Skip to main content

List Inbox Documents

GET 
/api/document-inbox

Return a unified, paginated, descending-by-created_at list combining customer PO, supplier invoice, vendor credit, and landed cost invoice OCR extractions. Each item is normalized to the same shape so a single mixed-type list can be rendered. Vendor credit extractions that are attached to an existing vendor credit (vendor_credit_attachment_id set) are excluded — they belong to that credit's documents, not the inbox.

Not yet available to API tokens

This endpoint currently requires session authentication; Personal Access Token scope support is in progress.

Example item:

{
  "id": 42,
  "type": "supplier_invoice",
  "status": "pending_review",
  "source": "email",
  "source_email_from": "Jane Doe <jane@tenant.com>",
  "source_email_forwarded_from": "orders-usa@tecmate.com",
  "source_email_subject": "Fwd: Invoice Number-0002020",
  "customer_name": null,
  "supplier_name": "TecMATE",
  "document_type": "invoice",
  "detection_signals": [],
  "auto_routed": false,
  "ocr_header": {
    "doc_number": {"value": "0002020", "confidence": 0.97},
    "total": {"value": 37.06, "confidence": 0.97},
    "tax_amount": {"value": 2.10, "confidence": 0.9},
    "currency_code": {"value": "USD", "confidence": 0.99},
    "doc_date": {"value": "2026-07-17", "confidence": 0.97},
    "po_number": {"value": "PO-0671", "confidence": 0.95}
  },
  "confirmed_record": null,
  "error_message": null,
  "created_at": "2026-07-17T22:28:48.000000Z"
}

Sender fields:

  • source_email_from — the sender of the email that reached the intake address. For forwarded emails this is the forwarder, not the original contact.
  • source_email_forwarded_from — the original sender parsed out of a forwarded email's body/headers (null for direct sends and uploads). Prefer this as the document's real contact when present.

OCR header summary (ocr_header):

  • A compact summary of what OCR extracted from the document, or null while the document is still processing (or when extraction failed before analysis).
  • Each field is {value, confidence}; confidence is 0–1 or null when the extraction pipeline did not score the field.
  • doc_number — the document's own number: invoice number (supplier_invoice, landed_cost_invoice), credit memo number (vendor_credit), or customer PO number (customer_po).
  • total, tax_amount, currency_code — extracted money fields.
  • doc_date — the document's own date (invoice date, credit date, or order date), distinct from created_at (when the document arrived).
  • po_number — for supplier invoices, the purchase order number referenced on the invoice (null for customer POs, where the PO number IS doc_number).

Classification fields:

  • document_type — what the classifier decided the document is: invoice or credit_memo. Populated on supplier_invoice and vendor_credit items; always null for customer_po and landed_cost_invoice.
  • detection_signals — array of signal keys that drove the classification (e.g. keyword:credit memo, doc_number:CM prefix, total:negative). Null when no detection ran.
  • auto_routed — true when the document arrived in the supplier invoice pipeline but was automatically routed to the vendor credit queue by the classifier. Always false for customer_po, supplier_invoice, and landed_cost_invoice items.

When status is 'confirmed', confirmed_record is populated with { id, type, number } of the created sales order, purchase invoice, vendor credit (type: vendor_credit, number: the vendor_credit_number), or landed cost bill (type: landed_cost_invoice).

Landed cost invoice items (type: "landed_cost_invoice") expose supplier_id/supplier_name (the matched supplier, when detected) and always have customer_id, customer_name, document_type, and detection_signals as null with auto_routed false.

Request​

Responses​

Successful response