Invoice OCR

A small, self-contained toolkit to:

• Split multi-invoice PDFs into individual invoice PDF files
• Extract invoice-level metadata (invoice number, date, consignee, dealer)
• Extract and parse tabular item rows from each invoice
• Aggregate and export all items to CSV or Excel

This repository contains two entry points: cli.py (interactive CLI) and ui.py (Tkinter GUI). Core logic lives in custom_modules/.

Contents

• cli.py — interactive command-line processor that prompts for output format and filename.
• ui.py — Tkinter-based GUI with file pickers and progress UI.
• custom_modules/invoice_splitter.py — PDF splitting logic using PyMuPDF.
• custom_modules/dataocr.py — invoice-level extraction (PyMuPDF + bold-span heuristics).
• custom_modules/table_extractor.py — table extraction and column parsing by X coordinate.
• pdfs/ — place source PDFs here (example default used by CLI: pdfs/invoice.pdf).
• individual_invoice/ — default output folder for split PDFs.

Processing steps (full flow)

1. Split PDF into individual invoices

• Module: custom_modules/invoice_splitter.py
• How it detects invoices: looks for the text marker "Tax Invoice" to detect the start of an invoice and "This is a Computer Generated Invoice" to detect the end. When both markers are present the pages between them are saved as a new PDF.
• Output filenames: <originalname>_<batchid>_<count>.pdf (batchid is a short unique id).

2. Extract invoice-level fields

• Module: custom_modules/dataocr.py
• Implementation: uses PyMuPDF (fitz) to walk page text blocks. For specific fields (invoice number, date, consignee, dealer) it looks for known keywords (e.g. Invoice No., Dated, Consignee, Authorised Signatory) and extracts bold spans nearby when available.

3. Extract and parse the items table

• Module: custom_modules/table_extractor.py
• It finds the table region using start/end markers (S.No. and Amount Chargable(in words)) and selects wide text blocks (>80% page width) inside that region.
• It then parses rows by inspecting each line's x0 coordinates and mapping ranges to columns (S.No, Description, HSN, Quantity, Rate, Per/Unit, Discount, Amount). The output per-invoice is a dict like { "items": [ {"items": ..., "Qnty": ..., "price": ..., "unit": ..., "discount": ..., "total": ...}, ... ] }.

4. Aggregate + Export

• Both cli.py and ui.py convert parsed invoice data into row dictionaries with columns:

  VCH_SERIES, SALE/PURC_TYPE, MC_NAME, VCH/BILL_DATE, VCH/BILL_NO, PARTY_NAME, ITEM_NAME, QUANTITY, UNIT, PRICE, DISCOUNT_PERCENT, LIST_PRICE_ALT_UNIT, LIST_PRICE, AMOUNT

• Export uses pandas.DataFrame and supports writing to Excel (.xlsx, via openpyxl) or CSV. Both overwrite and append modes are supported.

Installation

Prerequisites

• Python 3.10+ recommended (project pyproject.toml requires >=3.10).
• On Windows tkinter is typically included; if not, install the platform packages for Tk support.

Install dependencies (recommended)

Option A — install with uv (if you use it for process management):

pip install uv
uv sync

Option B — plain pip (explicit):

pip install pandas pymupdf openpyxl tqdm
# optional: pip install -r requirements.txt

If you want, I can add a requirements.txt listing these packages so you can use pip install -r requirements.txt.

Running the project

CLI (interactive)

python cli.py
# or with uv: uv run cli.py

Behavior

• cli.py default input: pdfs/invoice.pdf (change the input_pdf_file variable in main() to use a different file).
• The CLI prompts for output format (excel or csv) and output filename (default invoice_data).
• If the chosen output file already exists the CLI asks whether to write (overwrite) or append.

GUI (Tkinter)

python ui.py
# or: uv run ui.py

The GUI provides file pickers for the input PDF and output folder, options for output filename, format (Excel / CSV) and write mode (overwrite/append). It shows processing progress and a complete/summary dialog on success.

Examples (PowerShell)

Install deps and run GUI:

pip install pandas pymupdf openpyxl tqdm
python ui.py

Process a PDF and export CSV via CLI:

python cli.py
# choose `csv` when prompted and supply filename

Configuration and tuning

• Change default input or output paths by editing cli.py or by using the GUI.
• If your invoice documents use different start/end markers than the defaults, update them in custom_modules/invoice_splitter.py (search for Tax Invoice and This is a Computer Generated Invoice).
• If parsed table columns are wrong, open custom_modules/table_extractor.py and adjust the x0 ranges used to detect each column. These ranges are specific to your invoice layout and may need calibration.

Troubleshooting

• No data extracted: confirm the PDF contains selectable text. PyMuPDF reads embedded text; scanned images require OCR (e.g., pytesseract) and code changes to dataocr.py.
• Wrong columns / mis-parsed rows: adjust X-coordinate ranges in custom_modules/table_extractor.py and re-run.
• Excel export error: ensure openpyxl is installed.
• GUI doesn't start: ensure tkinter is available in your Python distribution.