Landed Cost Invoice OCR

📄️List Extractions (Inbox)

List landed cost invoice OCR extractions for the inbox queue. Requires Bearer token. Returns a paginated `LengthAwarePaginator` of landed cost invoice OCR extraction object items. All filters are optional. Useful for the OCR Inbox page that surfaces freight/customs/brokerage/insurance invoices awaiting review.

📄️Upload Landed Cost Invoice PDF

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

📄️Get Extraction

Retrieve a landed cost invoice OCR extraction record. Eager-loads the matched supplier and any extraction→bill pivots created during apply. Returns extracted header data merged with matched supplier info, line items with per-line cost-category matches (6-tier match: mapped → exact_name → name_substring → keyword → fuzzy_name → fuzzy_mapped → unmatched), candidate targets (inbound shipments / POs / warehouse transfers that fit the supplier + ETA window), and inbound-email source metadata.

📄️Delete Extraction

Soft-delete a landed cost invoice OCR extraction record (and its associated S3 PDF). Confirmed extractions cannot be deleted (returns 422).

📄️Stream Extraction PDF

Stream the original landed cost invoice PDF through the backend (avoids CORS when fetching from S3). Returns Content-Type: application/pdf, `Content-Disposition: inline; filename='landed-cost-invoice.pdf'`. Returns 404 if the underlying S3 object is missing. Cache-Control: private, max-age=300.

📄️Apply Extraction (Create Bills)

Accept the reviewed/edited extraction and create one or more `Bill` records — one per target entity (inbound shipment, purchase order, or warehouse transfer) — each with its own apportionment of the cost lines. Marks the extraction as `confirmed` and persists `Bill` rows linked via the polymorphic `bills.link_type` / `link_id` columns. Bridges supplier → vendor since Bills require `vendor_id`.

📄️Re-scan Extraction

Re-run Azure Document Intelligence OCR on the original PDF (e.g., after Azure model improvements, or to retry a failed extraction). Async — returns a `tracked_job_log_id` for progress tracking via the tracked job log endpoints.

📄️Save Draft (Working State)

Persist in-progress edits to the review form so users can resume later (across browser refresh / device switch). The `working_state` body is an opaque object stored on the extraction — the frontend defines its shape (typically header_overrides, line_overrides, selected_targets, etc.). Returns the updated extraction.

📄️List Cost Category Keyword Aliases

List active cost-category keyword aliases used by the OCR cost-line matcher. Returns an object keyed by `cost_category_id` with arrays of keyword strings. Used by the Settings page that lets users curate which keywords (e.g., 'freight', 'fuel surcharge', 'customs', 'duty', 'clearance') map to which CostCategory.

📄️Replace Cost Category Keyword Aliases

Replace the full set of keyword aliases for a single cost category. Idempotent — the repository diffs the supplied list against the existing rows: removes keywords not in the new list, inserts new ones. Default-flagged keywords (seeded by migration) are preserved if they remain in the new list, or removed if dropped. Keywords are normalized to lowercase + trimmed; duplicates are deduped server-side.