Expense Tracker MCP Server

"""PDF parsing and receipt extraction logic.""" import re from datetime import datetime from pathlib import Path from typing import Optional import pdfplumber from .models import Receipt, LineItem # Known store name patterns STORE_PATTERNS = [ r"\bwalmart\b", r"\bcostco\s*wholesale\b", r"\bcostco\b", r"\btarget\b", r"\bwhole\s*foods\b", r"\btrader\s*joe'?s\b", r"\bsafeway\b", r"\bkroger\b", r"\bpublix\b", ] # Date patterns (multiple formats) DATE_PATTERNS = [ r"(\d{1,2})[/-](\d{1,2})[/-](\d{2,4})", # MM/DD/YYYY or MM-DD-YYYY r"(\d{4})[/-](\d{1,2})[/-](\d{1,2})", # YYYY-MM-DD or YYYY/MM/DD ] # Price pattern PRICE_PATTERN = r"\$?\s*(\d+\.\d{2})" def extract_text_from_pdf(pdf_path: Path) -> str: """Extract text from PDF file. Args: pdf_path: Path to PDF file Returns: Extracted text as string Raises: FileNotFoundError: If PDF doesn't exist Exception: If PDF parsing fails """ if not pdf_path.exists(): raise FileNotFoundError(f"PDF file not found: {pdf_path}") text_content = [] with pdfplumber.open(pdf_path) as pdf: for page in pdf.pages: text = page.extract_text() if text: text_content.append(text) if not text_content: raise ValueError("No text could be extracted from PDF") return "\n".join(text_content) def identify_store(text: str) -> str: """Identify store name from receipt text. Args: text: Receipt text Returns: Store name (capitalized) or "Unknown Store" """ # Check first 10 lines for store name lines = text.split("\n")[:10] text_to_search = "\n".join(lines).lower() for pattern in STORE_PATTERNS: match = re.search(pattern, text_to_search, re.IGNORECASE) if match: store_name = match.group(0).strip() # Capitalize properly return store_name.title() return "Unknown Store" def extract_date(text: str) -> str: """Extract purchase date from receipt. Args: text: Receipt text Returns: Date in ISO format (YYYY-MM-DD) or current date if not found """ # Look for date in first 20 lines lines = text.split("\n")[:20] text_to_search = "\n".join(lines) for pattern in DATE_PATTERNS: match = re.search(pattern, text_to_search) if match: groups = match.groups() # Try to parse based on pattern type try: if len(groups[0]) == 4: # YYYY-MM-DD format year, month, day = groups else: # MM-DD-YYYY format month, day, year = groups # Handle 2-digit years year = int(year) if year < 100: year += 2000 if year < 50 else 1900 # Create date object to validate date_obj = datetime(year, int(month), int(day)) return date_obj.strftime("%Y-%m-%d") except (ValueError, IndexError): continue # Default to current date if not found return datetime.now().strftime("%Y-%m-%d") def extract_totals(text: str) -> dict: """Extract subtotal, tax, and total from receipt. Args: text: Receipt text Returns: Dictionary with subtotal, tax, total (may be None) """ result = {"subtotal": None, "tax": None, "total": None} # Look for these in the last 30 lines (where totals usually are) lines = text.split("\n")[-30:] text_to_search = "\n".join(lines).lower() # Extract total (required) total_match = re.search( rf"(?:^|\s)total\s*:?\s*{PRICE_PATTERN}", text_to_search, re.MULTILINE ) if total_match: result["total"] = float(total_match.group(1)) # Extract subtotal subtotal_match = re.search( rf"(?:^|\s)sub\s*total\s*:?\s*{PRICE_PATTERN}", text_to_search, re.MULTILINE ) if subtotal_match: result["subtotal"] = float(subtotal_match.group(1)) # Extract tax tax_match = re.search( rf"(?:^|\s)tax\s*:?\s*{PRICE_PATTERN}", text_to_search, re.MULTILINE ) if tax_match: result["tax"] = float(tax_match.group(1)) return result def extract_line_items(text: str) -> list[dict]: """Extract line items from receipt. Args: text: Receipt text Returns: List of dictionaries with item_name, quantity, price """ items = [] lines = text.split("\n") # Item line patterns (trying multiple formats) patterns = [ # Pattern 1: [qty] ITEM_NAME $PRICE r"^(\d+)\s+(.+?)\s+\$?\s*(\d+\.\d{2})\s*$", # Pattern 2: ITEM_NAME $PRICE r"^(.+?)\s+\$\s*(\d+\.\d{2})\s*$", # Pattern 3: ITEM_NAME PRICE (multiple spaces) r"^(.+?)\s{2,}\$?\s*(\d+\.\d{2})\s*$", ] for line in lines: line = line.strip() # Skip empty lines, header lines, and total lines if not line: continue if len(line) < 5: continue if any( keyword in line.lower() for keyword in ["total", "subtotal", "tax", "payment", "change", "card"] ): continue # Try each pattern for pattern in patterns: match = re.match(pattern, line, re.IGNORECASE) if match: groups = match.groups() # Parse based on number of groups if len(groups) == 3: # qty, name, price qty, name, price = groups items.append( { "item_name": name.strip(), "quantity": float(qty), "price": float(price), } ) elif len(groups) == 2: # name, price (no quantity) name, price = groups items.append( {"item_name": name.strip(), "quantity": 1.0, "price": float(price)} ) break # Found a match, move to next line return items def parse_receipt(text: str) -> tuple[Receipt, list[dict]]: """Parse receipt text into structured data. Args: text: Raw receipt text Returns: Tuple of (Receipt object, list of item dicts) Raises: ValueError: If total is not found or is invalid """ # Extract metadata store_name = identify_store(text) purchase_date = extract_date(text) totals = extract_totals(text) if totals["total"] is None or totals["total"] <= 0: raise ValueError("Could not extract valid total from receipt") # Create receipt object receipt = Receipt( store_name=store_name, purchase_date=purchase_date, subtotal=totals["subtotal"], tax=totals["tax"], total=totals["total"], ) # Extract line items items = extract_line_items(text) return receipt, items def parse_pdf_receipt(pdf_path: Path) -> tuple[Receipt, list[dict]]: """Parse a PDF receipt file. Args: pdf_path: Path to PDF file Returns: Tuple of (Receipt object, list of item dicts) Raises: FileNotFoundError: If PDF doesn't exist ValueError: If parsing fails """ # Convert string to Path if needed if isinstance(pdf_path, str): pdf_path = Path(pdf_path) # Extract text text = extract_text_from_pdf(pdf_path) # Parse receipt return parse_receipt(text)