Quick start — mailbox to dispatch

Parked — not currently shipped

The Correspondence staff UI has been removed and its API/worker handlers are parked under archive/ pending a rehome. This walkthrough describes the intended arc; it is not reachable in the running app today. For the pipeline internals see Correspondence module.

This is the end-to-end arc for one piece of mail, from the firm's mailbox to a dispatched item. The work surface was /dashboard/correspondence (parked — see the note above).

The arc

1. It arrives. A message with an attachment lands in the firm's mailbox; the poll picks it up, stores the PDF, and creates a correspondence document. (TODO: where to watch new arrivals appear.)
2. OCR & classify. OCR runs (automatically, or via Run OCR), text is extracted, client rows are identified, and the document is split into per-item PDFs. (TODO: reading the extracted items.)
3. Match the client. Each item is matched to a client (optionally via Odoo). Confident matches auto-fill; ambiguous ones wait for you — see Confirm or fix a client match.
4. Review. Check the extraction and the match for each item — see Review a correspondence item. Hold or defer anything that isn't clean.
5. Dispatch. Clean items go out after the grace window, one email per item, honouring the dispatch mode (item_ready or document_complete) — see Dispatch a clean item.

Demo data

pnpm db:seed provisions two demo documents (one completed with two extracted items, one fresh). (TODO: confirm seeded names and expected states.)

Next

• Review a correspondence item
• Handle an exception