
"net": extract_field(landed_section, "Net"),
|
||
}
|
||
|
||
# ----------- LOSS SECTION -----------
|
||
loss = {
|
||
"kg": extract_field(loss_section, "kg"),
|
||
"lb": extract_field(loss_section, "lb"),
|
||
"percent": extract_field(loss_section, "Percentage"),
|
||
}
|
||
|
||
# ----------- AVERAGES SECTION -----------
|
||
averages = {
|
||
"invoice_gross_per_bale": extract_field(avg_section, "Invoice average"),
|
||
"landed_gross_per_bale": extract_field(avg_section, "Landed average"),
|
||
}
|
||
|
||
# ----------- SIGNATURE -----------
|
||
signature = {
|
||
"signed_on": extract_field(signature_block, "Signed on"),
|
||
"signed_by": safe_search(r"\n([A-Za-z ]+)\nClient Services", signature_block, default="", context="signed_by"),
|
||
"role": "Client Services Coordinator",
|
||
"company": "Alfred H. Knight International Limited"
|
||
}
|
||
|
||
logger.info("Metadata extraction completed successfully")
|
||
return {
|
||
"report": top_info,
|
||
"parties": parties,
|
||
"shipment": shipment,
|
||
"weights": {
|
||
"invoice": invoice,
|
||
"landed": landed,
|
||
"loss": loss,
|
||
"averages": averages
|
||
},
|
||
"signature": signature
|
||
}
|
||
|
||
except Exception as e:
|
||
logger.exception("Unexpected error during metadata extraction")
|
||
raise HTTPException(status_code=500, detail=f"Metadata extraction failed: {e}")
|
||
|
||
def detect_template(text):
|
||
t = text.lower()
|
||
|
||
if "alfred h. knight" in t and "cotton landing report" in t:
|
||
return "AHK"
|
||
|
||
if "intertek" in t and "landing report" in t:
|
||
return "INTERTEK"
|
||
|
||
if "robertson international" in t or "ri ref no" in t:
|
||
return "ROBERTSON"
|
||
|
||
if "landing report" in t and "carcon cargo" in t:
|
||
return "SGS"
|
||
|
||
if "pacific inspection company" in t or "picl-bd.com" in t:
|
||
return "PICL"
|
||
|
||
return "UNKNOWN"
|
||
|
||
@app.post("/metadata")
|
||
async def metadata(text: str = Body(..., embed=True)):
|
||
return extract_report_metadata(text)
|
||
|
||
@app.post("/parse")
|
||
async def parse_endpoint(text: str = Body(..., embed=True)):
|
||
return parse_report(text)
|
||
|
||
PARSERS = {
|
||
"AHK": AHKParser(),
|
||
"INTERTEK": IntertekParser(),
|
||
"ROBERTSON": RobertsonParser(),
|
||
"SGS": SGSParser(),
|
||
"PICL": PICLParser()
|
||
}
|
||
|
||
def empty_weight_report(lab):
|
||
return {
|
||
"lab": lab,
|
||
"report": {"reference": None, "file_no": None, "date": None},
|
||
"contract": {"contract_no": None, "invoice_no": None, "lc_no": None, "origin": None, "commodity": None},
|
||
"parties": {"seller": None, "buyer": None, "carrier": None},
|
||
"shipment": {
|
||
"vessel": None, "bl_no": None, "port_loading": None,
|
||
"port_destination": None, "arrival_date": None,
|
||
"weighing_place": None, "weighing_method": None,
|
||
"bales": None
|
||
},
|
||
"weights": {
|
||
"gross_landed_kg": None, "tare_kg": None,
|
||
"net_landed_kg": None, "invoice_net_kg": None,
|
||
"gain_loss_kg": None, "gain_loss_percent": None
|
||
}
|
||
}
|
||
|
||
def parse_report(text):
|
||
template=detect_template(text)
|
||
if template not in PARSERS:
|
||
return {"template":"UNKNOWN"}
|
||
return PARSERS[template].parse(text) |