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 pdfplumber 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" )) import re from datetime import datetime class AHKParser: def __init__(self, text_content): self.text = text_content self.data = None def parse(self, lab="AHK"): """Parse le texte et retourne un dictionnaire structuré""" result = { "lab": lab, "report": self._extract_report_info(), "contract": self._extract_contract_info(), "parties": self._extract_parties_info(), "shipment": self._extract_shipment_info(), "weights": self._extract_weights_info() } self.data = result return result def _extract_report_info(self): """Extrait les informations du rapport""" report_info = { "reference": None, "file_no": None, "date": None } # Recherche de la référence client ref_match = re.search(r'Client Reference:\s*(S-\d+/\s*INV\s*\d+)', self.text) if ref_match: report_info["reference"] = ref_match.group(1).strip() # Recherche du numéro de fichier AHK file_no_match = re.search(r'AHK\s*S/([\w/]+)', self.text) if file_no_match: report_info["file_no"] = file_no_match.group(1).strip() # Recherche de la date du rapport date_match = re.search(r'Signed on\s*(\d{1,2}-[A-Za-z]{3}-\d{4})', self.text) if date_match: report_info["date"] = date_match.group(1).strip() return report_info def _extract_contract_info(self): """Extrait les informations du contrat""" contract_info = { "contract_no": None, "invoice_no": None, "lc_no": None, "origin": None, "commodity": None } # Extraction de la référence client (peut servir comme numéro de contrat) ref_match = re.search(r'Client Reference:\s*(S-\d+/\s*INV\s*\d+)', self.text) if ref_match: ref_parts = ref_match.group(1).split('/') if len(ref_parts) >= 2: contract_info["contract_no"] = ref_parts[0].strip() contract_info["invoice_no"] = ref_parts[1].strip() # Extraction de l'origine et de la marchandise origin_match = re.search(r'Growth\s*:\s*([\w\s]+)', self.text) if origin_match: origin_text = origin_match.group(1).strip() if "AUSTRALIAN" in origin_text.upper(): contract_info["origin"] = "AUSTRALIA" # La marchandise est généralement "RAW COTTON" contract_info["commodity"] = "RAW COTTON" return contract_info def _extract_parties_info(self): """Extrait les informations sur les parties""" parties_info = { "seller": None, "buyer": None, "carrier": None } # Extraction du vendeur (Client) seller_match = re.search(r'Client\s*:\s*([^\n]+)', self.text) if seller_match: parties_info["seller"] = seller_match.group(1).strip() # Extraction de l'acheteur (Buyer) buyer_match = re.search(r'Buyer\s*:\s*([^\n]+)', self.text) if buyer_match: parties_info["buyer"] = buyer_match.group(1).strip() # Extraction du transporteur (Vessel) vessel_match = re.search(r'Vessel\s*:\s*([^\n]+)', self.text) if vessel_match: # On considère le nom du navire comme transporteur parties_info["carrier"] = vessel_match.group(1).strip() return parties_info def _extract_shipment_info(self): """Extrait les informations d'expédition""" shipment_info = { "vessel": None, "bl_no": None, "bl_date": None, "port_loading": None, # Non spécifié dans le texte "port_destination": None, "arrival_date": None, "weighing_place": None, # Non spécifié dans le texte "weighing_method": None, "bales": None } # Extraction du navire vessel_match = re.search(r'Vessel\s*:\s*([^\n]+)', self.text) if vessel_match: shipment_info["vessel"] = vessel_match.group(1).strip() # Extraction du numéro de connaissement bl_no_match = re.search(r'B/L No\.\s*:\s*([^\n]+)', self.text) if bl_no_match: shipment_info["bl_no"] = bl_no_match.group(1).strip() # Extraction de la date du connaissement bl_date_match = re.search(r'B/L Date\s*:\s*(\d{1,2}-[A-Za-z]{3}-\d{4})', self.text) if bl_date_match: shipment_info["bl_date"] = bl_date_match.group(1).strip() # Extraction du port de destination dest_match = re.search(r'Destination\s*:\s*([^\n]+)', self.text) if dest_match: shipment_info["port_destination"] = dest_match.group(1).strip() # Extraction de la date d'arrivée arrival_match = re.search(r'Arrival Date\s*:\s*(\d{1,2}-[A-Za-z]{3}-\d{4})', self.text) if arrival_match: shipment_info["arrival_date"] = arrival_match.group(1).strip() # Extraction de la méthode de pesée weighing_method_match = re.search(r'Weighing method\s*:\s*([^\n]+)', self.text) if weighing_method_match: shipment_info["weighing_method"] = weighing_method_match.group(1).strip() # Extraction du nombre de balles bales_match = re.search(r'Total Bales\s*:\s*(\d+)', self.text) if bales_match: shipment_info["bales"] = int(bales_match.group(1).strip()) return shipment_info def _extract_weights_info(self): """Extrait les informations de poids""" weights_info = { "gross_landed_kg": None, "tare_kg": None, "net_landed_kg": None, "invoice_net_kg": None, "gain_loss_kg": None, "gain_loss_percent": None } # Extraction du poids brut débarqué gross_landed_match = re.search(r'LANDED WEIGHTS[\s\S]*?Gross\s*:\s*([\d.]+)\s*kg', self.text) if gross_landed_match: weights_info["gross_landed_kg"] = float(gross_landed_match.group(1).strip()) # Extraction du poids de tare tare_match = re.search(r'Tare\s*:\s*([\d.]+)\s*kg', self.text) if tare_match: weights_info["tare_kg"] = float(tare_match.group(1).strip()) # Extraction du poids net débarqué net_landed_match = re.search(r'LANDED WEIGHTS[\s\S]*?Net\s*:\s*([\d.]+)\s*kg', self.text) if net_landed_match: weights_info["net_landed_kg"] = float(net_landed_match.group(1).strip()) # Extraction du poids net facturé invoice_net_match = re.search(r'INVOICE WEIGHTS[\s\S]*?Net\s*:\s*([\d.]+)\s*kg', self.text) if invoice_net_match: weights_info["invoice_net_kg"] = float(invoice_net_match.group(1).strip()) # Extraction de la perte en kg loss_match = re.search(r'LOSS\s*:\s*-\s*([\d.]+)\s*kg', self.text) if loss_match: weights_info["gain_loss_kg"] = -float(loss_match.group(1).strip()) # Extraction du pourcentage de perte percent_match = re.search(r'Percentage\s*:\s*-\s*([\d.]+)%', self.text) if percent_match: weights_info["gain_loss_percent"] = -float(percent_match.group(1).strip()) return weights_info # 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.") # ============================= # 🧠 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)) @app.post("/ocr") async def ocr(file: UploadFile): """ Smart PDF processing optimized for cotton landing reports """ logger.info(f"Smart OCR request: {file.filename}") try: file_data = await file.read() # Strategy 1: Try pdfplumber (best for digital PDFs) try: with pdfplumber.open(io.BytesIO(file_data)) as pdf: text_parts = [] tables_found = [] for page in pdf.pages: # Extract text page_text = page.extract_text(x_tolerance=2, y_tolerance=2) if page_text: text_parts.append(page_text) # Look for tables (common in landing reports) tables = page.extract_tables({ "vertical_strategy": "text", "horizontal_strategy": "text", "snap_tolerance": 5, }) for table in tables: if table and len(table) > 1: tables_found.append(table) combined_text = "\n".join(text_parts) return {"ocr_text": combined_text} # if combined_text.strip(): # logger.info(f"pdfplumber extracted {len(combined_text)} chars") # # Try parsing structured data # structured_data = parse_cotton_report(combined_text) # # Check if we got key fields # if (structured_data.get("shipment", {}).get("bales") and # structured_data.get("weights", {}).get("net_landed_kg")): # logger.info("Successfully parsed structured data from pdfplumber") # return { # "method": "pdfplumber", # "structured_data": structured_data, # "raw_text_sample": combined_text[:500] # } except Exception as e: logger.warning(f"pdfplumber attempt: {e}") # from pdf2image import convert_from_bytes # images = convert_from_bytes(file_data, dpi=200) # ocr_results = [] # for img in images: # text = pytesseract.image_to_string( # img, # config='--psm 6 -c preserve_interword_spaces=1' # ) # ocr_results.append(text) # ocr_text = "\n".join(ocr_results) # return { # "method": "tesseract_ocr", # "structured_data": ocr_text, # "raw_text_sample": ocr_text[:500] # } except Exception as e: logger.error(f"Smart OCR failed: {e}", exc_info=True) return { "error": str(e), "success": False } # ============================= # 🧱 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, "bl_date": 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)