import_receipt_from_pdf
Extract and parse receipt data from PDF files to automatically categorize expenses and store them in a database for tracking spending patterns.
Instructions
Import and parse a receipt from a PDF file.
This tool:
Extracts text from the PDF
Parses receipt metadata (store, date, totals)
Extracts line items with prices
Categorizes each item using hybrid approach (static rules + LLM)
Stores everything in SQLite database
Args: pdf_path: Absolute path to the PDF receipt file ctx: FastMCP context for logging and LLM access
Returns: Summary of imported receipt including store, date, item count, and category breakdown
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| pdf_path | Yes | Absolute path to PDF receipt file |
Output Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Implementation Reference
- main.py:27-120 (handler)Primary tool handler function. Decorated with @mcp.tool for registration. Implements PDF validation, parsing via pdf_parser, item categorization, database storage, and returns structured summary.
@mcp.tool async def import_receipt_from_pdf( pdf_path: Annotated[str, "Absolute path to PDF receipt file"], ctx: Context, ) -> dict: """Import and parse a receipt from a PDF file. This tool: 1. Extracts text from the PDF 2. Parses receipt metadata (store, date, totals) 3. Extracts line items with prices 4. Categorizes each item using hybrid approach (static rules + LLM) 5. Stores everything in SQLite database Args: pdf_path: Absolute path to the PDF receipt file ctx: FastMCP context for logging and LLM access Returns: Summary of imported receipt including store, date, item count, and category breakdown """ try: await ctx.info(f"Starting import of receipt: {pdf_path}") # Validate path path = Path(pdf_path).expanduser().resolve() if not path.exists(): raise ToolError(f"PDF file not found: {pdf_path}") if not path.suffix.lower() == ".pdf": raise ToolError(f"File must be a PDF: {pdf_path}") # Parse PDF await ctx.info("Extracting text from PDF...") receipt, raw_items = parse_pdf_receipt(path) await ctx.info( f"Parsed receipt: {receipt.store_name} on {receipt.purchase_date}" ) await ctx.info(f"Found {len(raw_items)} line items") # Categorize items await ctx.info("Categorizing items...") categorized_items = [] item_type_counts = {} for idx, item_dict in enumerate(raw_items): # Categorize using hybrid approach item_type = await categorize_item(item_dict["item_name"], ctx) # Create LineItem object line_item = LineItem( item_name_raw=item_dict["item_name"], item_type=item_type, quantity=item_dict["quantity"], line_total=item_dict["price"], ) categorized_items.append(line_item) # Track category counts item_type_counts[item_type] = item_type_counts.get(item_type, 0) + 1 await ctx.debug( f" [{idx+1}/{len(raw_items)}] {item_dict['item_name']} -> {item_type}" ) # Insert into database await ctx.info("Saving to database...") receipt_id = insert_receipt(receipt) insert_items(receipt_id, categorized_items) await ctx.info(f"Successfully imported receipt #{receipt_id}") # Return summary return { "status": "success", "receipt_id": receipt_id, "store_name": receipt.store_name, "purchase_date": receipt.purchase_date, "total": receipt.total, "items_count": len(categorized_items), "item_types": item_type_counts, "message": f"Successfully imported {len(categorized_items)} items from {receipt.store_name}", } except FileNotFoundError as e: raise ToolError(f"File not found: {str(e)}") except ValueError as e: raise ToolError(f"Failed to parse receipt: {str(e)}") except Exception as e: await ctx.error(f"Unexpected error during import: {e}") raise ToolError(f"Failed to import receipt: {str(e)}") - expense_tracker/models.py:7-54 (schema)Dataclass models for Receipt and LineItem providing structured data validation and typing for parsed receipt data used throughout the tool.
@dataclass class Receipt: """Represents a parsed receipt.""" store_name: str purchase_date: str # ISO format: YYYY-MM-DD total: float subtotal: Optional[float] = None tax: Optional[float] = None def __post_init__(self): """Validate receipt data.""" if self.total <= 0: raise ValueError("Total must be positive") @dataclass class LineItem: """Represents a single item from a receipt.""" item_name_raw: str item_type: str line_total: float quantity: float = 1.0 unit_price: Optional[float] = None def __post_init__(self): """Calculate unit price if not provided.""" if self.unit_price is None and self.quantity > 0: self.unit_price = self.line_total / self.quantity if self.line_total <= 0: raise ValueError("Line total must be positive") if self.quantity <= 0: raise ValueError("Quantity must be positive") @dataclass class ItemStats: """Statistics for a specific item type.""" item_type: str total_purchases: int last_purchase_date: str first_purchase_date: str total_spent: float average_days_between: Optional[float] = None - Key helper function that extracts text from PDF using pdfplumber, parses store/date/totals/line items using regex patterns, returns Receipt and raw items list.
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) - Hybrid item categorization helper: static regex/pattern matching first, LLM fallback via ctx.sample for unknown items.
async def categorize_item(item_name: str, ctx=None) -> str: """Main categorization function with hybrid approach. Args: item_name: Raw item name from receipt ctx: Optional FastMCP Context for LLM fallback Returns: item_type category (guaranteed to return a value) """ # Try deterministic rules first category = deterministic_categorize(item_name) if category: return category # Fall back to LLM if context is available if ctx: return await llm_categorize(item_name, ctx) # Ultimate fallback return "other" - expense_tracker/database.py:81-135 (helper)Database persistence helpers: insert_receipt creates receipt record, insert_items adds categorized line items with foreign key.
def insert_receipt(receipt: Receipt, db_path: Path = DEFAULT_DB_PATH) -> int: """Insert a receipt and return its ID.""" conn = get_connection(db_path) try: cursor = conn.execute( """ INSERT INTO receipts (store_name, purchase_date, subtotal, tax, total) VALUES (?, ?, ?, ?, ?) """, ( receipt.store_name, receipt.purchase_date, receipt.subtotal, receipt.tax, receipt.total, ), ) conn.commit() return cursor.lastrowid finally: conn.close() def insert_items( receipt_id: int, items: list[LineItem], db_path: Path = DEFAULT_DB_PATH ) -> None: """Bulk insert items for a receipt.""" if not items: return conn = get_connection(db_path) try: conn.executemany( """ INSERT INTO items (receipt_id, item_name_raw, item_type, quantity, unit_price, line_total) VALUES (?, ?, ?, ?, ?, ?) """, [ ( receipt_id, item.item_name_raw, item.item_type, item.quantity, item.unit_price, item.line_total, ) for item in items ], ) conn.commit() finally: conn.close()