Email Invoice Auto-Organization

Problem

Expense reporting often slows down on repetitive manual work: locating invoices in email, downloading attachments, reading amounts, matching itineraries, and assembling submission-ready PDFs. This skill targets that workflow. It searches a configured mailbox for invoice-related messages, downloads PDF attachments, and produces a merged A4 PDF together with a single-sheet Excel expense list. For typical electronic invoices, it reads the PDF text layer first and falls back to OCR for scanned documents. The design emphasis is script-driven extraction, parsing, merging, and table generation, reducing manual field copying.

How It Works and Limits

Before processing, it checks mailbox credentials, PyMuPDF, openpyxl, and optional OCR components. It deduplicates by email UID and avoids truncating search results with a small --limit. After downloading, it can unzip archives from 12306 or airline messages, then runs a three-tier extraction path: L1 via PyMuPDF text layer, L2 via Tesseract OCR, and L3 via PaddleOCR. Parsed data is written to JSON files, and generate_excel.py builds a single-sheet Excel list with trip details merged into remarks. PDF merging places invoices and itineraries vertically on A4; 12306 train tickets are scaled to real ticket size and laid out top-to-bottom with a left margin for archival punching.

The scope is bounded: attachments should be PDF, while OFD, XML, and images are not automatically processed. Amount mismatches are marked in Excel remarks, and interrupted runs can resume from saved UIDs and downloaded files.