Purchase Invoice OCR

📄️Upload Supplier Invoice PDF

Upload a supplier invoice PDF to start the OCR extraction. Requires Invoice OCR to be enabled in Settings → Invoice OCR (returns 422 if not enabled).

📄️Get Extraction

Retrieve a purchase invoice OCR extraction record. Returns extracted header data merged with matched supplier and PO info, line items with per-line product matches (5-tier match: supplier_sku → SKU → barcode → MPN → fuzzy name), candidate purchase orders for the supplier (with attach-to-existing-invoice metadata), error_message (if failed/duplicate), and inbound-email source metadata.

📄️Delete Extraction

Delete a purchase invoice OCR extraction record and its associated PDF. Confirmed extractions cannot be deleted (returns 422).

📄️Stream Extraction PDF

Stream the original supplier invoice PDF through the backend (avoids CORS when fetching from S3). Returns Content-Type: application/pdf, inline disposition. 404 if the underlying S3 object is missing.

📄️Create Purchase Invoice from Extraction

Accept the reviewed/edited extraction data (header + lines + cost lines) and create a PurchaseInvoice record, marking the extraction as 'confirmed'. The extraction must be in `pending_review` or `duplicate_detected` status (a duplicate-detected extraction is allowed because the user may confirm it is not actually a duplicate and proceed to create a new invoice).

📄️Attach OCR Document to Existing Invoice

Attach the OCR'd PDF to an existing PurchaseInvoice that does not yet have a document. Used when the OCR extraction's PO number matches a PO that has already been invoiced — instead of creating a new invoice, the user can retroactively attach the extracted PDF to the prior invoice as the supporting document.

📄️Re-scan Extraction

Re-dispatch the OCR extraction job for an existing PurchaseInvoiceOcrExtraction. Resets status to 'processing', clears any previous error_message, and re-queues the ProcessSupplierInvoiceEmailOcrJob against the original PDF in S3. Useful when Azure misread the document (wrong supplier match, missing lines) and the user wants to re-run extraction without re-uploading.

📄️Reclassify as Customer PO

Reclassify a Supplier Invoice OCR extraction as a Customer PO. Used when an inbound email was routed to invoices@ but the attached PDF is actually a customer purchase order — the inverse of POST /api/sales-orders/ocr/{id}/reclassify-as-supplier-invoice.