Native PDF extraction
Reads embedded text per page with PdfPig for fast, accurate processing.
Open-source project · Phase 1
Turn native or scanned PDF invoices into structured, auditable data—entirely on your own infrastructure.
About the product
SalExtract is a deterministic invoice extraction engine designed for on-premise use. It reads embedded PDF text when available, applies OCR locally to scanned or low-text pages, normalizes the results, and stores a traceable audit record without sending document data to the internet.
Processing flow
A transparent pipeline validates each file, extracts content with the appropriate local method, and retains the evidence needed for review.
Validate extension, PDF signature, file size, and page count, then calculate SHA-256 for traceability.
Use PdfPig for embedded text and PDFium with Tesseract (Indonesian and English) for OCR fallback.
Apply deterministic rules to supplier, invoice number, date, total, and currency values.
Return Completed or Review status and persist the full result as JSON in local SQLite.
Phase 1 capabilities
Reads embedded text per page with PdfPig for fast, accurate processing.
Processes scanned pages locally with PDFium rendering and Tesseract OCR.
Extracts supplier name, invoice number, invoice date, total invoice, and currency.
Provides versioned POST and GET endpoints for extraction workflows and integrations.
Records SHA-256, timings, warnings, source method, confidence, page, and evidence.
Includes PDF signature validation, upload limits, rate limiting, safe filenames, and optional API keys.
Explore SalExtract